Research ArticleCancer

Amplification of the Driving Oncogene, KRAS or BRAF, Underpins Acquired Resistance to MEK1/2 Inhibitors in Colorectal Cancer Cells

See allHide authors and affiliations

Science Signaling  29 Mar 2011:
Vol. 4, Issue 166, pp. ra17
DOI: 10.1126/scisignal.2001752

This article has a correction. Please see:

Abstract

The acquisition of resistance to protein kinase inhibitors is a growing problem in cancer treatment. We modeled acquired resistance to the MEK1/2 (mitogen-activated or extracellular signal–regulated protein kinase kinases 1 and 2) inhibitor selumetinib (AZD6244) in colorectal cancer cell lines harboring mutations in BRAF (COLO205 and HT29 lines) or KRAS (HCT116 and LoVo lines). AZD6244-resistant derivatives were refractory to AZD6244-induced cell cycle arrest and death and exhibited a marked increase in ERK1/2 (extracellular signal–regulated kinases 1 and 2) pathway signaling and cyclin D1 abundance when assessed in the absence of inhibitor. Genomic sequencing revealed no acquired mutations in MEK1 or MEK2, the primary target of AZD6244. Rather, resistant lines showed a marked up-regulation of their respective driving oncogenes, BRAF600E or KRAS13D, due to intrachromosomal amplification. Inhibition of BRAF reversed resistance to AZD6244 in COLO205 cells, which suggested that combined inhibition of MEK1/2 and BRAF may reduce the likelihood of acquired resistance in tumors with BRAF600E. Knockdown of KRAS reversed AZD6244 resistance in HCT116 cells as well as reduced the activation of ERK1/2 and protein kinase B; however, the combined inhibition of ERK1/2 and phosphatidylinositol 3-kinase signaling had little effect on AZD6244 resistance, suggesting that additional KRAS effector pathways contribute to this process. Microarray analysis identified increased expression of an 18-gene signature previously identified as reflecting MEK1/2 pathway output in resistant cells. Thus, amplification of the driving oncogene (BRAF600E or KRAS13D) can drive acquired resistance to MEK1/2 inhibitors by increasing signaling through the ERK1/2 pathway. However, up-regulation of KRAS13D leads to activation of multiple KRAS effector pathways, underlining the therapeutic challenge posed by KRAS mutations. These results may have implications for the use of combination therapies.

Introduction

Targeting protein kinases with critical roles in tumor progression or maintenance is an attractive strategy for cancer therapy, and several small-molecule kinase inhibitors have now been approved for clinical use (1). However, the success of these inhibitors has been marred by disease relapse due to acquired drug resistance. There is thus a growing need to understand the molecular mechanisms underlying this resistance in order to design clinical strategies that can prevent, delay, or overcome the acquisition of drug resistance to small-molecule kinase inhibitors.

The protein kinase cascade linking RAF to MEK1/2 (mitogen-activated or extracellular signal–regulated protein kinase kinases 1 and 2) to ERK1/2 (extracellular signal–regulated kinases 1 and 2) (the ERK1/2 pathway) has received much attention in the search for new chemotherapeutic agents. This is because of both the high frequency of KRAS (2) and BRAF (3, 4) mutations identified in certain human cancers and the critical role this pathway plays in promoting cell proliferation (5) and, in some cases, cell survival (6). MEK1/2 occupy pivotal positions within the ERK1/2 pathway, and the presence of a unique inhibitor-binding pocket adjacent to the MEK1/2 adenosine 5′-triphosphate (ATP) binding site allows for their highly selective inhibition by small molecules (7). Furthermore, because MEK1/2 are the only physiologically defined substrates of RAF (4), their inhibition represents a targeted strategy to restrain the increased ERK1/2 signaling critical to the maintenance of many human cancers.

Selumetinib (AZD6244, ARRY-142886), an orally active MEK1/2 inhibitor currently in clinical trials (8, 9), inhibits MEK1 in vitro with an IC50 (half-maximal inhibitory concentration) of 14 nM (8, 10) and exerts antiproliferative and pro-apoptotic effects in various tumor cell lines grown in culture or as xenografts (9, 11). In addition to the ERK1/2 pathway, KRAS can activate multiple effector pathways (12, 13) including the phosphatidylinositol 3-kinase to protein kinase B signaling pathway (the PKB pathway), which can also promote cell proliferation and cell survival. Indeed, some tumor cell lines show intrinsic resistance to AZD6244 resulting from activation of the PKB pathway (11, 14). In addition to intrinsic resistance, it is unclear whether acquired resistance to MEK1/2 inhibitors develops; however, AZD6244’s progression into later-stage clinical trials and the emerging theme of acquired drug resistance render this an important issue.

In addition to common methods of drug resistance (such as drug metabolism or efflux), clinical and preclinical studies have identified two major mechanisms for acquired resistance to oncogene-targeted tyrosine kinase inhibitors (TKIs). The first involves the emergence of mutations in the targeted kinase that abrogate drug binding, such as the “gatekeeper” mutations EGFRT790M [substitution of the epidermal growth factor receptor threonine residue 790 with methionine, in tumors from non–small cell lung cancer (NSCLC) patients with acquired resistance to gefitinib] and BCR-ABLT315I [substitution of the BCR-ABL fusion protein threonine residue 315 with isoleucine, in chronic myelogenous leukemia (CML) patients treated with imatinib]. The second involves a “kinase switch” in which separate oncoproteins are up-regulated to substitute for the targeted driving oncogene; for example, the gene encoding hepatocyte growth factor receptor (MET) is amplified in some patients with acquired resistance to gefitinib, a TKI that specifically targets the EGFR (epidermal growth factor receptor), allowing tumor cells to activate cell survival pathways common to both MET and EGFR signaling [for recent reviews, see (15, 16)]. To date, studies have mainly focused on acquired resistance to drugs that directly target the mutant oncoprotein and bind, at least in part, to the ATP-binding pocket. Gatekeeper mutations in the ATP-binding domain often result in increased kinase activity, providing a rationale for their persistence in cultures during drug selection (13, 14). In contrast, MEK1/2 mutations are extremely rare in cancer (17) and do not consistently increase MEK1/2 activity (18), raising questions of whether acquired resistance to AZD6244 will arise and, if so, how. Moreover, differences in genetic background such as p53 status, chromosome stability, or the driving oncogene may promote different mechanisms of acquired resistance; for example, a distinct mechanism could arise in tumors with KRASMut compared to those with BRAFMut as a result of the broader repertoire of KRAS signaling (19).

To investigate these questions, we generated AZD6244-resistant derivatives of human colorectal cancer (CRC) cell lines that harbor single activating alleles encoding BRAF600E (a mutant form of BRAF in which valine 600 is substituted with glutamate; COLO205 and HT29 parental lines) or KRAS13D (a mutant form of KRAS in which glycine is substituted with aspartate; HCT116 and LoVo parental lines). We found that AZD6244-resistant cells showed amplification of their respective driving oncogene, were resistant to AZD6244-induced cell cycle arrest and cell death, and exhibited a pronounced hyperactivation of the ERK1/2 pathway. Inhibition of BRAF upstream of MEK1/2 confirmed that increased ERK1/2 pathway activation is critical in conferring resistance to AZD6244 in COLO205 cells. In contrast, AZD6244-resistant HCT116 cells exhibited hyperactivation of both the ERK1/2 and the PKB pathways, but their combined inhibition was not sufficient to restore sensitivity, which suggests that an additional KRAS effector pathway may contribute to AZD6244 resistance. These results have implications for the use of AZD6244 in combination with other therapies.

