Research ArticleCancer

Gatekeeper Mutations Mediate Resistance to BRAF-Targeted Therapies

See allHide authors and affiliations

Science Translational Medicine  09 Jun 2010:
Vol. 2, Issue 35, pp. 35ra41
DOI: 10.1126/scitranslmed.3000758

Abstract

BRAF is a serine-threonine–specific protein kinase that is mutated in 2% of human cancers. Oncogenic BRAF is a validated therapeutic target that constitutively activates mitogen-activated protein kinase kinase (MEK)–extracellular signal–regulated kinase (ERK) signaling, driving tumor cell proliferation and survival. Drugs designed to target BRAF have been developed, but it is difficult to prove that they mediate their antitumor effects by inhibiting BRAF rather than by working through off-target effects. We generated drug-resistant versions of oncogenic BRAF by mutating the gatekeeper residue. Signaling by the mutant proteins was resistant to the small-molecule inhibitor sorafenib, but sorafenib still inhibited the growth of tumors driven by the mutant protein. In contrast, both BRAF signaling and tumor growth were resistant to another RAF drug, PLX4720. These data provide unequivocal evidence that sorafenib mediates its antitumor effects in a manner that is independent of its ability to target oncogenic BRAF, whereas PLX4720 inhibits tumor growth by targeting oncogenic BRAF directly.

Introduction

In the last 15 years, the identification of hyperactivated cell signaling pathways as key drivers of cancer has prompted the search for drugs to target specific components of those pathways (1). Protein kinases, in particular, have proven to be tractable targets, and small-molecule inhibitors to target these enzymes have achieved spectacular clinical results, driving discovery programs aimed at developing new drugs to target specific protein kinases in particular human pathologies (2). One of the protein kinases to emerge as an important therapeutic target in cancer is BRAF (36). There are three RAF proteins in humans, ARAF, BRAF, and CRAF (7), and in normal cells, these kinases are activated downstream of the small G protein RAS, which is activated downstream of receptor tyrosine kinases, and cytokine and hormone receptors. Activated RAF phosphorylates and activates a second protein kinase, called mitogen-activated protein kinase (MAPK) kinase (MEK), which then phosphorylates and activates a third protein kinase, called extracellular signal–regulated kinase (ERK). ERK phosphorylates a multitude of cytosolic and nuclear substrates, thereby regulating complex cell processes such as proliferation, survival, differentiation, and senescence. The magnitude and duration of ERK activity controls cell behavior, so this pathway must be carefully regulated to ensure that cells respond appropriately to changing environmental conditions (8). However, in cancer, this pathway is constitutively activated, and consequently, survival and proliferation are the favored outcomes.

Whereas ARAF and CRAF mutations are rare in cancer, somatic gain-of-function mutations in BRAF are particularly common in melanoma (43% of cases), thyroid (45%), ovarian (10%), and colorectal (13%) cancers (9). A glutamic acid for valine substitution at position 600 (V600E) accounts for >90% of the BRAF mutations that occur in human cancer (3). This mutant activates BRAF by a factor of 500 and allows it to stimulate constitutive ERK and nuclear factor κB signaling, leading to transformation of cells such as fibroblasts and melanocytes (3, 10, 11). In melanoma cells, inhibition of V600EBRAF by small molecules or its depletion by RNA interference inhibits cell proliferation and induces apoptosis in vitro, and in vivo, it suppresses tumor cell growth. These data validate V600EBRAF as a therapeutic target, and several drugs that target BRAF have been developed (5, 6, 1214). BRAF drugs have raised a great deal of interest, particularly in melanoma, a potentially deadly form of skin cancer that has a mortality rate of 10 to 20%. If diagnosed early, melanoma can be cured by surgery. However, the malignant metastatic disease is refractory to treatment and has poor prognosis, with median survival rates of 6 to 9 months and 5-year survival rates of ~10%. Thus, new drugs are being sought with some urgency.

The first BRAF-targeting drug to enter the clinic was sorafenib, a compound that was designed to inhibit CRAF but which also inhibits V600EBRAF (6, 14). Sorafenib is a type II inhibitor, which binds to the inactive conformation of V600EBRAF with the so-called DFG motif in the “out” position (15). In clinical trials, sorafenib proved to be ineffective in melanoma in which BRAF is mutated (16, 17). However, sorafenib has been shown to be effective in renal cell and hepatocellular carcinoma (18, 19), demonstrating that it is a valuable drug. These data highlight one of the most pressing problems with kinase drug discovery programs—that it is difficult to prove that a particular drug mediates its antitumor effects by inhibiting its putative clinical target rather than by working through unexplained off-target effects. This uncertainty can cause confusion, and indeed, the failure of sorafenib in melanoma clinical trials has prompted suggestions that BRAF is not a viable therapeutic target in melanoma (20, 21).

The issues outlined above highlight the importance of developing effective and appropriate methods in which to test experimental drugs during preclinical development. The aim of this study was to develop approaches to test the selectivity of BRAF inhibitors in vitro and in vivo. The goal was to examine whether the responses to these agents are mechanism-based or mediated by off-target effects.

