Therapeutic targeting of biomolecules has become a dominant theme in modern day oncology. The success achieved with imatinib [1
], bevacizumab [2
] and a small number of other drugs [3
] has been of enormous clinical benefit. However, these outcomes are in stark contrast with the no less spectacular failures of many other targeted agents (e.g. gefitinib [6
], farnesyl transferase inhibitors [7
]). Agents targeting EGFR fall into a middle ground: some of them are extremely successful for specific types of cancer, and in a subset of patients. Better understanding of why some tumor types respond and others do not, and why some patients respond to EGFR-targeting agents and others do not, would produce vast clinical benefits. Given the particular importance of the EGFR signaling pathway in cancer, we have here systematically analyzed potential factors affecting tumor susceptibility or resistance to EGFR inhibitors.
The reason EGFR has been viewed a critical target involves its role as a central regulator of cell proliferation, survival and migration in normal and cancerous cells, and EGFR has long been a target of intense interest and effort for development of targeted cancer therapeutic agents. The structure of EGFR is shown in . EGFR is a transmembrane receptor tyrosine kinase (RTK). The extracellular, cysteine-rich domain binds activating ligands. EGFR ligands are expressed on the cell surface as tethered precursors that require proteolytic cleavage to be released to the interstitial space. A family of such proteases (sheddases, also known as a disintegrin and metalloprotease, or ADAM, proteases) release the ligands EGF, betacellulin, epiregulin, TGF-α, amphiregulin, and heparin-binding EGF-like growth factor (HB-EGF) [8
]. Once released from the membrane, the 49–85-amino acid mature growth factors are able to bind the EGF receptors. In 1997, Lemmon et al
. first suggested a model in which one EGF monomer binds to one EGFR monomer, and receptor dimerization involves subsequent obligate association of two monomeric (1:1) EGF EGFR complexes, associated with activation [9
]. The intracellular domain of EGFR includes a ligand- and dimerization-activated tyrosine kinase, which upon activation autophosphorylates EGFR at residues including Y992, Y1045, Y1068, Y1148 and Y1173 [10
], and also phosphorylates other binding partners, as discussed below. Cooperative physical interaction between adjacent kinase domains of adjacent erythroblastic leukemia viral (v-erb-b) oncogene homolog (ErbB) family molecules is required for EGFR activation [11
]. The normally auto-inhibited EGFR kinase undergoes conformational changes upon allosteric contact with another EGFR kinase domain, a mechanism strikingly similar to the activation of cyclin-dependent kinases by their cognate cyclins. The increased number of EGFR molecules on the cell surface associated with overexpression in cancer is predicted to increase the rate of kinase autoactivation [12
EGFR/(v-erb-b) oncogene homolog 1 (ErbB1) and the ErbB protein family
The intracellular domain of EGFR also encompasses a motif regulating internalization [13
]. Oligomerization of the ligand-activated EGFR triggers a rapid process of receptor internalization in parallel with the autophosphorylation/activation process, which allows regulated removal of the protein from the cell surface into intracellular compartments, limiting the duration of signaling [14
EGFR (also known as ErbB1) is the initially characterized member of a four-member group of proteins that also includes ErbB2/human EGFR 2 (HER2), ErbB3, and ErbB4 (). As part of its activation process, EGFR homodimerizes, but also in some cases heterodimerizes with ErbB2 and other family members to create an active signaling unit [16
]. The four ErbB proteins have overall structural similarity, with some key differences. For example, ErbB2 is not ligand-activated, but activated based on its heterodimerization with other family members [17
]; in normal cells, EGFR/ErbB2 heterodimers are more active than EGFR homodimers due to constitutive kinase activity of ErbB2, but this is regulated by the relative low expression level of ErbB2 in normal cells [18
]. As another example, ErbB3 is ligand-activated, but by a different set of ligands than EGFR, and lacks an active kinase domain, so that its main transmission of signals downstream is through its heterodimeric partners [19
There are three very well described and physiologically important signaling outputs of activated EGFR homo- or heterodimers:
- EGFR activates RAS viral oncogene homolog (Ras) signaling. The adaptor protein growth factor receptor-bound protein 2 (GRB2) either directly or with assistance of another scaffold protein, v-src sarcoma viral oncogene homology 2 domain-containing protein (Shc), binds activated EGFR at auto-phosphorylated phosphotyrosines, and activates son of sebvenless (SOS), a GTP-exchange factor for Ras, initiating signaling cascades that proceed through v-raf murine sarcoma viral oncogene homolog (Raf)/MAP kinse-ERK kinase (MEK)/extracellular signal-regulated kinase (ERK), influencing proliferation and cell cycle; v-ral simian leukemia viral oncogene homolog guanine nucleotide dissociation stimulator (RalGDS), influencing proliferation and migration; and phospho-inositol 3 kinase (PI3K), and thence to the important PI3K effector v-akt murine thymoma viral oncogene homolog (AKT), governing survival responses.
- EGFR activates signal transducer and activator of transcription (Stat)3 and 5 leading to their phosphorylation, nuclear translocation, and transcriptional activation of genes involved in proliferative response.