Results

AZD6244-resistant derivatives of COLO205 and HCT116 cells are refractory to AZD6244-induced cell cycle arrest and cell death

Treatment of NSCLC cells with gefitinib or CML cells with imatinib in vitro yields drug-resistant cells with the same gatekeeper mutations observed in patients with clinically acquired drug resistance [reviewed in (15, 16)]. Thus, generation of resistance in vitro can faithfully recapitulate the mechanisms of acquired resistance observed in the clinic. To study acquired resistance to AZD6244, we used the COLO205 and HCT116 CRC cell lines. These two lines are both sensitive to AZD6244 in proliferation assays (11) but harbor different driver oncogenes that activate ERK1/2; COLO205 cells are heterozygous for BRAF600E, whereas HCT116 cells are heterozygous for KRAS13D.

COLO205 and HCT116 cells were grown in the presence of increasing concentrations of AZD6244 without clonal selection until they grew apparently normally in 1 μM (C6244-R1 cells) or 2 μM drug (H6244-R1 cells). These concentrations are ~10 times the IC50 for inhibition of proliferation of the parental cells (<100 and ~200 nM for COLO205 and HCT116 cells, respectively) and pharmacokinetic studies in mice and humans indicate that 2 μM is approaching the maximum achievable and tolerable dose in vivo (20). The IC50 for AZD6244-induced inhibition of proliferation in C6244-R1 cells was 20 times that of the parental COLO205 cells, whereas even 10 μM AZD6244 failed to inhibit H6244-R1 cell proliferation by 50% (Fig. 1, A and B). C6244-R1 and H6244-R1 cells proliferated at rates comparable to those of their parental counterparts (fig. S1, A and B) and showed cross resistance to the MEK1/2-selective inhibitor PD184352 (fig. S2, A and B); however, cross resistance to the pan-MEK inhibitor U0126 was less apparent (fig. S2, C and D), which may reflect off-target effects of this drug (21, 22). At a concentration of 10 μM, AZD6244 inhibited colony formation of HCT116 cells by ~85% but had little effect on the number of colonies formed by H6244-R1 cells (Fig. 1C). Similar results were seen with COLO205 and C6244-R1 cells grown in the presence of 1 μM AZD6244 (Fig. 1D). Multidrug resistance (MDR) transporters frequently contribute to acquired resistance to conventional chemotherapeutic agents. However, H6244-R1 and C6244-R1 cells retained normal sensitivity to representatives of various classes of cytotoxic drugs, including microtubule disrupters (paclitaxel), anti-metabolites (5-fluorouracil), and DNA-damaging agents (cisplatin), that are known MDR substrates (fig. S3), arguing against a role for MDR in AZD6244 resistance in these cells.

Fig. 1

H6244-R1 and C6244-R1 cells are resistant to the antiproliferative effects of AZD6244. (A and B) Subconfluent cultures of (A) HCT116 and H6244-R1 cells or (B) COLO205 and C6244-R1 cells were treated with increasing concentrations of AZD6244 (10 nM to 10 μM) for 24 hours, and DNA synthesis was assayed by [3H]thymidine incorporation; data points represent means ± CV (coefficient of variation) of biological triplicates and are taken from a single experiment representative of three giving similar results. (C and D) HCT116 and H6244-R1 cells (C) or COLO205 and C6244-R1 cells (D) were treated with AZD6244, as indicated, and their ability to grow in colony-forming assays was assessed after 2 weeks in culture. The mean number of colonies formed ± SD (right panel) and photographic images (left panel) from a representative experiment of three are shown.

Treatment of parental COLO205 and HCT116 cells with 100 nM AZD6244 resulted in G1 arrest, whereas higher concentrations (1 and 10 μM) caused cell death (cells with sub-G1 DNA). In contrast, 1 μM AZD6244 had little effect on the cell cycle profile of C6244-R1 and H6244-R1 cells and, although 10 μM AZD6244 induced G1 arrest in both H6244-R1 and C6244-R1 cells, there was no substantial increase in cell death (Fig. 2, A and B). Thus, C6244-R1 and H6244-R1 cells are resistant to both the antiproliferative and the pro-death effects of AZD6244. AZD6244 resistance emerged more rapidly in HCT116 cells than COLO205 cells (fig. S4). Notwithstanding the modest differences in the proliferation rates of these cell lines, which would tend to favor this result, this may reflect the strong dependence of COLO205 cells on the ERK1/2 pathway and the possible ability of the HCT116 cells to use other KRAS effector pathways to offset the antiproliferative effects of AZD6244. Indeed, AZD6244-induced G1 arrest in parental COLO205 cells was more profound than that in parental HCT116 cells (Fig. 2, A and B) and so might act as a greater barrier to the emergence of resistance.

Fig. 2

H6244-R1 and C6244-R1 are resistant to AZD6244-mediated cell cycle arrest and cell death and exhibit increased basal MEK1/2-ERK1/2 activation and ERK1/2 pathway output. (A and B) After treatment for 48 hours with the indicated concentrations of AZD6244, subconfluent cultures of (A) HCT116 and H6244-R1 cells or (B) COLO205 and C6244-R1 cells were harvested and stained with propidium iodide, and their cell cycle distribution was assessed by flow cytometry. (C and D) Cyclin D1 (Cyc D1), p27KIP1, P(S795)-RB, P-MEK1/2, total MEK1/2, P-ERK1/2, total ERK1, FRA-1, FRA-2, and c-JUN levels were determined by Western blot analysis of whole-cell extracts from cells treated with vehicle control [dimethyl sulfoxide (DMSO)] or AZD6244 for 24 hours. All data are taken from a single experiment representative of three giving similar results.

ERK1/2 signaling promotes cell cycle progression by increasing the abundance of D-type cyclins and decreasing that of the cyclin-dependent kinase inhibitor (CDKI) p27KIP1, thereby activating the CDK4 and cyclin D (CDK4–cyclin D) and CDK2–cyclin E complexes and promoting phosphorylation of the retinoblastoma protein (RB). Indeed, AZD6244 decreased cyclin D1 abundance, increased p27KIP1 abundance, and promoted dephosphorylation of RB at serine residue 795 in parental cells (Fig. 2C). In contrast, H6244-R1 and C6244-R1 cells were resistant to AZD6244-mediated changes in these key cell cycle regulators and, in the case of C6244-R1 cells, exhibited strong, constitutive phosphorylation of RB. Consistent with this, both C6244-R1 and H6244-R1 cells showed enhanced growth factor–independent proliferation when serum-starved compared to their parental cells (fig. S1C).

H6244-R1 and C6244-R1 cells exhibit increased ERK1/2 pathway signaling

Because AZD6244-resistant cells showed increased abundance of cyclin D1 (encoded by an ERK1/2 target gene), we examined the phosphorylation status of components of the ERK1/2 pathway. In parental HCT116 and COLO205 cells, basal levels of phosphorylated ERK1/2 (P-ERK1/2) were maintained by low levels of the active phosphorylated form of MEK1/2 (P-MEK1/2) (Fig. 2D), and treatment with AZD6244 caused a decrease in P-ERK1/2. When maintained for 24 hours in the absence of drug, H6244-R1 and C6244-R1 cells showed a marked increase in P-MEK1/2 and P-ERK1/2 relative to their parent cells (Fig. 2D). The abundance of P-ERK1/2 in H6244-R1 or C6244-R1 cells under their routine culture conditions (in the presence of 2 or 1 μM AZD6244, respectively) was comparable to that in the corresponding parental cells under their routine culture conditions (in the absence of AZD6244) (Fig. 2D). This suggests that the resistant cells adapted to MEK inhibition by increasing flux through the ERK1/2 pathway to maintain P-ERK1/2 at levels comparable to those in parental cells; it also implies that this amount of P-ERK1/2 reflects the pathway throughput required to drive cell proliferation. When placed in drug-free medium, AZD6244-resistant cells showed increased abundance of FRA-1, FRA-2, and c-JUN (Fig. 2D), all of which are targets of the ERK1/2 pathway (23, 24), indicating that the increase in flux through the ERK1/2 pathway was functionally relevant.