Results

Potency and selectivity of BRAF inhibitors in vitro

For these studies, we focused on four recently described BRAF-targeting drugs: the type I inhibitors PLX4720 (13) and SB590885 (12) and the type II inhibitors sorafenib (14) and RAF265 (22) (fig. S1). First, we directly compared the activities of these drugs against a panel of 17 human melanoma, colon, or breast cancer cell lines in which BRAF is either wild-type or mutant. PLX4720 displayed the greatest level of selectivity for the BRAF-mutant cells, inhibiting their proliferation more potently than it inhibited the proliferation of cells expressing wild-type BRAF by a factor of 129 on average (Fig. 1A). RAF265 and SB590885 were also more selective for BRAF-mutant than for BRAF wild-type cells, with average efficiency increases by factors of 16 and 14, respectively (Fig. 1, B and C). In contrast, sorafenib was not selective and inhibited the growth of BRAF-mutant and BRAF wild-type cells equally (4.6 versus 4.7 μM; Fig. 1D). In addition to being the most selective, PLX4720 was the most potent. It inhibited BRAF-mutant cell proliferation with an average IC50 (median inhibitory concentration) of 0.13 μM (Fig. 1A), followed by RAF265 at 0.33 μM (Fig. 1B) and SB590885 at 0.81 μM (Fig. 1C). Sorafenib was the least potent and inhibited BRAF-mutant cell proliferation with an average IC50 of 4.6 μM (Fig. 1D). We also profiled these compounds against a panel of 63 protein kinases (Fig. 1E). In this analysis, SB590885 was the most selective and only inhibited CK1 strongly. PLX4720 inhibited SRC and LCK, and RAF265 inhibited SRC, LCK, and NEK2a. Sorafenib was the least selective, as it also inhibited LCK, HIPK2, Aurora B, ERK8, p38α, and p38β.

Fig. 1

Direct comparison of BRAF drugs in vitro. (A to D) Comparison of GI50 (growth inhibitory effect) values of 7 BRAF-mutant (COLO829, A375M, WM266.4, HT29, UACC62, SKMEL28, and COLO205) and 10 BRAF wild-type (WT) (D04, SKMEL2, HCT116, WM1361, SKMEL23, CHL, KM12, D35, BT474, and D24) cell lines treated with PLX4720 (A), RAF265 (B), SB590885 (C), and sorafenib (D). (E) Kinase inhibition profiles of PLX4720, RAF265, SB590885, and sorafenib tested at a concentration of 1 μM against a panel of 63 protein kinases (45). The percent remaining activity compared to controls for each kinase was used to generate a “heat map” of kinase inhibition, the values for which are indicated by the color bar. (F) In vitro inhibition of V600EBRAF kinase activity by sorafenib, RAF265, PLX4720, and SB590885. (G) In vitro inhibition of V600EBRAF kinase domain using Z-lyte and DELFIA technologies.

Next, we tested the drugs against full-length V600EBRAF using glutathione S-transferase (GST)–MEK as a direct substrate. SB590885 was the most potent drug and inhibited V600EBRAF at ~0.001 μM (Fig. 1F and table S1). Sorafenib was the second most potent, with an IC50 of 0.043 μM, followed by RAF265 at 0.155 μM (Fig. 1F and table S1). These values are close to the published values of 0.0002, 0.038, and 0.140 μM, respectively (12, 14, 22). Surprisingly, however, and despite the fact that it was the most potent drug against the BRAF-mutant cell lines, we found that PLX4720 was not particularly potent in vitro and only inhibited V600EBRAF at 2.8 μM (Fig. 1F), a value that is considerably higher than the published value of 0.013 μM against the isolated V600EBRAF kinase domain (13). However, in our DELFIA-based kinase assay and in a commercial Z-lyte technology assay (Invitrogen) where both approaches used the kinase domain of V600EBRAF, PLX4720 was also relatively inactive, with IC50 values of 7.1 and 0.134 μM, respectively (Fig. 1G). Thus, we observed a poor correlation between the in vitro activity of the compounds against V600EBRAF and their activity against BRAF-mutant cell lines. Whereas PLX4720 was the most potent and selective against the cells, it was the least active against purified V600EBRAF. Conversely, sorafenib was relatively potent against purified V600EBRAF, but it was neither potent nor selective for BRAF-mutant cells. This highlights the problem of demonstrating that drugs mediate their antiproliferative effects by inhibiting their intended target rather than through unknown off-target effects.

Gatekeeper mutations mediate BRAF resistance to drugs in vitro

To tease apart the effects on V600EBRAF in vitro and in vivo, we created drug-resistant mutants of V600EBRAF. Specifically, we targeted the gatekeeper residue of BRAF because it is known to mediate resistance to adenosine triphosphate (ATP) pocket binding drugs in other kinases (23). The gatekeeper residue in BRAF is Thr529 (T529; fig. S2), and we substituted this residue with methionine (T529M), isoleucine (T529I), or asparagine (T529N). First, we tested the effects of these substitutions on BRAF in vitro. Relative to V600EBRAF, T529N,V600EBRAF retains 87% activity, T529M,V600EBRAF retains 77% activity, and T529I,V600EBRAF retains only 10% activity (Fig. 2A). Because V600EBRAF is activated by a factor of ~480 relative to wild-type BRAF (15), these levels of activity equate to activation by factors of 420, 370, and 48, respectively, relative to the wild-type protein. None of the substitutions affected V600EBRAF affinity for ATP (Fig. 2B and table S1), so we assessed how they affected V600EBRAF sensitivity to the drugs in vitro. T529N,V600EBRAF, T529M,V600EBRAF, and T529I,V600EBRAF were all less sensitive to SB590885 than was V600EBRAF by factors of 200 to 650 (Fig. 2C and table S1). T529N,V600EBRAF was also less sensitive to sorafenib and RAF265 by factors of 250 and 65, respectively (Fig. 2, D and E, and table S1). In contrast, T529M,V600EBRAF and T529I,V600EBRAF were as sensitive as V600EBRAF to sorafenib and RAF265 (Fig. 2, D and E). Notably, none of the gatekeeper mutants mediated substantial resistance to PLX4720 in the in vitro assay. T529N,V600EBRAF, T529I,V600EBRAF, and T529M,V600EBRAF were less sensitive to PLX4720 than was V600EBRAF by factors of only 12, 3.4, and 2, respectively (Fig. 2F and table S1).

