Concise Review: Cancer Cells Escape from Oncogene Addiction: Understanding the Mechanisms Behind Treatment Failure for More Effective Targeting


  • Francesca Pellicano,

    1. Paul O'Gorman Leukaemia Research Centre, Institute of Cancer Sciences, College of Medical, Veterinary & Life Sciences, University of Glasgow, Glasgow, United Kingdom
    Search for more papers by this author
  • Leena Mukherjee,

    1. Paul O'Gorman Leukaemia Research Centre, Institute of Cancer Sciences, College of Medical, Veterinary & Life Sciences, University of Glasgow, Glasgow, United Kingdom
    Search for more papers by this author
  • Tessa L. Holyoake

    Corresponding author
    1. Paul O'Gorman Leukaemia Research Centre, Institute of Cancer Sciences, College of Medical, Veterinary & Life Sciences, University of Glasgow, Glasgow, United Kingdom
    • Correspondence: Tessa L. Holyoake, MBChB, MRCP, MRCPath, PhD, FRCP, FRCPath, RSE, Gartnavel General Hospital, 21 Shelley Road, Glasgow, G12 0ZD, U.K. Telephone: +141–301-7880; Fax: +141–301-7898; e-mail:

    Search for more papers by this author


Oncogene addiction describes the dependence of some cancers on one or a few genes for their survival. Inhibition of the corresponding oncoproteins can lead to dramatic responses. However, in some cases, such as chronic myeloid leukemia (CML), a disease characterized by the presence of the abnormal fusion tyrosine kinase BCR-ABL, cancer stem cells may never acquire addiction to the oncogene that drives disease development. The suggested mechanism(s) for treatment failure include a quiescent stem cell population capable of reinstating disease, high levels of oncoprotein expression, or acquired mutations in the oncogene. In this review, we discuss the evidence for oncogene addiction in several solid tumors and their potential escape mechanism(s) with a particular focus on CML stem cells. Stem Cells 2014;32:1373–1379


Carcinogenesis is a multistep process during which cells, over years, progressively acquire multiple genetic and epigenetic abnormalities. Carcinomas often display mutations in multiple oncogenes and tumor suppressor genes, epigenetic abnormalities in expression of genes, and chromosomal abnormalities [1]. However, despite such complexity, there is an apparent dependence of some cancers on one or a few genes for continued cell proliferation and maintenance of the malignant phenotype in a process defined as oncogene addiction. Abnormal gene expression in cancer cells alters the relative importance of particular survival pathways compared with normal, and proteins encoded by these genes often have multiple roles in complex and interacting networks. Abnormal dependence on the constitutive activation of certain genes can result in cancer cells being more susceptible to their inhibition than in normal cells and can increase their sensitivity to chemotherapeutic agents [2, 3].

However, in keeping with this genomic instability, inhibition of one pathway can lead to the emergence of de novo mutations that allow escape from oncogene addiction and cell survival. This evolution of cells carrying these new mutations is evidence for the dependence of these cancer cells on particular oncogenes for survival [4]. It also suggests that tumors may not always rely on the same oncogene for survival, making treatment difficult and cure elusive. In this review, we discuss the evidence for oncogene addiction in several cancers with a particular focus on chronic myeloid leukemia (CML) stem cells.

Oncogene Addiction in Cancer

There are three proposed mechanisms of oncogene addiction [4] (Fig. 1):

  1. Genetic streamlining: With the constant genetic evolution of cancer cells, nonessential pathways are inactivated, meaning dominant addictive pathways are not ameliorated by compensatory signals. Inactivation of dominant pathways then has catastrophic consequences, leading to cellular collapse.
  2. Oncogenic shock: Addictive oncoproteins are responsible for prosurvival and proapoptotic signals. Inactivation of a critical oncogene causes cell death due to differential attenuation rates of these and a dominance of proapoptotic signals.
  3. Synthetic lethality: The continued activity of two genes is required for cell survival. In the context of cancer, when one is inactivated by pharmacological inhibition or manipulated silencing, cell death occurs due to the pre-existence of a loss of function cancer-associated mutation in the second pathway.
Figure 1.

Schematic representation of how cancer cells may (A) escape oncogene addiction or (B) alternatively be oncogene addicted. Some cells, such as CML stem cells, may not be oncogene addicted and tumor progression results despite treatment.