Although H6244-R1 and C6244-R1 cells placed in drug-free medium showed increased P-ERK1/2 compared to the parental lines, P-ERK1/2 abundance remained sensitive to AZD6244 (Fig. 2D); this finding suggested that MEK1/2 were still susceptible to AZD6244 inhibition and argued against the acquisition of drug-resistant MEK1/2 mutations. Indeed, genomic sequencing of parental and drug-resistant cells failed to detect any mutations in MEK1 and only one synonymous mutation in MEK2 (C192T, encoding a valine at residue 64) present in both HCT116 and H6244-R1 cells. Inhibition of MEK1/2 by AZD6244 also increased P-MEK1/2 in HCT116 cells (Fig. 2D); this likely reflects disruption of an ERK1/2-driven feedback inhibitory loop that limits RAF activity (25, 26). The AZD6244-induced increase in P-MEK1/2 was not seen in H6244-R1 cells and this, together with the increased abundance of P-MEK1/2 in the absence of AZD6244, suggests that this feedback inhibitory loop is disabled in H6244-R1 cells. AZD6244 did not increase P-MEK1/2 in COLO205 cells, consistent with reports that the BRAF600E mutation uncouples this ERK1/2-driven feedback loop (27, 28).

Amplification of BRAF600E underlies AZD6244 resistance in COLO205 cells

The increase in P-MEK1/2 abundance in C6244-R1 cells and the absence of MEK1/2 mutations suggested that the change driving acquired resistance to AZD6244 occurred upstream of MEK1/2. Accordingly, we examined the abundance of upstream signaling components. Indeed, C6244-R1 derived from COLO205 cells (BRAF600E) showed a marked increase in BRAF abundance (Fig. 3A); however, there was no accompanying increase in KRAS or CRAF (Fig. 3A and fig. S10C). Quantitative polymerase chain reaction (qPCR) analysis of genomic DNA from COLO205 and C6244-R1 cells revealed a substantial increase in BRAF in C6244-R1 cells relative to the controls, consistent with a factor of 4 to 5 increase in copy number (fig. S5A and Table 1). In contrast, no amplification of BRAF was observed in H6244-R1 cells when compared to HCT116 cells, and neither KRAS nor CCND1 was amplified in C6244-R1 cells (fig. S5, A to C). Sequencing of BRAF exon 15 revealed that, whereas the parental COLO205 cells were heterozygous for the BRAFV600E mutation, C6244-R1 cells appeared to be homozygous for the mutant allele (fig. S6, A and B); no mutations in KRAS or NRAS were observed in either COLO205 or C6244-R1 cells.

Fig. 3

Resistance to AZD6244 in C6244-R1 cells is driven by an intrachromosomal amplification of the BRAF locus leading to increased BRAF abundance. (A) Whole-cell lysates from cultures treated with vehicle control (DMSO) or AZD6244 for 24 hours were analyzed by Western blotting for KRAS and BRAF abundance. (B and C) BRAF locus BAC DNA (RP11-1173P7; green) and chromosome 7 paint (blue) were hybridized to the (B) interphase nuclei and (C) metaphase spreads of COLO205 and C6244-R1 cells [gray, DAPI (4′,6-diamidino-2-phenylindole) stain]. (D and E) COLO205 (C) and C6244-R1 (CR) cells were transfected with BRAF-specific (siBRAF) or scrambled (siScram) siRNA oligos. Control cells (CNT) were left untransfected. (D) Seventy-two hours after transfection, BRAF, P-ERK1/2, and ERK1 abundances were determined by Western blot analysis of whole-cell extracts. (E) Forty-eight hours after transfection, cells were treated with AZD6244 for a further 24 hours, and DNA synthesis was assayed by [3H]thymidine incorporation. Where relevant, errors represent the mean ± CV of biological triplicates; all data are taken from a single experiment representative of three giving similar results.

Table 1

Summary of analysis of KRAS and BRAF gene sequence and copy number in parental and AZD6244-resistant cell lines. Mutations identified in KRAS are relative to National Center for Biotechnology Information NM_004985 and NP_004976. Mutations identified in BRAF are relative to NM_004333.3 and NP_004324.2. Fold increase in copy number is relative to healthy volunteers. All of the samples were wild type in the remainder of the screened exons. wt, wild type; het, heterozygous mutant; hom, homozygous mutant.

View this table:

To investigate the mechanism underlying BRAF amplification, we carried out fluorescence in situ hybridization (FISH) with a bacterial artificial chromosome (BAC) clone from the BRAF genomic locus. We examined both metaphase and interphase nuclei, because chromosomes are visible in metaphase, whereas chromatin is less condensed in interphase, and individual copies of an amplified locus can be distinguished to some extent. COLO205 cells contained three apparently normal copies of chromosome 7 (29), each with a single FISH signal from the BRAF BAC, whereas, in C6244-R1 cells, all interphase nuclei showed around 10 signals, consistent with BRAF amplification (Fig. 3, B and C). We observed two populations of C6244-R1 cells by metaphase FISH, both of which showed BRAF amplification (Fig. 3C). The major population (70% of cells) contained two apparently normal copies of chromosome 7 and two copies bearing a focal (likely in situ) amplification of BRAF on 7q, seen as a signal of consistently increased intensity and area. The minor population (30%) had two normal copies of chromosome 7 and a large derivative chromosome 7 that had two regions of the amplified BRAF locus and had undergone a further translocation with an unidentified chromosome. We speculate that the derivative chromosome 7 may have resulted from a fusion of the two BRAF-amplified chromosome 7s from the major population. In any case, these data indicate that the increase in BRAF abundance in C6244-R1 cells represents gene amplification, suggesting that the more abundant BRAF in these cells consists entirely of mutant BRAF600E.

We used RNA interference (RNAi) to determine whether the increase in BRAF600E could account for AZD6244 resistance in C6244-R1 cells. BRAF knockdown with short interfering RNA (siRNA) markedly attenuated the increased ERK1/2 phosphorylation in C6244-R1 cells (Fig. 3D) and also resensitized them to the antiproliferative effects of AZD6244 (Fig. 3E), whereas scrambled siRNAs failed to do so. Therefore, we conclude that increased BRAF abundance is required to maintain acquired resistance to AZD6244 in C6244-R1 cells.

Amplification of KRAS13D underlies AZD6244 resistance in HCT116 cells

In contrast to C6244-R1 cells, H6244-R1 cells derived from HCT116 (KRAS13D) showed a substantial increase in KRAS abundance but no change in that of BRAF (Fig. 4A). The increase in total KRAS in H6244-R1 cells was accompanied by an increase in the abundance of active guanosine triphosphate (GTP)–bound KRAS (Fig. 4B). qPCR analysis of genomic DNA revealed a factor of 9 to 10 increase in KRAS gene copy number in H6244-R1 cells compared to DNA from healthy volunteers (fig. S5B), consistent with a copy number of >16 (Table 1). No amplification of KRAS was observed in C6244-R1 cells nor of BRAF in H6244-R1 cells (fig. S5, A and B). Sequencing of exon 1 of KRAS confirmed the presence of a single G13D allele in parental HCT116 cells, whereas H6244-R1 cells were apparently homozygous for the mutant allele (fig. S6, C and D); no other mutations in KRAS, NRAS, or BRAF were observed in either HCT116 or H6244-R1 cells. FISH with a BAC clone from the KRAS genomic locus revealed that KRAS was amplified in H6244-R1 cells but not in the parent HCT116 cell line (Fig. 4, C and D). HCT116 cells contained two apparently normal copies of chromosome 12, each with a single FISH signal from the KRAS gene. In H6244-R1 cells, one copy of chromosome 12 had undergone translocation with an undetermined partner chromosome, and an amplification of KRAS was visible at or near the translocation junction (Fig. 4C). This amplification signal (thick arrow in Fig. 4Cii) was clearly distinct from a faint pair of FISH signals (thin arrow in Fig. 4Cii) that probably represent the normal KRAS locus. This distinctive derivative chromosome was present in all metaphase spreads examined, and a focal amplification of about 20 to 30 copies of the KRAS locus was visible in all interphase nuclei (Fig. 4D). Thus, the increase in KRAS abundance in H6244-R1 cells represents gene amplification, suggesting that the increased KRAS consists entirely of the mutant KRAS13D oncoprotein.