Fig. 2

Gatekeeper mutants of V600EBRAF are catalytically active and display differential sensitivity to kinase inhibitors. (A) Myc epitope–tagged V600EBRAF (V600E), T529M,V600EBRAF (T529M,V600E), T529N,V600EBRAF (T529N,V600E), or T529I,V600EBRAF (T529I,V600E) was transiently expressed in COS7 cells, and their kinase activity was measured in an immunoprecipitation kinase cascade assay using MEK, ERK, and MBP as sequential substrates. Activity is shown relative (%) to V600EBRAF. (B) The kinase activity of the BRAF mutants from (A) was measured in the presence of increasing concentrations of ATP, and the Km[ATP] so determined is presented in table S1. (C to F) Kinase assays were conducted for each BRAF mutant from (A) in the presence of increasing concentrations of SB590885 (C), sorafenib (D), RAF265 (E), and PLX4720 (F). The graphs are the representative of three independent experiments performed in duplicate, with error bars to indicate the SEM, and the IC50 values determined from these results are presented in table S1.

Gatekeeper mutations mediate BRAF drug resistance in cells

We also assessed the activity of the gatekeeper mutants in cells. First, we show that transient expression of T529NBRAF, T529MBRAF, and T529IBRAF did not activate MEK in human embryonic kidney 293 (HEK293) cells, whereas T529N,V600EBRAF, T529M,V600EBRAF, and T529I,V600EBRAF all activated MEK (Fig. 3A and fig. S3). Stable expression of V600EBRAF, T529M,V600EBRAF, T529I,V600EBRAF, and T529N,V600EBRAF in HEK293 cells also caused increased ERK activity (Fig. 3B), and this was blocked by the MEK inhibitor PD184352 (Fig. 3C). Together, these data demonstrate that the gatekeeper substitutions alone do not activate wild-type BRAF. The substituted proteins all constitutively activate MEK in cells, so we used the cells to test the sensitivity of the gatekeeper mutants to the BRAF inhibitors.

Fig. 3

Gatekeeper mutants mediate drug resistance in cells. (A) Myc-tagged BRAF (WT), V600EBRAF (V600E), T529NBRAF (T529N), T529N,V600EBRAF (T529N,V600E), T529MBRAF (T529M), T529M,V600EBRAF (T529M,V600E), T529IBRAF (T529I), or T529I,V600EBRAF (T529I,V600E) were transiently expressed in COS7 cells, and after 48 hours, the levels of MEK phosphorylation were determined by quantitative fluorescent Western blotting. MEK1 levels on the same blot were used as a loading control. Results are the mean of five independent experiments, with error bars to indicate the SEM. (B) HEK293 cells engineered for stable expression of Myc epitope–tagged V600EBRAF (V600E), T529N,V600EBRAF (T529N,V600E), T529M,V600EBRAF (T529M,V600E), or T529I,V600EBRAF (T529I,V600E) or a vector control were serum-starved for 16 hours and analyzed by Western blotting for the Myc-tagged BRAF (mBRAF), phospho-ERK (ppERK), and total ERK2 on the same blot as a loading control. Star indicates a nonspecific band. (C) HEK293 cells stably expressing Myc-tagged V600EBRAF (V600E), T529M,V600EBRAF (T529M,V600E), T529I,V600EBRAF (T529I,V600E), or T529N,V600EBRAF (T529N,V600E) were serum-starved for 16 hours and then treated with PD184352 for 6 hours. Cell lysates were Western-blotted for phospho-ERK and total ERK2 on the same blot as a loading control. (D) Serum-starved HEK293 cells stably expressing Myc-tagged V600EBRAF (V600E), T529M,V600EBRAF (T529M,V600E), T529I,V600EBRAF (T529I,V600E), or T529N,V600EBRAF (T529N,V600E) were treated with PLX4720, SB590885, sorafenib, or RAF265 at the indicated concentrations for 6 hours. Cell lysates were analyzed by Western blotting for phospho-ERK and total ERK2 on the same blot as a loading control.

V600EBRAF-driven ERK activation was almost completely blocked by 0.1 μM PLX4720, whereas even at 3 to 10 μM PLX4720, ERK was still active in T529M,V600EBRAF-, T529I,V600EBRAF-, and T529N,V600EBRAF-expressing cells (Fig. 3D). Similarly, SB590885 blocked ERK activity in V600EBRAF-expressing cells in the 0.1 to 0.3 μM range, whereas ERK was unaffected by 1 μM SB590885 in T529M,V600EBRAF-, T529I,V600EBRAF-, and T529N,V600EBRAF-expressing cells (Fig. 3D). Sorafenib inhibited ERK in the 1 to 3 μM range in V600EBRAF-, T529M,V600EBRAF-, and T529I,V600EBRAF-expressing cells, but even 30 μM sorafenib did not block ERK activity in T529N,V600EBRAF-expressing cells (Fig. 3D). Finally, RAF265 blocked ERK activity in the 0.1 to 0.3 μM range in V600EBRAF- and T529M,V600EBRAF-expressing cells and in the 0.3 to 1.0 μM range in T529I,V600EBRAF-expressing cells, whereas even at 3 μM RAF265, ERK was still only partially inhibited in T529N,V600EBRAF-expressing cells (Fig. 3D).

