Cancer cells can develop an 'addiction' to the drugs they are treated with, so that they need the drugs to survive. Analysis of the underlying mechanism reveals a potential clinical strategy for harnessing this phenomenon. See Letter p.270
As cancer cells become resistant to drugs that target cancer-specific signalling pathways, an intriguing phenomenon called drug addiction sometimes develops1. This occurs when cells that are resistant to drug treatment become dependent on the drug. On page 270, Kong et al.2 provide insights into the mechanisms underlying drug addiction in cancer. Their findings have major implications for the use of targeted tumour therapies.
The authors studied cultured cell samples of a type of human skin cancer called melanoma. They treated the cells with the drug dabrafenib, which is used to target the BRAF protein in cancers that have a cancer-causing mutation in the BRAF gene. Alternatively, the authors treated the cells with both dabrafenib and the drug trametinib, which targets an enzyme in the BRAF pathway called MEK. They identified drug-addicted cells, which died if they did not receive drug treatment.
To find the mechanism responsible for drug addiction in this context, Kong and colleagues used a genetic-engineering technique to delete genes in the cells. They withdrew dabrafenib from the engineered cells, then analysed the surviving cells for genes whose deletion prevented the cell death or growth arrest that usually occurs in drug-addicted cells when the drug is withdrawn. These experiments enabled the authors to identify two genes that they verified were involved in the drug-addiction process. The genes encode the proteins ERK2 and JUNB, which have roles in the BRAF-mediated signalling pathway. Then, using a panel of drug-resistant cell lines and in vitro and in vivo experiments, Kong et al. demonstrated that genetic depletion of MAPK1 (which encodes ERK2), JUNB or FOSL (which encodes JUNB's binding partner protein, FRA1) blocked the death of drug-addicted cells.
The authors uncovered an exquisite specificity of action of the proteins involved. For example, depletion of ERK2 prevented cell death associated with drug addiction, but depletion of ERK1 had no effect on this process. This is consistent with the observation that when the drug was withdrawn from drug-addicted cells, ERK2 but not ERK1 was hyperactivated, even though both act in the same signalling pathway.
The drug-addiction pathway was the same regardless of how drug resistance had developed in a particular cell, for example by increased activation of the ERK pathway, or by an increase in the number of copies of BRAF. So the authors' findings might have relevance for the treatment of many people who have drug-resistant melanoma.
To determine how drug-addicted cells die when the drug is withdrawn, the authors sequenced RNA in this type of dying cell. As the cells died, the expression of genes associated with cell proliferation decreased, whereas there was increased expression of genes associated with metastasis — the ability of cells to spread to new locations in the body. This finding links drug addiction to phenotypic switching, an ability of melanoma cells to switch between a proliferative state and an invasive cellular state associated with metastasis3.
Phenotypic switching underlies the plasticity of melanoma cells and their ability to adapt to their microenvironment. The cellular state of melanoma cells can be identified by monitoring the expression of the E-cadherin protein or other proteins that are characteristic of a change — known as the EMT — that occurs when epithelial cells transition to become mesenchymal cells4. Phenotypic switching results in a cellular state in which cells have an increased probability of becoming metastatic, and the EMT change is part of the process that enables cells to metastasize. The EMT is regulated4 in melanoma by the transcription factor protein MITF, which is the master regulator of the melanocyte-cell lineage5 from which melanoma arises.
Low levels of MITF are associated with an invasive metastatic state, whereas high levels are associated with proliferation4. The authors found that, in drug-addicted cells containing high levels of MITF, the MITF levels collapse on drug removal and the cells adopt an invasive form. When the authors used genetic engineering to prevent a decrease in MITF levels when drug was removed, the cell death associated with drug withdrawal was reduced. However, the role of MITF in this cell-death process is unclear, given that low levels of MITF expression are not known to cause cell death4.
Kong and colleagues' work has clinical relevance in that, when the drug treatment was stopped, they identified a window during which treatment of melanoma cells grown in vitro with the chemotherapeutic agent dacarbazine could boost cell death (Fig. 1). The authors propose that stopping the use of BRAF and MEK inhibitors, and starting the use of dacarbazine, would be more effective than merely halting all drug treatments in what are called 'drug-holiday' approaches aimed at promoting cell death by harnessing drug addiction.
The timing of drug treatment would be crucial. MEK inhibitors, in particular, have extremely long half-lives (up to 10 days) and, given that low levels of inhibition of this pathway could prevent cell death caused by drug addiction, correctly timed drug administration would be essential to achieve optimal benefit. There is clinical evidence6 that retreatment with the targeted therapy can be effective after a drug holiday that was implemented as a result of disease progression or toxicity. The phenotypic switch from proliferative to invasive states might reset the sensitivity of the cells to targeted therapies, although this needs to be investigated.
The authors also tested another type of cancer called non-small cell lung cancer. They found that drug withdrawal decreases cell growth and induces an EMT cellular transition in a drug-resistant sample of these cancer cells grown in vitro and treated with an inhibitor of the EGFR protein (mutations in this protein are often associated with lung cancer). Again, this drug-addiction phenomenon was prevented by depletion of ERK2 but not of ERK1. These data suggest that drug addiction is a general feature associated with treatment by drugs that target proteins, such as EGFR, in the MAPK pathway. It will be interesting to discover whether drug addiction extends to therapies that target other signalling pathways, and to identify the key pathway components responsible.
The exciting work reported by Kong et al. consists mainly of in vitro studies, so in vivo work will be needed to examine whether the tumour microenvironment affects drug addiction. Their study also raises many fascinating questions: for example,why does ERK2 but not ERK1 mediate drug addiction, even though both are key components of the MAPK pathway. A complex network of feedback mechanisms regulates this pathway, and the role of these regulatory mechanisms in drug addiction should be investigated. The key to translating the authors' findings into clinical use will be determining whether other drugs work better than dacarbazine if used together with drug withdrawal to target drug-addicted cells. It would also be useful to investigate the mechanisms that underlie this type of synergy in greater detail. The improvements in our understanding of drug addiction that come from this latest work clearly offer a promising start in exploring the therapeutic opportunities that might arise from this Achilles heel of cancer.
Das Thakur, M. et al. Nature 494, 251–255 (2013).
Kong, X. et al. Nature 550, 270–274 (2017).
Hoek, K. S. et al. Cancer Res. 68, 650–656 (2008).
Müller, J. et al. Nature Commun. 5, 5712 (2014).
Garraway, L. A. et al. Nature 436, 117–122 (2005).
Schreuer, M. et al. Lancet Oncol. 18, 464–472 (2017).
About this article
Cite this article
Lee, R., Marais, R. Tumours addicted to drugs are vulnerable. Nature 550, 192–193 (2017). https://doi.org/10.1038/nature24148
Science Signaling (2017)