Fig. 4

Resistance to AZD6244 in H6244-R1 cells is driven by an intrachromosomal amplification of the KRAS locus leading to increased KRAS abundance. (A) Whole-cell lysates from cultures treated with vehicle control (DMSO) or AZD6244 for 24 hours were analyzed by Western blotting for abundance of KRAS and BRAF. (B) Active KRAS (KRAS-GTP) in HCT116 and H6244-R1 cells was assessed by GST-RAF pull-down and compared with total KRAS, P-ERK1/2, and ERK1 in the corresponding whole-cell lysates by Western blot analysis. (C and D) KRAS locus BAC DNA (RP11-707G1; green) and chromosome 12 paint (blue) were hybridized to the (C) metaphase spreads and (D) interphase nuclei of HCT116 and H6244-R1 cells (gray, DAPI). (E and F) HCT116 and H6244-R1 cells were transfected with KRAS-specific (K) or scrambled (Scr) siRNA oligos. Control cells were left untransfected (UT). (E) Seventy-two hours after transfection, KRAS, P-MEK1/2, P-ERK1/2, P(T308)-PKB, P(S473)-PKB, and Hsp90 levels were determined by Western blot analysis of whole-cell extracts. (F) Forth-eight hours after transfection, cells were treated with AZD6244 for a further 24 hours, and DNA synthesis was assayed by [3H]thymidine incorporation. Where relevant, errors represent the mean ± CV of biological triplicates; all data are taken from a single experiment representative of three giving similar results.

We used RNAi to determine whether the increase in KRAS13D could account for AZD6244 resistance in H6244-R1 cells. H6244-R1 cells showed increased PKB activation (indicated by increased phosphorylation on threonine residue 308 and serine 473) as well as increased ERK1/2 phosphorylation (Fig. 4E), consistent with activation of multiple effector pathways by the amplified KRAS13D; this was not seen in C6244-R1 cells, consistent with the fact that BRAF does not couple to PI3K signaling (fig. S7). KRAS knockdown with siRNA inhibited the increase in ERK1/2 and PKB phosphorylation in H6244-R1 cells (Fig. 4E) and resensitized these cells to the antiproliferative effects of AZD6244, whereas scrambled siRNAs failed to do so (Fig. 4F). Therefore, we conclude that increased KRAS abundance is required to maintain acquired resistance to AZD6244 in H6244-R1 cells.

Small-molecule inhibition of RAF reverses AZD6244 resistance in C6244-R1 cells, whereas H6244-R1 cells are resistant to combined inhibition of the ERK and PI3K signaling pathways

Ideally, the identification of the mechanisms underlying acquired resistance should enable rational intervention with small-molecule inhibitors to overcome resistance. To this end, we used a 100 nM concentration of the pan-RAF inhibitor AZ628 (30) to “reset” ERK1/2 signaling in C6244-R1 cells to the approximate level characteristic of parental COLO205 cells under basal conditions; this reduced the increases in the abundances of FRA-1 and cyclin D1 (Fig. 5A). Using a [3H]thymidine incorporation assay, we found that inhibition of RAF with 100 nM AZ628 completely resensitized C6244-R1 cells to the antiproliferative effects of AZD6244 (Fig. 5B). Because small-molecule inhibitors of KRAS are not available, we tried to overcome the effects of KRAS amplification with combined inhibition of the ERK1/2 and PI3K pathways—two RAS effector pathways whose coincident activation can confer AZD6244 resistance (11, 14). A combination of AZD6244 and the combined PI3K and mTOR (mammalian target of rapamycin) inhibitor AZ12321046 completely reversed hyperactivation of the ERK1/2 and PKB pathways in H6244-R1 cells (Fig. 5C). However, despite this, cyclin D1 abundance remained increased, p27KIP1 abundance remained decreased, and both were nonresponsive to the combination of AZD6244 and AZ12321046 (Fig. 5C). Consistent with this, PI3K inhibition showed little evidence of resensitizing H6244-R1 cells to the antiproliferative effects of AZD6244 and had similar effects on the proliferation of the H6244-R1 and HCT116 cells (Fig. 5D). Notably, whereas cyclin D1 overexpression can confer resistance to BRAF inhibitors (31), the persistent increase in cyclin D1 abundance under these conditions was not due to amplification of the CCND1 gene (fig. S5C). Therefore, combined inhibition of the ERK1/2 and PI3K pathways was unable to overcome the acquired resistance to AZD6244 associated with increased KRAS13D abundance in H6244-R1 cells.

Fig. 5

RAF inhibition overcomes acquired resistance to AZD6244 in C6244-R1 cells, but combined inhibition of MEK1/2 and PI3K fails to overcome acquired resistance to AZD6244 in H6244-R1 cells. (A) Whole-cell lysates from C6244-R1 cultures treated with vehicle control (DMSO) or the indicated concentrations of RAF inhibitor AZ628 for 24 hours were analyzed by Western blotting for BRAF, P-MEK1/2, P-ERK1/2, FRA-1, P-S6K, cyclin D1 (Cyc D1), and ERK1. Whole-cell lysate from parental COLO205 cells treated with DMSO (COLO) was used as a point of reference. (B) Subconfluent cultures of COLO205 or C6244-R1 cells were treated with AZD6244 (10 nM to 10 μM) either alone or in combination with 100 nM AZ628 for 24 hours, and DNA synthesis was assayed by [3H]thymidine incorporation. (C and D) Subconfluent cultures of HCT116 or H6244-R1 cells were treated with increasing concentrations of AZD6244 as indicated, either alone or in combination with 200 nM AZ12321046 for 24 hours. (C) Whole-cell lysates were analyzed by Western blotting for P-ERK1/2, ERK1, P(S473)-PKB, P(T308)-PKB, P(T389)-S6K, cyclin D1 (Cyc D1), and p27KIP1. (D) DNA synthesis was assayed by [3H]thymidine incorporation; errors represent the mean ± CV of biological triplicates from a single experiment representative of three.

Cells with acquired resistance to AZD6244 exhibit increased expression of MEK1/2 pathway output gene signatures

Changes in gene expression that reflect signal pathway output may serve as pharmacodynamic biomarkers (changing upon drug treatment) and may be predictive of drug response. As an alternative measure of target pathway activity, we conducted genome-wide mRNA expression analysis using Human Illumina bead arrays and compared the changes in gene expression we observed to two published signatures representing MEK1/2 pathway output (“MEK,” 18 genes) and RAS-mediated compensatory signaling upstream of RAF (“CRes,” 13 genes) predictive of response to MEK inhibition by AZD6244 (32). Consistent with our biochemical analysis, we found an overall increase in expression of genes associated with the MEK1/2 output signature in both C6244-R1 and H6244-R1 cells as compared with their parental cell lines (increased expression of 11 of 18 and 8 of 18 genes, respectively), confirming an increase in MEK1/2 pathway output (Fig. 6, A and B). In addition, although only 5 of the 13 CRes signature genes mapped to the Illumina platform, 3 of these showed significantly increased expression compared to the parental line in H6244-R1 cells but not in C6244-R1 cells (Fig. 6, A and B). This suggests an increase in signaling from RAS in H6244-R1 cells, again corroborating our biochemical findings. Because we used the Illumina platform, these results also demonstrate the biological robustness of these gene signatures, first defined in studies using Affymetrix and Agilent platforms (32). Indeed, both H6244-R1 and C6244-R1 demonstrated significant differential regulation of genes within an independent signature of RAS pathway activity and dependence [147 genes (33)], with both cell lines scoring particularly well for genes known to be transcriptionally activated by the ERK1/2 pathway (fig. S8). These gene signatures may be useful in identifying and characterizing emerging clinical resistance to MEK1/2 inhibitors.