The gatekeeper mutations block drug binding through steric hindrance

The data above show that T529N,V600EBRAF is resistant to all four drugs, whereas T529I,V600EBRAF and T529M,V600EBRAF are only resistant to the type I inhibitors. To understand the mechanisms underlying these differences, we examined the crystal structures of BRAF bound to these drugs. A key difference between the type I and the type II inhibitors is the position of the DFG motif on binding. Type I inhibitors bind with the DFG motif in the “in” and therefore active conformation (12, 24), whereas, when the type II inhibitors bind, the DFG is the “out” and therefore inactive conformation (2, 15). The consequences of the different positioning of the DFG motif are that the type I inhibitors bind to a closed pocket (Fig. 4A), whereas the type II inhibitors elaborate into a channel created by displacement of the DFG motif (Fig. 4B). Note that T529 lines the walls of both the closed pocket and the channel (Fig. 4, A and B).

Fig. 4

Substitution of the gatekeeper residue in V600EBRAF results in a steric clash. (A) Image of the crystal structure of the type I inhibitor SB590885 bound to the DFG in conformation of BRAF. This image shows the presence of the closed pocket into which the inhibitor binds. T529 is highlighted in red. (B) Image of the crystal structure of the type II inhibitor sorafenib bound to the DFG out conformation of BRAF, showing the channel into which the right-hand side of the molecule elaborates. T529 is highlighted in red. (C) In silico docking of PLX4720, SB590885, sorafenib, and RAF265 into V600EBRAF (V600E) or the models for T529N,V600EBRAF (T529N,V600E), T529I,V600EBRAF (T529I,V600E), or T529M,V600EBRAF (T529M,V600E). Van der Waals forces are indicated as red spheres. The region of steric clash is indicated by the arrows.

We used in silico docking to introduce drugs into the appropriate DFG in or DFG out conformations of BRAF by means of the program GOLD (25). The published binding modes for SB590885 and PLX4720 bound to the DFG in conformation were faithfully reproduced by this analysis (fig. S4). Similarly, GOLD faithfully reproduced the published binding mode for sorafenib bound to the DFG out motif of BRAF and also faithfully reproduced previously described binding mode for early analogs of RAF265 (fig. S4). We introduced the gatekeeper substitutions, relaxing the protein within 9 Å of this residue and calculated the favored rotamers of the newly introduced side chains. We then docked the drugs, allowing the protein to adopt its favored rotamer during the docking run. Only SB590885 failed to dock, so for this compound, the images generated are superimpositions of the native docking pose into the gatekeeper-mutated proteins. The results reveal a significant hypothetical clash between the side chain of the substituted asparagine and all four drugs (Fig. 4C). In contrast, methionine and isoleucine only formed significant hypothetical clashes with PLX4720 and SB590885 and did not clash with sorafenib or RAF265 (Fig. 4C).

Gatekeeper mutations in BRAF render cells insensitive to selective BRAF inhibitors

Next, we examined the in vivo consequences of signaling by the gatekeeper-substituted proteins. For these studies, we selected Ba/F3 pro-B cells because they display low levels of spontaneous transformation and their growth is absolutely dependent on interleukin-3 (IL-3) (26). We find that HEK293 cells are not suitable for these growth studies because although V600EBRAF activates MEK-ERK signaling in these cells, their growth cannot be rendered completely dependent on this signaling. V600EBRAF and T529N,V600EBRAF constitutively activate ERK in Ba/F3 cells, whereas wild-type BRAF does not (Fig. 5A). Furthermore, the growth of Ba/F3 cells can be rendered independent of IL-3 by stable expression of V600EBRAF or T529N,V600EBRAF but not by wild-type BRAF (Fig. 5B). Notably, the elevated ERK activity (Fig. 5C) and the IL-3–independent growth (Fig. 5D) driven by V600EBRAF and T529N,V600EBRAF in these cells are sensitive to MEK inhibition by PD184352, whereas the proliferation driven by IL-3 is less sensitive to this MEK inhibitor by a factor of 100 (Fig. 5D). Thus, in the absence of IL-3, Ba/F3 proliferation can be driven by oncogenic BRAF signaling through MEK, so we examined the sensitivity of these cells to the BRAF drugs.

Fig. 5

Ba/F3 cells as tools to study kinase inhibitor selectivity. (A) Stable expression of Myc epitope–tagged BRAF (WT), V600EBRAF (V600E), or T529N,V600EBRAF (T529N,V600E) or an empty vector control in the Ba/F3 cells. Cell extracts were Western-blotted for Myc-tagged BRAF, phospho-MEK (ppMEK), phospho-ERK (ppERK), and ERK2 on the same blot as a loading control. (B) Ba/F3 cells stably expressing Myc epitope–tagged BRAF (WT), V600EBRAF (V600E), or T529N,V600EBRAF (T529N,V600E) were assessed for growth by cell counting in the presence (+) or absence (−) of IL-3 (1 ng/ml). (C) Ba/F3 cells stably expressing V600EBRAF (V600E) or T529N,V600EBRAF (T529N,V600E) were treated with the indicated concentrations of PD184352 for 2 hours in the absence of IL-3 and analyzed for phospho-ERK (ppERK) and ERK2 on the same blot by Western blotting. (D) Ba/F3 cells expressing V600EBRAF (V600E) or T529N,V600EBRAF (T529N,V600E) were treated with the indicated concentrations of PD184352 in the absence (−) or presence (+) of IL-3 (1 ng/ml) for 24 hours, and DNA synthesis was determined relative to vehicle-treated controls (%).