Most examples of oncogene addiction fall within the first two categories. The majority of gastrointestinal stromal tumors (GISTs) strongly express an activating KIT mutation or activating mutations in PDGFRα (platelet-derived growth factor receptor) and their abrogation leads to a dramatic response [5]. However, treatment failure in GIST may occur due to novel point mutations, KIT genomic amplification with overexpression of the KIT oncoprotein, activation of an alternate receptor tyrosine kinase protein accompanied by loss of KIT oncoprotein expression, or functional resistance [6].

Similarly, 40%–60% of cutaneous melanomas carry mutations in the oncogene BRAF leading to constitutive activation of downstream signaling through the Mitogen-activated protein kinases (MAPK) pathway [7, 8]. Vemurafenib was developed to target this specific mutation with response rates in BRAF mutated melanoma of 48% and overall survival rates of 84% at 6 months [7]. However, resistance mechanisms have emerged. In BRAF(V600E) mutated melanomas, levels of RAS activation are low and a subset of cells resistant to vemurafenib expresses a 61-kDa variant form of BRAF(V600E), p61BRAF(V600E), which lacks the RAS-binding domain and enhances protein dimerization, conferring resistance. Crucially, a mutation that abolishes the dimerization of p61BRAF(V600E) restores its sensitivity to vemurafenib [9].

Another example of oncogene addiction is given by Her2. Her2 is overexpressed in 25%–30% of breast cancer—a feature that correlates with adverse prognosis [10]. Treatment with a humanized antibody to the extracellular domain of Her2, trastuzumab, was one of the first targeted oncological therapies but displays response rates of only 30% likely due to activation of compensating pathways, such as Phosphatidylinositide 3-kinases (PI3K)-dependent pathways [10].

In the case of non-small cell lung cancer (NSCLC), approximately 13% of patients display an activating mutation in the epidermal growth factor receptor (EGFR) [11]. These patients respond to kinase inhibition with compounds targeting the catalytic activity of EGFR (erlotinib, gefitinib), with response rates of 50%–70% and significant increases in progression-free and overall survival [12]. However, primary resistance to EGFR inhibition exists due to coexisting EGFR mutations, and secondary resistance occurs with mutations developing upon EGFR blockade and amplification of the MET oncogene, which activates the PI3K pathway [12, 13]. In addition, in a small number of cases of NSCLC, an inversion within the short arm of human chromosome 2 results in constitutive activation of the fusion kinase EML4-ALK [14]. A targeted EML4-ALK kinase inhibitor, crizotinib, has led to response rates of 61% in phase 2 trials with 2-year overall survival rates of 54% [15]. However, data are emerging of two point mutations that have developed within 5 months [14].

Finally, PARP inhibition together with a BRCA mutation is synthetically lethal by conferring faulty homologous repair mechanisms [16]. Early phase trials in ovarian and breast cancer with the PARP inhibitor olaparib showed promise, with a greater proportion of BRCA positive ovarian cancer patients achieving stable disease [17]. However, in patients with “triple negative” breast cancer that does not express oestrogen, progesterone receptors or Her2/neu, additional PARP inhibition did not show any survival benefit in phase 3 trials, despite earlier trials showing increased response [18]. This suggests that tumor heterogeneity may lessen the importance of this driver oncogene for a large population of these patients.

Table 1 highlights additional in vitro and in vivo models that suggest oncogene addiction together with targeted approaches used either in a clinical scenario or within the laboratory. Findings presented are reviewed in Sharma et al. [19].

