New Targetable Oncogenes in Non–Small-Cell Lung Cancer

Abstract

The identification of oncogenic driver mutations underlying sensitivity to epidermal growth factor receptor and anaplastic lymphoma kinase tyrosine kinase inhibitors has led to a surge of interest in identifying additional targetable oncogenes in non–small-cell lung cancer. A number of new potentially oncogenic gene alterations have been characterized in recent years, including BRAF mutations, HER2 insertions, PIK3CA mutations, FGFR1 amplifications, DDR2 mutations, ROS1 rearrangements, and RET rearrangements. In this review, we will discuss the techniques used to discover each of these candidate oncogenes, the prevalence of each in non–small-cell lung cancer, the preclinical data supporting their role in lung cancer, and data on small molecular inhibitors in development.

INTRODUCTION

The treatment of non–small-cell lung cancer (NSCLC) has been transformed by the identification of rationally targeted therapies for a subset of molecularly defined lung cancers. Initial experience with targeted therapies for unselected patients with NSCLC resulted in modest and inconsistent improvements in outcomes. Although the addition of bevacizumab to first-line doublet chemotherapy improved median survival by 2 months in one randomized study, it failed to improve survival in another.^1–3 Treatment with epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs) in unselected patients with NSCLC resulted in similarly modest effects: Erlotinib in previously treated NSCLC improved median survival by 2 months,⁴ whereas gefitinib did not significantly improve survival in a similar population.⁵

In contrast, treatment of molecularly selected populations with rationally targeted therapies has led to unprecedented results. Prospective studies of erlotinib and gefitinib in patients with EGFR-mutant lung cancer have resulted in response rates greater than 60%, and randomized trials in both Asia and Europe have shown a marked improvement in progression-free survival (PFS) over first-line chemotherapy, albeit with no overall survival benefit because of patient crossover.^6,7 Initial data with crizotinib in ALK-rearranged lung cancer have demonstrated similar response rates⁸; studies randomly assigning patients to crizotinib versus standard chemotherapy are under way.^8a The remarkable effectiveness of these drugs in molecularly selected populations has led to a surge in the number of lung cancer trials studying targeted therapies in other molecularly selected populations.

In this review, we will discuss emerging data on new targetable oncogenes in NSCLC. We will defer to other reviews in this series for a discussion of cancers with KRAS mutations and MET alterations. We will focus on genes encoding protein kinases, given the important recent progress in treating lung cancer through the use of targeted kinase inhibitors. The emerging therapies directed at immune modulation, apoptosis, and chaperone proteins are outside of the spectrum of this review. Although several of these new oncogenes are found rarely in NSCLC, in total they constitute 9% to 14% of lung adenocarcinomas (Table 1) and 16% to 30% of squamous cell lung carcinomas (Table 2; Fig 1). To highlight the varied strategies that have been used in recent years to identify new targetable oncogenes, this review will discuss oncogenes grouped by the technique in which they were identified (Fig 2).

Table 1.

New Targetable Oncogenes in Lung Adenocarcinoma

Oncogenic Target	Prevalence (%)	Reported Clinical Associations	Potential Kinase Inhibitors
HER2 insertions	2.89	Never-smokers9	Afatinib (BIBW-2992)
		Asian race9	Neratinib (HKI-272)
		Female sex9	Dacomitinib (PF-00299804)
BRAF mutations	2 to 4.910–12	Ever-smokers10	Vemurafenib
		White race10	GSK2118436
		V600E: never- smokers11
		V600E: female sex11
PIK3CA mutations	1.5 to 2.612–15	No association seen13–15	GDC-0941
			XL147
			BKM120
RET rearrangements	1.216	Never-smokers16	Vandetanib
	1.917*	Asian race17	Sorafenib
			Sunitinib
			Cabozantinib (XL184)
ROS1 rearrangements	1.2 to 2.616,18	Never-smokers18	Crizotinib
		Asian race18
		Younger age18

^*Never-smokers or former limited smokers.

Table 2.

New Targetable Oncogenes in Squamous Cell Lung Carcinoma

Oncogenic Target	Prevalence	Reported Clinical Associations	Potential Kinase Inhibitors
PIK3CA mutations	3.6 to 6.513,14,19	No association seen13,14	GDC-0941
			XL147
			BKM120
FGFR1 amplification	9.7 to 2120,21	Ever-smokers20	Brivanib (BMS-582664)
			Dovitinib (TKI258)
			Ponatinib (AP24534)
			E3810
DDR2 mutations	2.222	No association seen22	Dasatinib
			Imatinib
			Nilotinib

Fig 1.