PLX4720 inhibited the proliferation of V600EBRAF-expressing cells more effectively (by a factor of 26) than it inhibited the proliferation of T529N,V600EBRAF-expressing cells (Fig. 6A and table S2). SB590885 was more active against V600EBRAF- than against T529N,V600EBRAF-expressing cells by a factor of 11.4 (Fig. 6B and table S2), and RAF265 was more effective against V600EBRAF-expressing cells by a factor of 6.8 (Fig. 6C and table S2). Only sorafenib failed to display any significant selectivity, producing a factor of 1.17 differential (Fig. 6D and table S2).

Fig. 6

Gatekeeper mutations in V600EBRAF render Ba/F3 cells resistant to selective BRAF inhibitors. (A to D) Ba/F3 cells expressing V600EBRAF (V600E) or T529N,V600EBRAF (T529N,V600E) were treated with increasing concentrations of PLX4720 (A), SB590885 (B), RAF265 (C), or sorafenib (D) in the absence of IL-3 for 24 hours, and DNA synthesis was determined by [3H]thymidine incorporation. IC50 values from these results that were determined by nonlinear regression are presented in table S2. (E) Nude mice bearing established Ba/F3 tumors driven by V600EBRAF (V600E) or T529N,V600EBRAF (T529N,V600E) were treated daily by intraperitoneal dosing with vehicle (5% DMSO–water), sorafenib (50 mg/kg), or PLX4720 (20 mg/kg) for 4 days. Tumor sections were stained for phospho-MEK1/2, and nuclei were counterstained with methyl green. Representative images are shown; scale bar, 20 μm. (F) Tumor sections from (E) were scored for the percentage of cells displaying low, medium, or high intensity of staining, which was normalized to vehicle controls. (G and H) Nude mice bearing established subcutaneous tumors driven by V600EBRAF (V600E) (G) or T529N,V600EBRAF (T529N,V600E) (H) were treated daily by intraperitoneal dosing with vehicle (5% DMSO–water), sorafenib (50 mg/kg), or PLX4720 (20 mg/kg). Tumor volumes were measured every 4 to 5 days, and results are the average for six animals per group; error bars represent SEM.

Finally, we compared the responses of these cells to BRAF inhibitors in vivo. V600EBRAF- or T529N,V600EBRAF-expressing Ba/F3 cells were grown as subcutaneous allografts in nude mice for 7 days, and then the mice were treated with the BRAF drugs. First, mice were treated with drugs for 4 days and then the tumors were harvested 1 hour after the last dose, and MEK phosphorylation was examined by immunohistochemistry. We show that both sorafenib and PLX4720 suppress MEK phosphorylation in tumors driven by V600EBRAF (Fig. 6, E and F). In line with our in vitro results, neither sorafenib nor PLX4720 blocked MEK phosphorylation in tumors driven by T529N,V600EBRAF (Fig. 6, E and F). We show that sorafenib delayed the growth of V600EBRAF tumors, and despite failing to inhibit MEK, it also delayed the growth of T529N,V600EBRAF tumors (Fig. 6, G and H). In contrast, PLX4720 suppressed only the growth of tumors driven by V600EBRAF and had no effect on the growth of tumors driven by T529N,V600EBRAF (Fig. 6, G and H).

Discussion

Here, we investigated the in vitro and in vivo responses of V600EBRAF to several recently described BRAF-targeting drugs. In vitro, we find that SB590885 is the most potent drug and inhibited purified V600EBRAF with an IC50 of 0.001 μM. At 0.043 and 0.155 μM, respectively, sorafenib and RAF265 had intermediate potency, and PLX4720 was the least potent at 2.8 μM. Despite its apparent lack of in vitro activity, PLX4720 was the most potent and selective drug against cancer cell lines expressing oncogenic BRAF. It inhibited their growth with an average IC50 of 0.13 μM and was on average more selective for BRAF-mutant than for BRAF wild-type cells by a factor of 129. RAF265 and SB590885 were also reasonably potent and selective. They inhibited proliferation of BRAF-mutant cells at an average of 0.33 and 0.81 μM, respectively, and were more potent against BRAF-mutant than against wild-type BRAF cells by factors of about 16 and 14, respectively. Finally, sorafenib was neither selective nor potent and inhibited the growth of BRAF-mutant and BRAF wild-type cells at similar potency (4.6 and 4.7 μM, respectively).