Table 1. Models suggesting oncogene addiction using either clinically approved drugs, lab-based and/or transgentic models
OncogeneCancer typeIn vitro modelsIn vivo modelsClinically approved treatments
KITGastrointestinal stromal tumourKit inhibitor induced apoptosis in GIST cellsKIT inhibitor showed anti-tumour activity in mouse xenografts imatinib
PDGFRGlioma, gastrointestinal stromal tumourPDGFR inhibitors induced apoptosis in rat glioma cellsPDGFR inhibitors exhibited anticancer activity in miceimatinib
BRAFMelanoma, thyroid, colorectalSuppression of BRAF(V599E) in human melanoma cells abrogated transformationBRAF inhibition reduced tumour formation in micesorafenib
Her2Breast, ovarian, non-small cell lung cancer (NSCLC)RNAi (RNA interference) in Her2 amplified breast carcinoma cells lines inhibited growth and DNA synthesis by 60%Conditional neu activation in mammary mouse model led to reversible pulmonary metastasistrastuzumab
EGFRNSCLC, gliblastoma, colon, pancreasRibosome-mediated suppression of aberrant EGFR expression inhibited tumour growthMonoclonal antibodies inhibited growth of tumour xenografts with EGFR mutationgefitinib
ALKAnaplastic large cell lymphoma, NSCLCALK inhibitor inhibited growth of ALK-dependent cell lineALK inhibitor blocked lymphomagenesis in micecrizotinib
VEGFTumour angiogenesis Monoclonal antibody against VEGF and tyrosine kinase inhibitors inhibited tumour growth in vivobevacizumab
OncogeneCancer typeIn vitro modelsIn vivo modelsLab-based drugs/transgenic
HRAS KRASPancreas, thyroid, colon, NSCLCFarnesyl-transferase inhibitors blocked growth of RAS dependent tumoursFarnesyl-transferase inhibitors blocked growth of RAS-dependent tumours in nude mice. Lung adenocarincoma reversibly induced by activation of KRAS transgeneL-739,749 transgenic mice expressing K-RAS4b (G12D)
MYCLymphoma, leukaemiaRNAiof MYC in haemopoeitic cells led to inhibition of cell growth and DNA synthesisMYC expression in transgenic mouse led to haematological malignancy that was reversed upon MYC inhibitionTetracycline induced MYC proto-oncogene
METGastric, NSCLCMET inhibitor induced apoptosis in gastric cancer cell linesMET inhibitor inhibited tunour formation and angiogenesis in mouse kung cancer xenograftsPHA665752
FGFR3MyelomaMyeloma cells were sensitive to an FGFR3 inhibitorFGFR3 inhibitor inhibited tumour growth of mouse xenograftsPRO-001
AURORA KINASEColon, breastInhibitor reduced colony formation in human cancer cellsAnti-tumour activity inhibited tumour growth in human xenograft modelJNU-7706621
RETThyroidRET inhibitor induced cell death in papillary thyroid cancer cellsRET inhibitor induced dose-dependent growth inhibition in RET-expressing human xenograftsNVP-AST487

All the above examples—KIT, BRAF, Her2, EGFR, and PARP, provide evidence for the existence of oncogene addiction. However, as described later in the review, this is not always as clear for other forms of cancer, such as CML.

Are CML Stem Cells Really Oncogene Addicted?

CML is a myeloproliferative disease arising at the hemopoietic stem cell (HSC) level due to the expression of BCR-ABL, an oncogene encoding a constitutively active tyrosine kinase and occurring as the result of the chromosomal translocation t(9;22)(q34;q11).

The discovery of the constitutively active fusion protein BCR-ABL in CML was among the first described examples of oncogene addiction and the first rationally developed tyrosine kinase inhibitor (TKI) imatinib, was approved in 2001 after inducing dramatic response rates in blast crisis CML patients in phase 1 trials. Until recently, CML and BCR-ABL have been considered icons for the phenomenon of oncogene addiction in human cancer. The belief that BCR-ABL expression was the sole initiating and driving event in every leukemic cell leading to CML development has promoted the development of several highly effective TKI [20]. Current therapy for CML involves imatinib, dasatinib, nilotinib, bosutinib, and ponatinib, all rationally designed TKI with variable potency and activity against wild type and TKI-resistant mutants of BCR-ABL [21-28].

The generation of several retroviral and transgenic mouse models of CML has contributed to our understanding of the nature of the disease and how it can be targeted [29]. Retroviral models have demonstrated that BCR-ABL protein expression is sufficient to transform bone marrow cells which, once transplanted, are able to reconstitute the disease [30]. Similarly, transgenic models have confirmed that BCR-ABL expression drives CML development and maintenance. In these models, expression of BCR-ABL in hemopoietic progenitor cells caused a CML-like disease, but subsequent suppression of the oncogene led to apoptosis and complete remission [31]. Similar behavior was observed when a fragment of the 3′ enhancer of the murine SCL gene was used to direct expression of BCR-ABL in HSC and myeloid progenitors [32]. In this model, CML stem cells were necessary to induce CML-like disease in syngeneic transplant recipient mice [33]. Finally, a transgenic model expressing BCR-ABL under the control of the Sca-1 promoter has also been generated to confine BCR-ABL expression to the HSC compartment [34]. Here stem cells caused leukemia in secondary recipients, but the disease did not respond to imatinib.