Fig. 6

A gene signature for MEK1/2 pathway output is up-regulated in cells with acquired resistance to AZD6244. RNA was extracted from lysates of cells incubated in the absence of AZD6244 for 24 hours, and gene expression profiles were examined with Human-6 Illumina Bead Arrays. Gene expression profiles were measured in triplicate for each cell line and averaged; results are expressed as the fold change in expression of resistant versus parental cells (heat map values: red, increased expression in resistant cells; blue, decreased expression). (A) Heat map of differential expression for the 18-gene “MEK functional activation” signature and the 13-gene “compensatory resistance” signature in (H) HCT116 versus H6244-R1 cells and (C) COLO205 versus C6244-R1 cells. Genes are ordered left to right by performance predicting preclinical MEK inhibition and clinical genotype in independent colorectal samples. (B) Plots of mean expression of MEK functional activation and compensatory signaling signature (weighted relative to the predictive performance specifically within colorectal samples). (C) Heat map of differential expression of genes surrounding BRAF and KRAS loci in (H) HCT116 versus H6244-R1 cells and (C) COLO205 versus C6244-R1 cells.

Analysis of the Illumina gene expression array data from H6244-R1 cells highlighted a clustering of significantly up-regulated genes surrounding the KRAS genomic locus at 12p12 (P < 3.6 × 10−8) (Fig. 6C). A similar enrichment of increased gene expression was identified around the BRAF genomic locus at 7q34 (P < 0.001) in C6244-R1 cells (Fig. 6C). This suggests that, in C6244-R1 or H6244-R1 cells, the respective oncogene amplification occurs within an amplicon containing neighboring genes. However, because siRNA-mediated knockdown of KRAS in H6244-R1 cells or BRAF in C6244-R1 cells resensitized cells to the antiproliferative effects of AZD6244 (Figs. 3E and 4F), these co-amplified genes do not contribute directly to AZD6244 resistance.

Amplification of BRAF and resistance to AZD6244 in COLO205 cells are reversible upon drug withdrawal

EGFRT790M amplification and acquired resistance to a second-generation irreversible EGFR inhibitor are reversible after drug withdrawal (34). To test the stability of resistance to AZD6244, we maintained C6244-R1 cells (no clonal selection) and two cell lines derived by single-cell cloning of the C6244-R1 population (C6244-R1 C1 and C6244-R1 C2) in the absence of AZD6244 for 20 weeks and assessed their sensitivity to the antiproliferative effects of AZD6244 (fig. S9A). We found that resistance to AZD6244 was reversible, although the three cell lines reverted with different kinetics; the IC50 for C6244-R1 cells was similar to that for COLO205 cells within 10 weeks. Although C6244-R1 C1 cells showed full reversion to parental-like sensitivity after 20 weeks in the absence of drug, C6244-R1 C2 cells showed a partial resensitization to AZD6244 but had not reverted fully. Resensitization of C6244-R1, C6244-R1 C1, and C6244-R1 C2 cells to the antiproliferative effects of AZD6244 correlated with decreased BRAF abundance and ERK1/2 phosphorylation (fig. S9B).

Up-regulation of the driving oncogene is a common mechanism of acquired resistance to AZD6244

In addition to dose escalation, drug-resistant cell lines can be generated by delivery of solely the maximum drug concentration. We treated parental COLO205 or HCT116 cells with 1 or 2 μM AZD6244, respectively, until the emergence of a cell population able to sustain growth under these conditions. The corresponding cell lines (H6244-R2 and C6244-R2) showed a similar or greater degree of AZD6244 resistance compared to the R1 cells (fig. S10, A and B); in addition, H6244-R2 showed a comparable increase in KRAS abundance, ERK1/2 phosphorylation, and PKB phosphorylation to that of H6244-R1 and C6244-R2 cells showed an increase in BRAF abundance and ERK1/2 phosphorylation comparable to that of C6244-R1 cells (fig. S10, C and D). We also generated HCT116 cells with acquired resistance to a separate MEK1/2-selective inhibitor, PD184352, and observed a comparable increase in KRAS abundance, ERK1/2 signaling, and PKB signaling (fig. S11). Finally, to determine whether increased abundance of the driving oncoprotein was restricted to COLO205 or HCT116 cells or was potentially a common mechanism of acquired resistance to AZD6244, we generated resistant derivatives of HT29 (BRAF600E) and LoVo (KRAS13D) cells. AZD6244-resistant HT29 cells (HT6244-R cells) exhibited increased BRAF abundance, hyperactivation of the ERK/12 pathway (which remained responsive to inhibition by AZD6244), and changes in the abundance of cyclin D1 and p27KIP1 (Fig. 7, A and B) as in the C6244-R1 cells. Likewise, AZD6244-resistant LoVo cells (L6244-R cells) showed increased KRAS abundance, hyperactivation of both the ERK1/2 and the PKB pathways, and changes in the abundance of cyclin D1 and p27KIP1 in comparison to parental LoVo cells (Fig. 7, C and D), replicating our results with H6244-R1 cells. Therefore, amplification of the driving oncogene is a common mechanism of acquired resistance to different MEK1/2 inhibitors regardless of the treatment regimen used to generate resistance.

Fig. 7

Up-regulation of the driving oncogene is a common mechanism of acquired resistance to AZD6244. (A to D) AZD6244-resistant derivatives of HT29 (A and B) or LoVo (C and D) were generated by passaging the parental cells in 1 or 4 μM AZD6244, respectively. Subconfluent cultures of (A) HT29 or HT6244-R or (C) LoVo or L6244-R cells were treated with increasing concentrations of AZD6244 (10 nM to 10 μM) for 24 hours, and DNA synthesis was assayed by [3H]thymidine incorporation; errors represent the mean ± CV of biological triplicates. (B and D) Whole-cell lysates from cultures treated with vehicle control (DMSO) or AZD6244 for 24 hours were analyzed by Western blotting for KRAS, BRAF, P-MEK1/2, MEK1/2, P-ERK1/2, ERK1, P(S473)-PKB, PKB, p27, and cyclin D1 (Cyc D1).

Discussion

Inhibition of MEK1/2 remains an attractive treatment strategy for tumors addicted to ERK1/2 signaling, although combination therapy is likely to be required to fully exploit its therapeutic effect. The progression of AZD6244 into later-stage clinical trials makes it important to define potential mechanisms of acquired resistance; we therefore modeled acquired resistance to AZD6244 using an in vitro cell culture approach. The strength of this approach is demonstrated by the observation that growth of NSCLC or CML cell lines in the presence of gefitinib or imatinib results in resistant cells with the same gatekeeper mutations found in biopsies from tumors of patients undergoing treatment with these drugs. Furthermore, amplification of MET, a clinically relevant mechanism of gefinitib resistance, was first identified with an in vitro culture method (35). We used four CRC cell lines with differences in driving oncogene, chromosome stability, and p53 status (table S3). Sequencing of C6244-R1 and H6244-R1 cells ruled out mutations in MEK1 and MEK2 as a resistance mechanism, and, although it was entirely conceivable that different mechanisms of resistance could have arisen depending on the genetic background of the cells, we identified a common mechanism involving amplification of the driving oncogenes, BRAF or KRAS. While this manuscript was undergoing revision, another study demonstrated BRAF amplification as a mechanism of resistance to AZD6244 in two other CRC cell lines harboring BRAF600E mutations (36), corroborating these findings.

