A drug for treating melanoma is ineffective in colorectal cancers that have the same causative mutation. Studies of how cells adapt to the drug reveal why this is so, and suggest combination therapies that may be more effective. See Letter p.100
Genetic mutations that disrupt fundamental cellular processes, such as cell division and differentiation, can lead to cancer. Mutations in the BRAF gene, which codes for the kinase protein BRAF, are one such example, and are detected in multiple cancers. In patients with advanced BRAF-mutant melanomas, treatment with the BRAF-inhibiting drug vemurafenib significantly prolongs survival1. However, vemurafenib is ineffective2 in patients with colorectal cancers that harbour the same BRAF mutation — BRAF(V600E). On page 100 of this issue, Prahallad et al.3 provide an explanation for this lack of efficacy by showing that, in colorectal cancer cells, drug treatment leads to rapid activation of a protein that counteracts the drug's BRAF-inhibitory activity.
Prahallad and colleagues identified this molecular basis for resistance by screening for genes for which inhibition of expression leads to enhanced sensitivity of human colorectal cancer cells to vemurafenib. Their screen assessed all 518 human kinase-encoding genes and 17 additional kinase-related genes. Among the most potent 'hits' they identified was the gene coding for the epidermal growth factor receptor (EGFR). To test the biological significance of this finding, the authors treated BRAF(V600E)-mutant colorectal cancer cell lines with vemurafenib, and found that the drug induced hallmarks of EGFR activation.
EGFR acts earlier than BRAF, and other RAF proteins, in a cell signalling pathway called the ERK pathway. The authors hypothesize that inhibition of BRAF by vemurafenib relieves a negative feedback loop that keeps EGFR inactive. EGFR activates a protein called RAS, which, when activated, has been shown to cause vemurafenib resistance by inducing the formation of dimers of RAF proteins, against which the drug is ineffective (Fig. 1). EGFR also activates other cell signalling pathways, such as the PI3K/AKT pathway, and EGFR-mediated activation of these pathways may also contribute to vemurafenib resistance.
The identification of EGFR activation as the basis for vemurafenib resistance in colorectal cancer cells is consistent with a model of resistance that has emerged from studies of patients with melanoma who acquire resistance to this drug4,5,6. In BRAF-mutant melanomas, RAS activity is low, and consequently RAF dimer formation is low. Vemurafenib effectively inhibits the monomeric forms of BRAF6. However, perturbations that increase RAS activation (such as RAS mutation or activation of upstream kinases) promote the formation of RAF dimers, leading to drug resistance4,5,6.
In contrast to melanoma cells, EGFR is ubiquitously expressed in BRAF-mutant colorectal cancers, although it is inactive before drug treatment. Exposure to vemurafenib results in rapid EGFR activation and immediate drug resistance. EGFR-expressing colorectal cancer cells are thus 'primed' for resistance to BRAF inhibitors. This model of vemurafenib resistance is supported by another recent study7, which reports that vemurafenib treatment leads to increased EGFR-mediated RAS activity in BRAF-mutant colorectal cancer cells. This can be blocked by concurrent treatment with an EGFR inhibitor.
Drugs that target EGFR, such as the monoclonal antibody cetuximab, are ineffective in patients with advanced BRAF-mutant colorectal cancer8. However, when Prahallad and colleagues3 treated mice bearing grafts of human BRAF-mutant colorectal cancers with vemurafenib and cetuximab, the combination was significantly more effective than either drug alone. These results imply that EGFR and BRAF inhibitors, despite being ineffective in patients with BRAF-mutant colorectal cancer when used alone, may be more successful when administered in combination.
Prahallad and colleagues' paper complements a series of recent publications that have enhanced our understanding of the seemingly idiosyncratic activity of BRAF inhibitors in patients. Specifically, vemurafenib inhibits ERK-pathway activity in BRAF-mutant tumour cells, but paradoxically activates ERK signalling in normal cells and in tumour cells expressing normal BRAF9. The drug thus inhibits ERK activity in a BRAF-mutant-specific manner, an atypical property that accounts for its unusual toxicity profile. Approximately one-third of patients treated with vemurafenib develop a skin rash that differs markedly from the rashes associated with drugs that inhibit ERK signalling. Some vemurafenib-treated patients also develop hyperproliferative skin lesions — including keratoacanthomas and cutaneous squamous-cell carcinomas — that do not harbour BRAF mutations1. The emergence of these skin lesions is explained by two recent studies10,11 that found that most of the lesions harbour mutations in the genes encoding RAS proteins. Animal experiments have indicated that, as a result of the elevated RAS activity in these skin lesions, vemurafenib activates ERK signalling, which leads to cellular hyperproliferation.
Taken together, these findings3,4,5,6,7,8,9,10,11 have broad clinical implications. They clearly justify studies combining EGFR and BRAF inhibitors in patients with BRAF-mutant colorectal cancer. Furthermore, as vemurafenib activates ERK signalling and EGFR inhibitors have the opposite effect, this drug combination may prove less toxic than either drug alone, a result similar to the reduced toxicity observed in studies combining inhibitors of MEK (another ERK-pathway protein) and BRAF12.
Prahallad and colleagues' findings also highlight the potential for differences in efficacy of kinase-targeted therapies in tumours that share the same mutation but arise in different tissue types. For example, because EGFR is broadly expressed in epithelial cells, carcinomas may, as a group, be intrinsically less sensitive to BRAF inhibition than melanomas, which occur in cells of a different (neural crest) origin from carcinomas. This hypothesis awaits clinical validation in ongoing trials of BRAF inhibitors.
As efforts to characterize individual tumours on a molecular basis accelerate in the clinic, with the goal of personalizing cancer therapy, there will be a strong desire to treat patients who have a BRAF-mutant tumour with a BRAF inhibitor, regardless of the site of tumour origin. The findings reported by Prahallad et al. suggest that this approach may sometimes prove ineffective. Ideally, such efforts at personalized medicine should proceed within the context of clinical trials and should include molecular analysis of both pre- and post-treatment tumour samples, because such studies may aid the identification of additional cell-type-specific factors that affect drug responses and thereby facilitate the design of combination therapies.