Thus, we observed a poor correlation between the in vitro activity of these drugs and their activity in cells. Sorafenib was relatively potent against V600EBRAF in vitro, but it was neither potent nor selective against BRAF-mutant cell lines, whereas PLX4720 was both selective and potent for BRAF-mutant cell lines but was not very active against V600EBRAF in vitro. It is unclear why we could not confirm the previously reported in vitro activity (0.013 μM) of PLX4720 against V600EBRAF (13). It is not because the previous study used the isolated catalytic domain and we used full-length protein; because in our assay these constructs were similarly sensitive to PLX4720 and in an alternative assay format using Z-lyte technology, we still obtained an IC50 value that was higher than published values by a factor of ~10 (13). Furthermore, the values we report for SB590885, sorafenib, and RAF265 are close to their previously reported values (12, 14, 22), validating our in vitro assay system. This suggests that PLX4720 is exquisitely sensitive to assay conditions, making it difficult to interpret the in vitro kinase assay data with this drug. Thus, although we show PLX4720 to be highly selective against a panel of 63 kinases, it is unclear how these results would reflect inhibition of these kinases in cells.

In the past decade, we have witnessed an enormous increase in activities dedicated to the development of molecularly targeted drugs. These agents hold great promise for cancer treatment, but our data highlight one of the key issues with these drugs. It is difficult to prove that a so-called specific agent mediates its therapeutic effects by inhibiting its intended target rather than by working through unknown off-target effects. Thus, although PLX4720 has been shown to be exquisitely selective for cancer cells that carry mutations in BRAF, this does not formally prove that it is targeting V600EBRAF in vivo. Conversely, the failure of sorafenib to mediate clinical responses in melanoma patients prompted suggestions that V600EBRAF is not a good therapeutic target in this disease (20, 21) despite the extensive preclinical evidence supporting its validation as a target (27). This uncertainty clearly confounds judgment about the importance of clinical targets and the drugs designed to inhibit them. To address this problem directly, we investigated how gatekeeper substitutions affected responses by V600EBRAF to drugs in vitro and in vivo.

We replaced T529 in BRAF with isoleucine or asparagine because these amino acids have bulky side chains and can arise by single-base substitutions, so they are more likely to occur naturally in patients. All other single base-pair changes would introduce amino acids with small side chains (AGC/TCC: serine; GCC: alanine) or a proline, which is likely to be structurally disruptive (CCC). Notably, an isoleucine for threonine substitution of the gatekeeper in BCR-ABL mediates imatinib resistance in chronic myelogenous leukemia (28). We also generated a methionine-for-threonine gatekeeper substitution in BRAF because although this requires two base-pair changes, in the epidermal growth factor receptor (EGFR) this substitution mediates clinical resistance to gefitinib and erlotinib in lung cancer (29). By themselves, none of these substitutions activated wild-type BRAF. Furthermore, in the V600EBRAF background, all three mutants retained substantial amounts of kinase activity and activated constitutive MEK-ERK signaling in cells. Our cell line data show that the type I inhibitors were more sensitive to the gatekeeper substitutions than the type II inhibitors. Thus, all three substitutions mediated resistance to PLX4720 and SB590885, but only the asparagine mutant mediated resistance to sorafenib and RAF265.

A methionine-for-threonine gatekeeper substitution in L858REGFR reduces EGFR binding to ATP but only weakly affects gefitinib binding (30), a differential effect that causes a factor of 100 reduction in sensitivity to gefitinib in patient-derived cell lines (29). Unlike the EGFR, we show that gatekeeper substitutions did not affect ATP binding to V600EBRAF, and our modeling studies suggest that resistance is mediated by steric clashes between the drugs and the gatekeeper residue. Structural studies have shown that the type II binders elaborate into a channel in the kinase domain created by displacement of the DFG motif into the out conformation, whereas the type I binders bind to a closed pocket. Our data suggest that the channel accommodates the gatekeeper substitutions better than the pocket because the channel is more flexible, or because type II inhibitors pack onto the gatekeeper residue less tightly.

Notably, we obtained strong correlation between the in vitro and cell-based resistance of the gatekeepers for sorafenib, RAF265, and SB590885. In both assays, only the asparagine substitution was resistant to sorafenib and RAF265, whereas all three substitutions were resistant to SB590885. However, we did not find such a correlation with PLX4720 because the gatekeeper mutations only mediated modest resistance in vitro (by a factor of 2 to 12) but displayed substantial resistance in cells (factor of >100). These data support the notion that PLX4720 is particularly sensitive to assay conditions and highlight the difficulty of translating in vitro assay data to activity in cells. This demonstrates the importance of developing cell-based approaches to test the selectivity of newly synthesized drugs.

Critically, the gatekeeper-substituted proteins enabled us to assess the efficacy of drugs in vivo. For these studies, we selected Ba/F3 cells because their spontaneous transformation rate is low and, in the absence of IL-3, their in vitro and in vivo growth can be rendered absolutely dependent on V600EBRAF. In vitro, we demonstrated that PLX4720, SB590885, and RAF265 all selectively blocked the proliferation of the V600EBRAF-expressing cells over the T529N,V600EBRAF-expressing cells, whereas sorafenib did not display any such selectivity. In vivo, we showed that sorafenib blocked MEK activity in tumors driven by V600EBRAF, but not in tumors driven by T529N,V600EBRAF, confirming that resistance is maintained in vivo. However, despite failing to block T529N,V600EBRAF signaling in these tumors, sorafenib still suppressed their growth, demonstrating that it mediates its antitumor effects independently of its ability to block V600EBRAF signaling. In contrast, PLX4720 failed to inhibit either MEK activity or tumor growth driven by T529N,V600EBRAF, confirming that it mediates its antitumor effects through its ability to inhibit V600EBRAF.