The consequences of the observed gene amplification and the implications for therapeutic intervention depend in large part on the nature of the driving oncogene and the pathways it activates. Tumor cells typically adapt to be dependent on their specific driving oncogene and the pathways it regulates; this is termed oncogene addiction and pathway addiction. Despite substantial overexpression of BRAF600E in C6244-R1 cells, the fidelity of downstream signaling remained intact; there was an increase in ERK1/2 signaling and in MEK1/2 output gene signature, but no increase in PKB activity. These results suggest that COLO205 cells are addicted to a specific level of ERK1/2 signaling for proliferation, and chronic exposure to AZD6244 causes them to remodel the pathway by increasing BRAF600E abundance and thereby maintaining flux through the ERK1/2 pathway in the presence of drug. This allows C6244-R1 cells to maintain a level of ERK1/2 activity comparable to that of the parental cells even in the presence of drug. There are two clear implications of these results. First, for tumors with BRAF600E, the combined inhibition of MEK1/2 and BRAF may be more efficacious than that of inhibiting either alone and could reduce the incidence of acquired resistance to MEK1/2 or BRAF inhibitors. Second, a rational combination of inhibitors that shifts the phenotypic outcome to cell death would mitigate the risk of acquired resistance due to pathway up-regulation.

The increase in KRAS13D abundance in H6244-R1 cells led to increased phosphorylation of MEK1/2 and ERK1/2, increased expression of the MEK output gene signature, increased abundance of cyclin D1, and decreased abundance of p27KIP1. H6244-R1 cells also showed a marked PI3K-dependent increase in P-PKB and P-S6K and an increase in expression of a subset of genes from a defined “compensatory-resistance” gene expression signature (32). These cells also showed increased expression of a subset of genes from the recently defined RAS pathway gene signature (33). Coincident activation of the ERK1/2 and PKB pathways is associated with intrinsic resistance to MEK1/2 inhibitors including AZD6244 (11, 14), in part because the two pathways are redundant for the regulation of cyclin D1 and p27KIP1 expression; indeed, overexpression of cyclin D1 can confer resistance to BRAF inhibitors (31). RNAi-mediated knockdown of KRAS decreased ERK1/2 and PI3K-PKB signaling and resensitized H6244-R1 cells to AZD6244; however, combined inhibition of these pathways (AZD6244 + PI3K and mTOR inhibition) did not reverse the increase in cyclin D1 abundance or restore growth inhibition. Genomic qPCR eliminated the possibility that H6244-R1 cells harbor amplification of the gene encoding cyclin D1 (fig. S5C); thus, it is possible that the persistent increase in cyclin D1 abundance is due to hyperactivation of a further KRAS13D effector pathway, although we cannot exclude the possibility of some stable epigenetic change. Thus, the consequences of KRAS amplification in H6244-R1 cells are more complex than those of BRAF amplification in C6244-R1 cells and include both a quantitative increase in ERK1/2 signaling and qualitative differences in signaling. Combining AZD6244 with inhibition of PI3K and mTOR was more efficacious at inhibiting the proliferation of parental HCT116 cells than was AZD6244 alone, perhaps providing justification for the use of a combination approach to tackle innate resistance. Indeed, the combination of an ERK1/2 pathway inhibitor and a PI3K-and-mTOR inhibitor is increasingly viewed as an attractive drug combination (37, 38).

Saturation mutagenesis studies have identified mutants of MEK1 that confer resistance to PD184352 or AZD6244 (18, 39); one of these, MEK1P124L, was identified in a resistant metastatic lesion from a melanoma patient treated with AZD6244 (18). We did not detect any acquired mutations in MEK1 or MEK2, which suggests that resistance to MEK1/2 inhibitors can emerge by various mechanisms. What are the relative advantages of increasing BRAF600E or KRAS13D abundance compared with drug-refractory mutations in MEK1/2? In the case of BRAF600E, one advantage is increased pathway activation. BRAF600E is up to 500 times as active as wild-type BRAF (3) and is refractory to ERK1/2-mediated feedback inhibition (27, 28). In addition, wild-type CRAF can inhibit BRAF, and tumor cells harboring BRAF600E typically increase their BRAF/CRAF ratio to overcome this inhibition (40); in C6244-R1 cells, BRAF600E was amplified without any change in CRAF, effectively freeing the amplified BRAF600E from CRAF-mediated inhibition. Together, these factors would provide a stronger signal for ERK1/2 activation in cells already addicted to that pathway (11). In contrast, whereas mutations in MEK1 can confer drug resistance by maintaining pathway output in the presence of drug, it is not clear that they consistently increase the specific activity of MEK1 in vitro or basal pathway activation in cells (18); hence, in periods of lowered drug concentrations (for example, in between doses of a treatment regime), MEK1 activity would not be increased. Amplification of KRAS will drive increased activation of multiple effector pathways, and specific amplification of mutant KRAS may confer other advantages; for example, wild-type Kras acts as a tumor suppressor, and allelic loss of wild-type Kras is frequently observed in mouse models of lung cancer driven by mutant Kras (41). Thus, an increase in the ratio of mutant/wild-type Kras is favored during tumor progression and also underpins AZD6244 resistance. The broader signaling repertoire afforded by KRAS13D amplification may drive gains in other tumorigenic properties (such as resistance to apoptosis), although we found that H6244-R1 cells were not cross-resistant to a range of cytotoxic drugs. Indeed, the use of AZD6244 and Taxotere or irinotecan is an effective combination for inhibiting the growth of CRC cell lines as tumors in mouse xenograft studies (9).

There is increasing evidence that “acquired” resistance can arise from expansion of a small subset of cells with a preexisting abnormality (34, 36, 4247); these cells have a growth advantage when a continuous selective pressure is applied. But how do such genetic abnormalities arise in the first place? Gene amplification can occur by various mechanisms [reviewed in (48)]; two well-documented mechanisms are the breakage-fusion-bridge cycle (repeated fusion and breakage of chromosomes after fusion of sister chromatids at a double-strand break) and double minute formation (episomal amplification of acentric DNA fragments). Double minutes can remain extrachromosomal or re-integrate, often away from their site of origin (49). For example, amplification of EGFR in response to EGFR inhibition probably occurs through formation of double minute chromosomes (34). In our samples, all metaphase spreads and interphase nuclei contained focal amplifications, without extrachromosomal copies (double minutes). The KRAS amplicon we observed appears to be distinct from the normal KRAS locus (Fig. 4C), which might suggest it was generated by reintegration of double minutes. However, the breakage-fusion-bridge cycle could also account for this; KRAS amplification in H6244-R1 cells was near the junction of a chromosome translocation, making it possible that fusion of chromosome termini was involved in generating the amplification. In addition, we saw some evidence of ongoing fusion of chromosome termini in C6244-R1 cells, where the abnormal chromosome 7 derivative in the minor population may have been formed by the fusion of two chromosome 7p termini. Some NSCLC cells are predisposed to the development of low-level MET amplification, which contributes to gefitinib resistance (47); MET is located at a fragile site in chromosome 7, and the breakage-fusion-bridge cycle can drive the intrachromosomal amplification of MET in a human gastric carcinoma (50). Our results suggest that the copy number gain only occurs in the chromosome harboring the mutant BRAF or KRAS allele, not the wild type. Indeed, there is correlative evidence that activating mutations can be associated with local copy number gains (51, 52), but this is poorly understood.