Prevalence of new targetable oncogenes in non–small-cell lung cancer. Although several of the individual oncogenic alterations are relatively rare, in total they constitute (A) 9% to 14% of lung adenocarcinomas and (B) 16% to 30% of squamous cell lung carcinomas.

Fig 2.

Strategies used to identify targetable oncogenes in non–small-cell lung cancer (NSCLC). (A) Alterations in recognized proto-oncogenes (eg, BRAF mutations) can be identified through focused sequencing efforts. (B) Unexpected oncogenes (eg, FGFR1 or ALK) can be discovered through genome-wide screening strategies, with the requirement that initial positive results are subsequently confirmed in a separate cohort of clinical specimens. (C) Analysis of specimens from a trial of a targeted therapy (eg, epidermal growth factor receptor [EGFR] tyrosine kinase inhibitors [TKIs]) has the advantage of being inherently tied to an available treatment, although there may be limitations to what genomic analyses are feasible on biopsies from patients with advanced NSCLC. All candidate oncogenes should demonstrate oncogenicity in a model system before they can be considered driver mutations.

ONCOGENIC ALTERATIONS IDENTIFIED THROUGH FOCUSED ANALYSIS OF KNOWN PROTO-ONCOGENE

The most straightforward approach toward the identification of driver mutations in NSCLC has been to search for alterations in known proto-oncogenes, coding proteins with an established important role in cellular growth signaling. Through this approach, several new molecular targets in NSCLC have been identified, including HER2 insertions, BRAF mutations, and PIK3CA mutations.

HER2 Insertions

Human epidermal growth factor receptor 2 (HER2, or ERBB2) is a membrane-bound tyrosine kinase in the ERBB family; unlike EGFR (or EBRR1), HER2 has no known ligand but will heterodimerize with other members of the ERBB family on ligand binding. The HER2 gene has been a recognized proto-oncogene in human cancers for more than two decades, since it was found to be amplified in approximately 30% of breast cancers. HER2-positive breast cancer has historically been associated with a poorer prognosis, although outcomes are improved significantly through the use of HER2-targeted agents like trastuzumab.^23,24 HER2 has also been found to be amplified or overexpressed in a subset of gastric carcinomas, in which it is associated with improved outcomes through the addition of trastuzumab to chemotherapy.²⁵ Although HER2 is also overexpressed in 13% to 20% of NSCLCs (with 3+ expression in only 2% to 6%),^26–29 a study of trastuzumab in HER2-overexpressing NSCLC did not identify significant clinical activity.³⁰ HER2 amplification on fluorescent in situ hybridization (FISH) is seen in approximately 2% to 4% of NSCLCs (more commonly in adenocarcinomas),^26,27 but trastuzumab sensitivity in this population has not been clearly studied.

The initial identification of EGFR mutations in a subset of lung adenocarcinomas led to subsequent interest in identifying such mutations in HER2, particularly given the homology between these two ERBB-family kinases. In the initial report, 120 patients with NSCLC underwent HER2 sequencing; mutations were identified in five (4.2%) and were found to be nonoverlapping with KRAS and EGFR mutations.³¹ A second study of 671 patients with resected NSCLC found HER2 mutations in 11 (1.2%); all occurred in those with adenocarcinoma, where the prevalence was 2.8%.⁹ All of the mutations identified were in-frame insertions of three to 12 base pairs into exon 20 and were found to be more prevalent in never-smokers. HER2 insertions have been found to induce constitutive activation of the HER2 kinase in a ligand-independent fashion,³² similar to the effect of EGFR mutations. Indeed, parallel exon 20 insertion mutations in the EGFR gene can be identified with a similar frequency in NSCLC, in approximately 1.9% of patient cases.³³ Inducible tissue-specific overexpression of mutant HER2 in a transgenic mouse model was found to result in rapid development of NSCLC, confirming the oncogenicity of this driver mutation.³⁴