Thus, we have established that the clinical target of sorafenib is not V600EBRAF. It has recently been shown that sorafenib inhibits V600EBRAF by driving it into a complex with CRAF (29). However, the lack of clinical activity of sorafenib in melanoma (30) suggests that sorafenib is unable to inhibit V600EBRAF in vivo through this mechanism either. Thus, although sorafenib can inhibit V600EBRAF in vitro, it cannot do so in vivo, and we establish that BRAF inhibition is not required for the antitumor activity of sorafenib. Our findings robustly dispel the confusion surrounding the suitability of BRAF as a therapeutic target and highlight the importance of relating target inhibition to therapeutic efficacy. Furthermore, our results are consistent with recent data showing that melanoma patients with BRAF-mutant tumors can achieve objective and sustained clinical responses to the potent and selective BRAF drugs PLX4032 and XL281 (31, 32). These trials have highlighted the critical need to use targeted therapies in only those patients who express that specific lesion in their tumor. Indeed, three recent studies have shown that in RAS-mutant cells, BRAF inhibitors can actually activate the MAPK pathway through formation of BRAF-CRAF heterodimers or CRAF-CRAF homodimers to drive tumor progression (3335), highlighting that patients with RAS-mutant tumors should not be treated with BRAF-selective drugs.

We have developed an approach to investigate kinase inhibitor specificity in vitro and in vivo with the use of gatekeeper mutants. Questions about drug selectivity can be addressed by comparing responses of tumor cell lines that do or do not express a given oncogene, but an obvious advantage of our approach is that the analysis can be conducted in vivo within a representative tumor microenvironment, providing an important tool for in vitro and in vivo assessment of inhibitor specificity. Our data also point to potential mechanisms of resistance. Although kinase inhibitors can mediate excellent clinical responses, resistance can occur through acquisition of secondary mutations in residues such as the gatekeeper that block drug binding (2830, 3336), and our data provide formal proof that resistance to BRAF drugs can be mediated through this mechanism. Often, the identification of the clinical target of a specific drug is only confirmed by the emergence of mechanism-based resistance in the clinic, but our approach offers a powerful means to investigate mechanism during the drug discovery process. Clearly, this approach could also be used to test or design second-generation compounds to identify those that can tolerate gatekeeper residue changes and be used in patients whose tumors acquire resistance through this mechanism.

Materials and Methods

Cell culture and reagents

Cell lines were cultured in Dulbecco’s modified Eagle’s medium or RPMI 1640, as appropriate, supplemented with 10% fetal bovine serum (Invitrogen). The following antibodies were used: Myc, phospho-MEK1/2 (Cell Signaling Technologies), MEK1 (BD Biosciences), phospho-ERK1/2 (Sigma), Myc (Abcam), and ERK2 (Santa Cruz Biotechnology). Sorafenib was synthesized as described (6), and PLX4720 was synthesized by published routes (13). RAF265 (CHIR265) was purchased from American Custom Chemicals, and SB590885 was from Symansis. All drugs were dissolved in dimethyl sulfoxide (DMSO).

Expression and purification of BRAF proteins

Recombinant baculoviruses were generated using the Bac-to-Bac system according to the manufacturer’s instructions (Invitrogen). Baculovirus-infected SF9 cells (2 × 106/ml) were harvested after 72 hours, and the histidine-tagged full-length BRAF proteins were purified using Ni-NTA affinity chromatography (Qiagen).

BRAF kinase assays

A DELFIA-based (Perkin Elmer) 96-well assay was used to measure BRAF kinase activity as described in the presence of 100 μM ATP (37). Assays were performed in duplicate with an 11-point concentration response, and IC50 values were determined using GraphPad Prism software (GraphPad Software). IC50 values are the mean of three independent assays.

Transfection procedures and BRAF immunoprecipitation kinase assays

The expression constructs for G12VRAS, wild-type BRAF, and V600EBRAF have been described (38, 39). Gatekeeper residue mutants of BRAF were generated by polymerase chain reaction (PCR) mutagenesis and confirmed by automated dideoxy sequencing. COS7 cells were transfected using Lipofectamine (Invitrogen) as described (38) and harvested after 48 hours for quantitative Western blotting using the Odyssey infrared imaging system (Li-Cor Biosciences). For all Western blots, test and loading control antibodies were applied to the same gels to ensure consistent loading. HEK293 stable lines were established by transfecting as described (38), and then stable lines were selected by limiting dilution and selection in G418 (4 mg/ml). Ba/F3 stable cell lines were generated by transfecting 2 × 106 cells with 2 μg of plasmid DNA using the Nucleofector Kit V (Amaxa). Clones were selected in G418 (1 mg/ml) and for IL-3–independent growth. Expression of the appropriate BRAF mutants was confirmed by automated sequencing of the PCR-amplified BRAF expression construct. Immunoprecipitation kinase assays were performed using GST-MEK, GST-ERK, and myelin basic protein (MBP) as sequential substrates as described (15, 38).

Proliferation assays

Cell proliferation was assessed by measuring tritium-labeled thymidine incorporation. Ba/F3 cells (1 × 104 to 5 × 104) were seeded into the wells of 96-well plates, and compounds were added to the desired concentration. After 20 hours, 0.08 μCi of [3H]thymidine (GE Healthcare) was added to each well, and after a further 4 hours, the cells were captured onto Multiscreen glass fiber 96-well plates (Millipore) and washed twice with phosphate-buffered saline and twice with methanol using a vacuum manifold. Microscint 20 (25 μl) (Perkin Elmer) was added to the wells before counting on a TopCount NXT (Perkin Elmer). Assays were performed in quadruplicate with 10-point dilution series, and IC50 values were calculated using GraphPad Prism software. Values reported are the mean of three independent assays. Growth inhibition using the sulforhodamine B assay was as described (40).