Intriguingly, C6244-R1 cells regain sensitivity to the antiproliferative effects of AZD6244 after prolonged drug withdrawal. The same result was observed in single-cell clones of the C6244-R1 population, suggesting that a true re-adaptation occurs, as opposed to an outgrowth of a small population of parental-like cells able to persist in the original C6244-R1 culture. This has therapeutic implications and may provide a rationale for rechallenge with pathway inhibitors at some point after primary acquired resistance. Another therapeutic approach that may be of benefit is to identify drug combinations that promote apoptosis rather than arrest of cells; this may prevent the emergence of resistance and hence be a more effective strategy in the long term. These conclusions are supported by the findings of an in vitro study examining resistance to gefitinib in NSCLC cells (34).

In summary, we describe the generation of tumor cell lines with a common mechanism of acquired resistance to MEK1/2 inhibitors characterized by amplification of the driving oncogene. These findings have implications for therapeutic strategies involving MEK1/2 inhibitors currently in clinical trials and may be relevant for other kinase inhibitors that target downstream pathway components rather than the upstream driving oncogene itself.

Materials and Methods

Materials

AZD6244 and AZ12321046 were provided by AstraZeneca; AZ628 was provided by L. Drew (AstraZeneca R&D Boston). AZ12321046 is a mixed inhibitor of PI3K and TOR; it exhibits an IC50 of <10 nM against the enzymes PI3K α, β, δ, and γ and mTOR in vitro, decreases P-PKB abundance with an IC50 of 4 nM in MDA-MA-468 cells, and inhibits proliferation of MDA-MB-468 cells with an IC50 of 368 nM. PI-103 was purchased from Calbiochem. Antibodies against BRAF (sc-5284), pRB (sc-50), FRA-1 (sc-183), and FRA-2 (sc-604) were from Santa Cruz Biotechnology; KRAS (OP24), cyclin D1 (CC12), and p27KIP1 (NA35) from Calbiochem; P-ERK1/2 (9106), MEK1/2 (9122), P-MEK1/2 (9121), PKB (9272), P(S473)-PKB (9271), P(T308)-PKB (9275), p70 S6K (9202), phospho-p70 S6K (9205), 4EBP1 (9452), phospho-4EBP1 (9455), phospho pRB (9301), and c-JUN (9165) from Cell Signaling Technology; and ERK1 (610031) and CRAF (610151) from BD Biosciences. Horseradish peroxidase–conjugated secondary antibodies were from Bio-Rad, and detection was with the enhanced chemiluminescence (ECL) system (GE Healthcare). All other reagents were from Sigma.

Cell culture and generation of drug-resistant cell lines

COLO205 and HT29 cells were obtained from the American Type Culture Collection; HCT116 cells were provided by B. Vogelstein (Johns Hopkins University); LoVo cells were provided by K. Ryan (The Beatson Institute for Cancer Research). Cells were cultured in “complete medium,” comprising Dulbecco’s modified Eagle’s medium (HCT116 and LoVo cells) or RPMI 1640 (COLO205 cells) or McCoy’s 5A (HT29 cells) supplemented with glucose (4.5 mg/ml), penicillin (100 U/ml), streptomycin (100 μg/ml), 2 mM glutamine, and 10% (v/v) fetal bovine serum, and routinely tested for mycoplasma infection every 2 to 3 months (all tests were negative).

Dose-escalation method

COLO205 and HCT116 cells in exponential growth phase were initially exposed to 0.1 μM AZD6244. The cultures were observed daily and maintained and passaged normally with gradually increasing doses of AZD6244, up to maximal doses of 1 μM for COLO205 cells or 2 μM for HCT116 cells. The established resistant cell lines were then maintained in medium containing the maximal dose of AZD6244 (1 or 2 μM) to maintain selective pressure for AZD6244 resistance and named C6244-R1 and H6244-R1, respectively. HCT116 cells resistant to 1 μM PD184352 were generated by the same protocol described above for H6244-R1 cells and were named H184-R.

Chronic maximum-dose method

Cells in exponential growth phase were exposed to AZD6244 (1 μM for COLO205 and HT29 cells, 2 μM for HCT116, and 4 μM for LoVo cells). Cells were maintained in these doses of AZD6244 for a minimum of five passages until a population of cells emerged that could sustain growth under these conditions. The established resistant cell lines were then maintained in medium containing AZD6244 to maintain selective pressure for AZD6244 resistance and were designated C6244-R2, HT6244-R, H6244-R2, and L6244-R cells for AZD6244-resistant derivatives of COLO205, HT29, HCT116, and LoVo cells, respectively.

Colony formation assays

HCT116 and H6244-R1 cells were seeded in 12-well plates at 200 cells per well. Cells cultured for 14 days in the presence or absence of AZD6244, as indicated in the figure legend, were stained with crystal violet to visualize colony growth. Following treatment for 72 hours as indicated in the figure legend, COLO205 or C6244-R1 cells were trypsinized, counted with a hemocytometer, and resuspended at low density (100 or 200 cells) in 6 ml of 0.15% agar, 1× RPMI, penicillin (100 U/ml), streptomycin (100 μg/ml), 2 mM glutamine, and 10% (v/v) fetal bovine serum. After 14 days in culture, macroscopic colonies were stained with methylthiazolyldiphenyl-tetrazolium bromide (MTT) and counted.

Western blotting

Cells were lysed in ice-cold TG lysis buffer, assayed for protein content, and fractionated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) as described previously (53). SDS-PAGE gels were transferred to Immobilon P membranes (Millipore), which were blocked in 0.1% (v/v) Tween-20–TBS (tris-buffered saline) containing 5% (w/v) powdered milk and probed with the indicated antibodies. Immune complexes were visualized with the ECL system (GE Healthcare).

Assay of cell proliferation

Flow cytometric analysis of cell cycle profiles and [3H]thymidine incorporation were performed as described previously (11, 53). Assay of cell growth inhibition by paclitaxel, 5-fluorouracil, or cisplatin was performed by crystal violet staining as described previously (54).

Assay of KRAS-GTP

To measure the abundance of active KRAS, we washed cells in ice-cold phosphate-buffered saline (PBS) and harvested them on ice in TG lysis buffer. Cell extracts were cleared by centrifugation, and supernatant protein concentrations were measured by Bradford protein assay (Bio-Rad). Equal quantities of cell extracts were incubated for 2 hours at 4°C with glutathione S-transferase (GST)–Raf-1–RBD fusion protein (provided by J. Downward, Cancer Research UK London Research Institute) bound to glutathione-Sepharose beads. Beads were washed twice with ice-cold lysis buffer before separation on SDS-PAGE and immunoblotting with KRAS antibodies.

siRNA sequences and RNAi

For transient RNAi, the following oligos were used: siBRAF, AAAGAAUUGGAUCUGGAUCAU; siScram (BRAF), AAUGAAUUGCAUCUGGAACAU; KRAS and Non-targeting ON-TARGETplus SMARTpool siRNA reagents were purchased from Dharmacon. COLO205 and C6244-R1 cells were routinely passaged 2 days before transfection. For transfection, 4 × 106 cells were resuspended in 100 μl of Nucleofector solution T and mixed with 7 μg of the appropriate siRNA oligo. Cells were transfected by electroporation with a Nucleofector Device (Amaxa) and protocol T-020. For each condition, transfected cells were pooled before plating as biological triplicates for subsequent experimental procedures. For HCT116 and H6244-R1 cells, siRNA oligos were mixed with Opti-MEM medium, and an equivalent volume of Opti-MEM was combined with DharmaFECT2 transfection reagent and incubated for 5 min. Both siRNA and DharmaFECT2 were combined, mixed well, and incubated for 20 min. siRNA/DharmaFECT complexes were added to wells of a six-well tray, and 4 × 105 cells were added per well in penicillin/streptomycin-free medium. Forty-eight hours after transfection, cells were trypsinized and redistributed into the appropriate experimental receptacles.