It is not entirely clear whether BCR-ABL is always the first hit and only driver of the disease or if additional genetic and epigenetic changes are essential for this process and do contribute to drug resistance and disease progression [35, 36]. It is now evident that TKI have been very successful in inducing rapid hematological and cytogenetic responses in the majority of chronic phase CML patients [37], but while TKI show an impressive efficacy against proliferating BCR-ABL positive leukemic cells, they fail to eradicate quiescent and primitive CML stem cells, even at high concentrations [21, 22, 38]. This explains why TKI do not eliminate BCR-ABL transcripts in the majority of patients treated, suggesting the persistence of a minimal residual disease, likely to be restricted to the stem cell level [21, 39]. The formation of a reservoir of these stem cells could easily explain the presence of residual disease and the rapid relapse observed in most patients who discontinue TKI treatment, even if an apparently stable cytogenetic and molecular remission was previously achieved for up to 5 years [40-42]. In support of this, a trial named “non-randomised Stop IM (STIM),” has shown that when imatinib treatment was discontinued in patients who had previously reached complete molecular remission for more than 2 years, of the 69% of patients that had a complete follow-up, 61% relapsed [43].

The mechanism(s) through which CML stem cells are, or become, inherently insensitive to BCR-ABL inhibition is unknown. Although in acquired TKI resistance almost 50% of cases demonstrate the presence of mutations in the kinase domain, our data have shown no detectable mutations in CML stem cells after in vitro drug exposure [44, 45]. In addition, the rapid relapses following TKI withdrawal in patients were shown to happen in the presence of wild type and not mutant BCR-ABL.

Recently, parallel studies carried out in our, Brian Druker's and Michael Deininger's laboratories have shown that, although most mature CML cells are oncogene addicted, CML stem cells survival is independent from BCR-ABL kinase activity as they proliferate despite long-term kinase inhibition [35, 36]. The results derived from these studies in primary human and murine CML stem and progenitor cells suggest that CML stem cells may not be oncogene addicted, or at least not addicted to the oncogene kinase activity.

Initially, we used the inducible transgenic mouse model of CML in which p210-BCR-ABL expression is targeted to stem and progenitor cells of murine bone marrow using the “tetracycline-off” system [32]. On tetracycline withdrawal, BCR-ABL expression is switched on and mice develop a CML-like disease within a few weeks. The leukemic stem cells of these mice are transplantable and the disease can be reverted after tetracycline or TKI treatment [33]. In this model, CML stem cells in which BCR-ABL expression was first induced and then shut off for sustained periods were able to survive in vivo and induce a CML-like disease in secondary recipients upon BCR-ABL re-expression. These results suggest that the leukemic stem cells retain their transplant ability even if the expression of the oncogene was removed. In primary human CD34+ CML progenitor cells, partial knockdown of BCR-ABL by a lentiviral system did not affect the survival of the small fraction of primitive leukemic cells, which continued to grow in long term culture initiating cell assays and expanded despite the absence of growth factors in the culture medium. Similar results were observed when complete inhibition of BCR-ABL kinase was achieved by dasatinib treatment in the absence of growth factors. Even by combining BCR-ABL lentiviral knockdown with dasatinib we were unable to kill these cells. Therefore, these data have shown that although BCR-ABL kinase activity is inhibited in CML stem cells by TKI, the cells survive the treatment. This is not meant to infer that TKI are inert against CML stem cells. Indeed, TKI activate a number of critical pathways in CML stem cells, exerting an antiproliferative activity in the cells and making them even more quiescent and therefore more difficult to eradicate [22, 46-48].

Comparable conclusions were made by Druker and Deininger, who also reported that human CML stem cells (in this case the more primitive CD34+38133+ population) are not addicted to BCR-ABL as their survival was not impaired upon kinase inhibition [36]. In this study, it was observed that TKI treatment reduced the growth of CML stem/progenitors, but interestingly the addition of cytokine support allowed these cells to survive in the complete absence of BCR-ABL activity, resembling a normal stem cell phenotype. Complete inhibition of BCR-ABL activity with TKI was shown by several approaches, including Fluorescence-activated cell sorting (FACS), intracellular staining, and immunoblot analysis of phospho-CRKL, a downstream protein of BCR-ABL.

On the other hand, in addition to ABL kinase, other domains in the BCR-ABL protein may have a role in leukemogenesis and could explain why CML stem cells survive despite long-term kinase inhibition. Although ABL kinase activity is essential for CML development in vivo, the kinase domain is not sufficient to induce the full disease in vivo when expressed on its own, suggesting that multiple domains within the BCR-ABL protein are required to reproduce a CML-like disease [49-51]. Within these domains are the GRB2 SH2-binding site, the SH3 domain, and the amino-terminal coiled-coil oligomerization domain. They regulate BCR-ABL kinase activity itself and are essential to connect BCR-ABL to other downstream signaling pathways (reviewed in [29]). Recently, together with Jiang's group, we have shown that uncoupling the oncoprotein Abelson helper integration site-1 (AHI-1) from BCR-ABL and JAK2 led to increased sensitivity of CML stem cells to TKI [52]. By inhibiting JAK2 with the drug TG101209, direct and AHI-1-mediated interactions with BCR-ABL were blocked. When TG101209 was used in combination with imatinib enhanced cytotoxicity was achieved, suggesting that BCR-ABL kinase inhibition plus inhibition of AHI-1-mediated interactions could represent a rationally designed combination treatment for eliminating CML stem cells.