Several kinase inhibitors are in development for HER2-dependant lung adenocarcinoma. The most promising compounds are irreversible TKIs targeting HER2 and EGFR, such as neratinib (HKI-272), dacomitinib (PF-00299804), and afatinib (BIBW-2992).³⁵ Neratinib and dacomitinib have both been found to effectively inhibit the growth of HER2-mutant lung cancer cell lines as well as cell lines transformed by the introduction of HER2.^32,36 Afatinib was found to induce modest regressions when introduced into the previously mentioned transgenic mouse model, and this effect was potentiated by the addition of rapamycin, a mammalian target of rapamycin (mTOR) inhibitor, suggesting a particular dependence on the Akt/mTOR pathway in HER2-mutant lung cancer.³⁴ Although there are limited data regarding mTOR inhibitors in patients with HER2-mutant lung adenocarcinoma, a phase I trial of neratinib plus temsirolimus resulted in partial responses in two of six patients with HER2-mutant NSCLC.³⁷ An early report from a phase II trial of afatinib alone in molecularly selected lung adenocarcinoma (NCT00730925) describes partial responses in two of five patients with tumors harboring HER2 mutations³⁸; efficacy may have been limited by toxicity, which required several patients to come off study before progression. Interestingly, HER2-amplified NSCLC may also gain benefit from HER2 kinase inhibitors, based on a report of a marked response seen in one patient receiving dacomitinib.³⁹ However, a prospective trial of dacomitinib has now reported only three responses out of 18 HER2-mutant lung cancers,^39a suggesting that dacomitinib alone may be inadequate to fully inhibit this particular oncogene in patients.

BRAF Mutations

BRAF is a serine/threonine kinase downstream from KRAS in the mitogen-activated protein kinase (MAPK) signaling cascade. Although KRAS mutations have been well established in human cancer for several decades, BRAF mutations were only first identified in 2002, with a particularly high prevalence identified in melanoma.⁴⁰ BRAF mutations were found most commonly in exon 15, resulting in a substitution of a glutamine for a valine at residue 600 (V600E); this substitution is believed to destabilize the inactive conformation of the kinase, resulting in constitutive kinase activation and downstream phosphorylation.⁴¹ BRAF mutations have since been identified in a variety of human cancers, including colorectal cancer,⁴² papillary thyroid cancer,⁴³ and hairy cell leukemia.⁴⁴

The high incidence of oncogenic BRAF mutations in melanoma led investigators to search for BRAF mutations in NSCLC. In two initial reports, each studying more than 100 patient cases of NSCLC, sequencing of exons 11 and 15 of BRAF identified missense mutations in 1.6% and 3%, respectively; a majority of the mutations were non-V600E.^45,46 More recently, 697 and 739 lung adenocarcinomas were genotyped by two groups of US and Italian investigators, identifying missense mutations in 3% and 4.9% of patient cases, respectively^10,11; the lower prevalence in one study may have been the result of a focal genotyping technique looking only at codons V600, D594, and G469.¹⁰ Approximately half of the mutations identified were V600E, with one group finding that this particular mutation was more common in never-smokers.¹¹ Importantly, non-V600E mutations seem to be relatively more common in NSCLC than in colorectal cancer or melanoma and were found to occur exclusively in patients with lung cancer with a smoking history.

The potential role of BRAF V600E as targetable driver mutation in lung adenocarcinoma is strengthened by the finding that in vivo expression of V600E in mouse models leads to development of invasive adenocarcinoma, a phenotype that is reversed when V600E expression is stopped.⁴⁷ This finding, which parallels the behavior of EGFR-activating mutations in mouse models, fuels optimism that BRAF-mutant lung adenocarcinoma may be highly sensitive to TKIs like vemurafenib, which has demonstrated potent anticancer activity in melanomas harboring BRAF V600E mutations.⁴⁸ However, this enthusiasm must be tempered by the experience in colon cancer, in which BRAF inhibitors have been found to have little activity against tumors carrying V600E mutations.⁴⁹ The contrasting phenotypes seen in these two types of BRAF-mutant cancer suggest that genotype alone may be inadequate for the prediction of sensitivity to targeted therapy; a better understanding of tissue-specific factors modulating oncogenic sensitivity to TKIs is needed. One instance of a positron emission tomography response to vemurafenib in a patient with NSCLC harboring BRAF V600E has been reported,⁵⁰ and a phase II trial of the BRAF inhibitor GSK2118436 is currently under way in BRAF V600E–mutant NSCLC (NCT01336634) to prospectively study this approach in lung cancer. Because cancers with non-V600E BRAF mutations are unlikely to be inhibited by V600E-specific inhibitors, inhibitors of downstream targets (ie, MEK) are being explored in this population (NCT00888134) because these have be found to be active in BRAF-mutant melanoma.⁵¹