Genomic qPCR

qPCR assays for BRAF, KRAS (table S1), cyclin D1 (cat. no. Hs02666923), and the control genes TBP (cat. no. 4333769), TERT (cat. no. 4403315), and RNAseP (cat. no. 4316831) were purchased from ABI. All assays were performed with the Mx3000P Real-Time PCR system (Agilent Technologies/Stratagene Products, cat. no. 401512). Each assay was performed on the cell lines and on two genomic DNA samples from control healthy volunteers. Assays were run in triplicate under standard conditions and absolute quantity calculated by relating the CT value to a standard curve. The ratio of the mean quantity for each test gene to each control gene was calculated per sample and the mean of the control individuals was used to normalize the cell line data.

Mutation screen of KRAS, NRAS, BRAF, MEK1, and MEK2

Genomic DNA from cell lines was extracted and PCR was performed with primers to amplify the required regions from each of the genes (table S2). Forward and reverse primers were tagged with M13F and M13R sequence, and PCR products were sequenced in both directions by dye-terminator sequencing using the ABI3730 capillary sequencer. Sequence traces were analyzed for variation manually after assembly and quality calling with polyphred/phrap/consed packages. Only exons 1 and 2 of KRAS were screened, because these contain the commonly reported mutations in codons 12, 13, and 61. Only exon 15 of BRAF was screened, because this contains the commonly reported mutations in codons 600 and 601. Only exons 2 and 3 of NRAS were screened, because these contain all the mutations reported in COSMIC database with the exception of the rare A146T mutation. All 11 exons of MEK1 and MEK2 were screened.

FISH

Metaphase preparation and FISH was performed as described previously (55, 56). BAC clones for FISH experiments, RP11-707G1 (HG36, chr12:25269925-25447810) and RP11-1173P7 (HG36, chr7:140392402-140558165), were selected for proximity to KRAS and BRAF, respectively, and were obtained from BACPAC resources. BACs were labeled with fluorescein isothiocyanate-12-deoxyuridine triphosphate (dUTP) (Roche) by nick translation with a commercial kit (Abbott Molecular). The identity of BACs was confirmed by PCR of STS (sequence tagged site) markers SHGC-144445 and SWSS783. Whole chromosome paints (libraries of fluorescently labeled DNA fragments from a given normal chromosome used to identify that chromosome by hybridization) were made from flow-sorted chromosomes amplified by degenerate oligonucleotide-primed PCR (DOP-PCR) as described (57). Sorted human chromosomes were supplied by M. Ferguson-Smith (Department of Veterinary Medicine, University of Cambridge, UK). DOP-PCR products were labeled with Spectrum orange dUTP (Abbott Molecular) as above. Images were captured with Cytovision software (Applied Imaging/Genetix). For each sample, we scored at least 25 metaphase spreads and 200 interphase nuclei.

Expression array analysis of cell lines

Cells from each cell line were plated in triplicate in their normal growth medium (that is, with or without AZD6244 for resistant or parental cells, respectively). Sixteen hours later, the medium was replaced in all dishes with complete medium (no AZD6244). Cells were incubated for 24 hours and then lysed in TRIZOL. After chloroform addition and phase separation, RNA was precipitated from the upper aqueous phase, washed, and diluted in nuclease-free water. Samples were sent to Cambridge Genomic Services where gene expression was measured with Illumina Human-6 v2 Expression BeadChips following standard protocol. The raw data were filtered to remove any nonexpressed probes, transformed with a variance stabilization algorithm (58), and normalized by quantile normalization, conducted with the Bioconductor lumi package (59). Comparisons results and linear regressions were completed on the normalized data with the limma package (60). Microarray data are available in the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-468.

Calculation of mean expression values for gene signatures

Published gene expression gene lists representing MEK and RAS activation were retrieved from (32, 33) and mapped to Illumina gene expression data annotation by gene (Entrez) identifier. The fold change (FC) between the mean expression of the parent and the acquired resistant lines was calculated, and significance was assessed as t test, with P value <0.05. Where multiple Illumina probe sets mapped to a single gene, the most significant result for that gene was selected. Lists of genes demonstrating FC > 2 with P < 0.05, or FC > 1.5 with P < 0.001, were uploaded to Oncomine (61) and enrichment to chromosomal cytobanding was assessed. For the “MEK functional activation” and “Compensatory resistance” signatures (32), a mean expression measure was taken across all genes. To identify the signature genes most likely to respond in colorectal cell lines, we also assessed the ability of each gene to consistently predict preclinical AZD6244 sensitivity and preclinical/clinical MEK activation (BRAF mutation; compound inhibition) specifically within published colorectal sample data (29) (table S4).

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/4/166/ra17/DC1

Fig. S1. Growth curves of HCT116 and H6244-R1 or COLO205 and C6244-R1 cells and growth factor–independent proliferation.

Fig. S2. AZD6244-resistant cells are cross-resistant to other MEK inhibitors.

Fig. S3. AZD6244-resistant cells are not cross-resistant to cytotoxic drugs.

Fig. S4. Kinetics of emergence of AZD6244 resistance.

Fig. S5. Genomic PCR demonstrates amplification of BRAF and KRAS but not cyclin D1 in C6244-R1 and H6244-R1 cells.

Fig. S6. BRAF and KRAS mutation analysis.

Fig. S7. H6244-R1 cells show increased PI3K-PKB pathway signaling, but C6244-R1 cells do not.

Fig. S8. Heat map of differential expression of a gene expression signature indicative of RAS activation in HCT116 versus H6244-R1 cells and COLO205 versus C6244-R1 cells.

Fig. S9. Resistance to AZD6244 in C6244-R1 cells is reverted upon prolonged withdrawal of the drug.

Fig. S10. Different methods of generating AZD6244-resistant cells give rise to a common mechanism of resistance.

Fig. S11. HCT116 cells rendered resistant to PD184352 (H184-R) also show increased abundance of KRAS.

Table S1. Summary of primers and probes used for the BRAF and KRAS qPCR assays.

Table S2. Summary of primers used in the mutation screen to amplify the required regions for sequencing.

Table S3. Summary of genetic background of cell lines used.

Table S4. Weighting values relative to the performance of genes for the “MEK functional activation” and “Compensatory resistance” gene signatures specifically within colorectal cancer (CRC) samples.

References and Notes

  1. Acknowledgments: We thank S. Guichard and S. Cosulich (AstraZeneca) and members of the Cook group for advice and R. Slack, G. McWalter, and R. Sinton from AstraZeneca R&D Genetics for technical assistance. Funding: This work was funded by a grant from AstraZeneca (A.S.L.) and by the Babraham Institute (S.J.C. and K.B.). M.J.S. was supported by a Biotechnology and Biological Sciences Research Council Ph.D. studentship. S.N. was supported by a Medical Research Council Ph.D. studentship. Work in P.A.W.E.’s lab is also supported by Cancer Research UK. Author contributions: A.S.L., K.B., M.J.S., S.N., J.R.D., M.H., P.A.W.E., P.D.S., and S.J.C. designed the experiments; A.S.L., K.B., M.J.S., S.N., J.R.D., and M.H. performed the experiments; A.S.L., K.B., M.J.S., S.N., J.R.D., M.H., P.A.W.E., P.D.S., and S.J.C. analyzed the results; and A.S.L., P.D.S., and S.J.C. wrote the manuscript with contributions from J.R.D., S.N., and P.A.W.E. Competing interests: J.R.D., M.H., and P.D.S. are employees of AstraZeneca. P.D.S. is a shareholder of AstraZeneca. AZD6244 (selumetinib), AZ628, and AZ12321046 were supplied under a material transfer agreement (MTA) from AstraZeneca. Accession numbers: The gene expression array data are available in the ArrayExpress database under accession number E-MTAB-468.
View Abstract

Navigate This Article