Molecular modeling

The crystal structures of BRAF in complex with sorafenib (15) [Protein Data Bank (PDB) code 1UWJ], PLX4720 (13) (PDB code 3C4C), and SB590885 (12) (PDB code 2FB8) have been reported. After removing the ligand from these structures, the T529M, T529N, and T529I gatekeeper substitutions were generated using MOE v.2006.08 (Chemical Computing Group). To release possible strains introduced by altering the T529 residue, we subjected the models to an energy minimization within a radius of 9 Å from the mutation site using the AMBER99 (41) force field, and the favored rotamers of the gatekeeper residues were evaluated using the rotamer explorer utility of MOE. The inhibitors were then docked using GOLD version 3.1.1 (42). Sorafenib and RAF265 were then docked onto structures modeled starting with UWJ, PLX4720 onto model starting with 3C4C, and SB590885 onto model starting with 2FB8. During the docking runs, the gatekeeper side chains were allowed to rotate according to the torsion angles previously calculated. Partial charges of the ligands were derived using the Charge-2 CORINA 3D package in TSAR 3.3, and their geometries were optimized using the COSMIC module of TSAR. The calculations were terminated if the energy difference or the energy gradient was smaller than 1E–005. Ten docking solutions were generated per docking run with GOLD, and the best three were stored for analysis. To validate this approach, we docked the same set of ligands on the wild-type protein structures subjected to an energy minimization of 9 Å around the gatekeeper residue, and binding modes were compared with those observed by crystallography. In all cases, the docking poses faithfully reproduced the binding modes of sorafenib, SB590885, and PLX4720 in their x-ray structures.

In vivo efficacy

Nude mice were injected subcutaneously with 5 × 106 V600EBa/F3 or 2 × 107 T529N,V600EBa/F3 cells. Tumors were allowed to establish for 7 days, after which mice were size-matched and allocated to six per treatment group. Treatment was by intraperitoneal daily dosing with either vehicle (5% DMSO, 95% water), sorafenib (50 mg/kg), or PLX4720 (20 mg/kg). Tumor size was determined by caliper measurements of tumor length, width, and depth every 3 to 4 days. Tumor volume was then calculated using the following formula: volume = 0.5236 × length × width × depth (mm). Experiments were conducted in accordance with the UK Home Office regulations and UK Coordinating Committee on Cancer Research Guidelines (43). For biomarker studies, tumors were allowed to reach a volume of ~200 mm3 and then treated as above for 4 consecutive days. Tumors were harvested 1 hour after the last dose and fixed in neutral-buffered formalin overnight at room temperature. Tumor sections were stained with phospho-MEK1/2 antibody and counterstained using 0.5% methyl green (Sigma) as described (44). To quantify phospho-MEK staining in the tumors, we recorded the percentage of cells in each section that stained with low (scored 1), intermediate (scored 2), and high (scored 3) intensity for each of two V600EBRAF and two T529N,V600EBRAF tumors treated with vehicle, PLX4720, or sorafenib (12 tumors in total). The relative intensity of staining for phospho-MEK was calculated using the formula [(% cells × score 1 staining) + (% cells × score 2 staining) + (% cells × score 3 staining)] and then normalized to the vehicle control. The samples were scored blind, and error bars represent SEM.

Supplementary Material

www.sciencetranslationalmedicine.org/cgi/content/full/2/35/35ra41/DC1

Fig. S1. Chemical structures of sorafenib (A), PLX4720 (B), RAF265 (C), and SB590885 (D).

Fig. S2. Alignment of the protein sequences of BRAF, EGFR, ABL, and KIT.

Fig. S3. Assessment of MAPK pathway activity following transfection of COS7 cells with expression constructs for BRAF proteins.

Fig. S4. Small-molecule inhibitors of V600EBRAF bind in close proximity to the gatekeeper residue.

Table S1. Inhibition of mutant BRAF kinase activity by sorafenib, RAF265, PLX4720, and SB590885.

Table S2. Inhibition of V600EBRAF- and T529N,V600EBRAF-driven DNA synthesis in Ba/F3 cells.

Footnotes

  • Citation: S. Whittaker, R. Kirk, R. Hayward, A. Zambon, A. Viros, N. Cantarino, A. Affolter, A. Nourry, D. Niculescu-Duvaz, C. Springer, R. Marais, Gatekeeper mutations mediate resistance to BRAF-targeted therapies. Sci. Transl. Med. 2, 35ra41 (2010).

References and Notes

  1. Funding: Wellcome Trust (refs: 071487/Z/03/A and 080333/Z/06/Z), Cancer Research UK (refs: C309/A2187, C107/A3096, and C107/A10433), Institute of Cancer Research, and Isle of Man Anti-Cancer Association. Author contributions: S.W. and R.M. conceived, designed, and analyzed the data; S.W., R.K., R.H., A.A., N.C., and A.V. performed the biological experiments; A.N., D.N.-D., and C.S. synthesized the drugs; C.S. and R.M. provided funding; and S.W., A.Z., C.S., and R.M. wrote the manuscript. Competing interests: The authors have no competing interests.
View Abstract

Navigate This Article