- EGFR directly binds and activates the p85 subunit of PI3K, thus providing a second stimulus to activate AKT. Although these are some of the most studied interactions, EGFR also interacts with a number of other signaling proteins; a recent peptide library screen identified multiple overlapping sites for over 40 interactions between EGFR and cytosolic proteins [10,21]. Soon after its initial discovery in 1980 , abnormal function of EGFR was identified as cancer-promoting, based on the discovery that the v-erbB oncogene, encoded by the erythroblastosis virus, represented a truncated form of EGFR [22–24]. The v-erbB truncation constitutively activated the tyrosine kinase activity of the protein, driving proliferation and apoptosis resistance in tumors. After three decades of study, several distinct categories of oncogenic lesions influencing endogenous EGFR activity in human tumors have been documented as clinically significant (). One of the best studied is activation of the EGFR signaling arising either from increased gene copy number (induced by amplification or polysomy [25,26]) or from feedback upregulation in response to cellular stresses [27–29]. Second, in some cancers, the EGFR ligands are overexpressed . Third, increased activity of the EGFR signaling is mediated through activating mutations in its kinase domain, or via structural alterations of the extracellular domain such as in variant III (EGFRvIII) discussed below. These direct mechanisms of EGFR activation are observed in many different types of cancer, at significant frequencies. For example, increase in EGFR gene copy number is seen in approximately one fifth of non-small cell lung cancers (NSCLC)  and is found in populations partially overlapping with those possessing EGFR kinase mutations (most commonly seen in female non-smoking lung adenocarcinoma or bronchoalveolar carcinoma patients) . Finally, in addition to cancers with lesions directly related to EGFR hyperactivity, EGFR and its family members are also central components of critical autocrine feedback circuits that promote growth (). Hence, even in cancers in which EGFR itself is not oncogenically activated, inhibition of EGFR would be predicted to have therapeutic value.
For these reasons, enormous effort has gone into the development of targeted therapeutic agents that inhibit EGFR. The most useful clinical agents fall into two classes: small-molecule tyrosine kinase inhibitors (TKIs), and EGFR-directed antibodies (). The small-molecule inhibitors of EGFR include reversible inhibitors such as erlotinib and gefitinib, and irreversible ones, such as HKI-272, EKB-569 and CI-1033. These TKIs target the intracellular kinase domain of EGFR by blocking the ATP-binding pocket thus preventing autophosphorylation on cytoplasmic tyrosines, and subsequent assembly of downstream macromolecular signaling complexes mediated through v-src sarcoma viral oncogene homology 2 domain (SH2)- or polypyrimidine tract binding protein (PTB)-domain interactions with EGFR phospho-tyrosines.
Current use of anti-EGFR agents in clinical settings
The first of the EGFR-targeting antibodies, m225, was generated in John Mendelsohn’s laboratory in the early 1990’s and was subsequently transformed into an acclaimed anti-cancer targeting agents, cetuximab [33
]. There are presently at least five EGFR-targeting antibodies in clinical use or under development, including cetuximab, panitumumab, matuzumab (EMD 72000), nimotuzumab (hR3), and zalutumumab [34
]. Upon receptor activation, EGFR domains I and III are brought within proximity of each other, and create a binding site for the ligand molecule [35
] (). Structural analyses have shown that most of the EGFR-targeting antibodies, including cetuximab, sterically hinder the interactions between ligand and binding site on EGFR [25
]. Alternatively, several recently developed antibodies (e.g., 806 [37
] and zalutumumab [34
]) act by blocking the conformational rearrangements that are required during EGFR dimerization, while some target EGFR dimerization with partners (e.g., pertuzumab, which disrupts EGFR ErbB2 interactions [38
]). Besides directly inhibiting EGFR-dependent signaling, a second action of these antibodies is to promote receptor removal from the cell surface, reducing the active pool of the protein available to signal [14
]. A third, in vivo
, activity of therapeutic antibodies is to induce antibody-dependent, cell-mediated cytotoxicity (ADCC) [39
]. In this process, the Fc domain of the antibody engages the Fc receptors on macrophages and natural killer (NK) cells, driving their activation and promoting cancer cell destruction.
In rare cancers, in which there is absolute dependence of the malignant phenotype on a unique oncogenic lesion, the use of a targeted agent as a monotherapy is appropriate and yields a dramatic clinical response. This is the underlying basis for efficacy of imatinib and other drugs of its class, which block the highly potent oncogenic driver breakpoint cluster region-Abelson murine leukemia viral (v-abl) oncogene homolog (BCR-ABL), in chronic myeloid leukaemia CML [1
]. For the many reasons discussed below, EGFR-targeting inhibitors are very effective as single agents only in a small subset of patients. However, because of the central role of EGFR signaling in apoptosis resistance, neo-angiogenesis, cell proliferation and damage repair, inactivation of EGFR was immediately predicted as being likely to enhance the action of cytotoxic chemotherapies, radiation or other targeted agents. EGFR inhibitors are commonly used in combination therapies with these other treatments in the clinic, and hence, in studying resistance to EGFR-targeting therapies, it is also necessary to consider interactions between EGFR and its co-administered agents.
In this review, we summarize current and emerging concepts of tumor resistance to pharmacologic signaling inhibitors, particularly bearing on resistance to EGFR. These topics include (1) pharmacokinetic resistance mechanisms, in which drugs fail to reach their targets; (2) resistance due to changes in the drug target; (3) compensatory mechanisms and lateral rescue pathways in cancer cells, that render targeting ineffective; (4) the potential for intelligent development of combinatorial strategies, that utilize understanding of biological networks to devise synergistic therapies that override resistance. Although EGFR is the main focus of this review, this discussion emphasizes commonalities that can be applied to many other pathway-targeted inhibitors.