BCR-ABL kinase activity alone is unlikely to be the main molecular mechanism for leukemic stem cell survival, which may instead be maintained by activation of key survival pathways that are inherent to normal stem cells. Therefore, to investigate signaling pathways activated by BCR-ABL independently from the kinase activity, Chen et al. have identified pathways that are activated by BCR-ABL but are insensitive to inhibition by imatinib. Within these pathways, the arachidonate 5-lipoxygenase (5-LO) gene (Alox5) was identified as a critical regulator of CML stem cells [53]. The absence of Alox5 blocked BCR-ABL-dependent development of CML in vivo and treatment of mice with a 5-LO inhibitor impaired CML stem cell function. Similarly, inhibition of the tumor suppressor protein phosphatase 2A (PP2A) is caused by BCR-ABL expression, but not kinase activity, through recruitment of JAK2. Of relevance, persistence of CML stem cells is mediated by the inhibition of PP2A and expression of BCR-ABL and drugs targeting PP2A are able to target CML stem cells [54].


In several cancers, the presence of a driver oncogene is evident. Nevertheless, inhibition does not achieve cure for patients due to either upstream proteins re-activating the same downstream pathway, or activation of parallel pathways. Resistance may develop due to selection of a resistant clone that either pre-exists or evolves over time. Also, epigenetic modulation of gene expression in cells mediates heterogeneous responses to drug treatment allowing some cells to become drug resistant.

Recent studies have suggested that CML stem cells are inherently insensitive to TKI, and therefore it is unlikely that CML is going to be cured using TKI monotherapy. It is appropriate to think that combination therapy with agents able to induce apoptosis in CML stem cells in a selective manner will be required for disease eradication [21, 22, 43, 55]. This scenario now looks to be achievable with growing evidence that CML stem cells use survival signals other than BCR-ABL kinase to maintain their viability in the presence of TKI. Also, although BCR-ABL may be the initiating oncogene in CML, additional genetic and/or epigenetic events (including aberrant methylation of ABL1, p15, ATG16L2, and DAPk1) could be responsible for maintaining the disease independently of, or in addition to, the initial transformation caused by the chromosomal translocation [35, 46, 56-60]. As with many other forms of cancer, CML development and progression does not result from the activity of a single driver but may require tightly regulated cooperation of several mutations at the genetic level [61].

The resistance of cancer cells to treatment is likely to be caused by the synergistic/cooperative activity of several signaling pathways and molecules and, in the case of CML stem cells, this has already been shown by the tightly regulated network that includes TGFβ, AKT, FOXO3a, and BCL6 [36, 46-48]. It will be important to explore targeting of novel and potentially multiple co-operating genes deregulated in cancer. Several laboratories have now approached the targeting of cancer by investigating synergistic mutations required for malignant transformation reflected in downstream gene networks or signatures [62, 63]. This scenario offers the possibility of developing novel therapeutic approaches, aiming to use completely different strategies, such as targeting stem cell self-renewal or disrupting interactions with the microenvironment. Similar contributions to the understanding of the complex biologic mechanisms of cancer will also be provided by the use of combined computational and experimental systems biology approaches leading to the uncovering of unknown active networks. It is important to consider that chemotherapy may cause evolution of the cancer cells leading to several different phenotypes whose characteristics will be dependent on the surrounding landscape [64]. In conclusion, it seems that cancers are never truly “oncogene addicted”, and that understanding the mechanisms behind treatment failure will allow combination treatments to facilitate more effective cell kill.


We acknowledge CR-UK (A11008 and C596/A17196), Wellcome Trust (101703/Z/13/Z), and ELF (F217 and ELF 6/29/1) for funding.

Author Contributions

F.P.: conception and design, collection and/or assembly of data, manuscript writing, and final approval of manuscript; L.M.: collection and/or assembly of data, manuscript writing, and final approval of manuscript; T.L.H.: manuscript writing and final approval of manuscript.

Disclosure of Potential Conflicts of Interest

The authors do not have anything to disclose.