PIK3CA Mutations

Phosphatidylinositol 3-kinases (PI3Ks) are a family of lipid kinases that play an important role in regulating cell growth, proliferation, and survival. Mutations in PI3Ks were first identified through a sequencing analysis of eight PI3K genes in 297 cancer specimens (primarily colon cancers).⁵² Point mutations in the PIK3CA gene, encoding the catalytic subunit of the class 1 PI3Kα, were found in several cancer types, including one of 24 lung cancer specimens studied. In three subsequent large analyses of 229, 235, and 691 NSCLC specimens, PIK3CA mutations were found in three (1.3%), eight (3.4%), and 11 (1.6%) patient cases, respectively,^13,14,53 with prevalence being higher in squamous cell carcinoma. A recent study genotyped 95 resected squamous cell lung cancers (all p63+/TTF1−), and PIK3CA mutations were found in four (4%), whereas no mutations in EGFR, KRAS, ALK, HER2, or BRAF were identified.¹⁹

An important finding has been that PIK3CA mutations in lung adenocarcinoma often coexist with another oncogenic mutation, such as an EGFR or KRAS mutation,^13,15 raising suspicion that this may be a secondary process and not a true driver mutation. Yet genetically engineered mouse models have demonstrated that expression of mutant PIK3CA on its own can induce multifocal adenocarcinoma, and these tumors rapidly regress when expression ceases.⁵⁴ In lung cancer cell lines with PIK3CA mutations, it has also been shown that PI3K activity is elevated, and RNAi knockdown of PI3K causes growth arrest.¹⁴ One could hypothesize that a portion of PIK3CA-mutant lung cancers are dependent on this driver oncogene, whereas in other patient cases, the mutation may modulate the effect of another oncogenic process.

Multiple PI3K inhibitors are in development, although it remains uncertain what population should be targeted with these agents. In the previously described genetically engineered mouse model of PIK3CA-mutant lung cancer, a PI3K inhibitor with broad activity was found to inhibit PI3K phosphorylation and induce tumor response.⁵⁴ However, there is debate in the breast cancer literature as to whether PI3K inhibitor activity is truly limited to those cancers with PIK3CA mutations,⁵⁵ particularly given the broader activity seen with mTOR inhibitors. Ongoing trials in lung cancer include single-agent PI3K inhibitors (NCT01501604) as well as combinations with chemotherapy (NCT00974584, NCT00756847) and other targeted agents (NCT00974584).

ONCOGENIC ALTERATIONS IDENTIFIED THROUGH GENOMIC SCREENING STRATEGIES

Genomics is a discipline with a broad interest in describing the landscape of genetic alterations in a given organism or disease. Although this discipline was not specifically developed for the purpose of identifying targetable oncogenes, genomic tools can be used to screen broadly for genetic alterations of potential interest in cancer treatment. Cancer genomics primarily studies structural DNA alterations—missense mutations, copy number alterations, and translocations.⁵⁶ There are a wide variety of genomic analysis technologies available, each with specific strengths and weaknesses that have been recently reviewed in detail.⁵⁷ Potentially targetable alterations identified using these techniques include FGFR1 amplifications, DDR2 mutations, and RET rearrangements.

FGFR1 Amplification

The important advances achieved over the past decade through identifying oncogenic mutations in lung adenocarcinoma have led to several efforts to screen for targetable oncogenes in squamous cell carcinoma of the lung. An initial screen of 155 squamous cell lung cancer specimens using single nucleotide polymorphism arrays identified focal amplifications of the FGFR1 gene.²⁰ FGFR1 is a membrane-bound receptor tyrosine kinase that regulates proliferation via the MAPK and PI3K pathways, much like EGFR. High-level amplification of FGFR1 (chromosome 8p) was seen in 9.7% of patient cases in addition to the previously described amplification of SOX2 (chromosome 3q).⁵⁸ Seventy-seven lung adenocarcinoma specimens were also studied, and FGFR1 amplification was seen in only 1%. A second analysis has confirmed the finding that FGFR1 amplification is much more common in squamous cell lung cancers (21%) than in lung adenocarcinoma (3.4%).²¹ Survival of FGFR1-amplified lung cancer cell lines was additionally shown to be dependent on overexpression of the FGFR1 kinase.^21,58

Several small molecular inhibitors of FGFR1 are in development, and initial studies indicate activity against FGFR1-amplified lung cancer models. The FGFR inhibitor PD173074 (with activity against multiple FGFR-family kinases as well as vascular endothelial growth factor receptor 2 [VEGFR2]⁵⁹) has been found to inhibit the growth of FGFR-amplified lung cancer cell lines^21,58 and also led to responses in mouse xenografts.⁵⁸ Sensitivity to PD173074 was not seen in cell lines without FGFR1 amplification. Clinically, the FGFR inhibitor farthest along in development is brivanib (BMS-582664), a dual inhibitor of FGFR and VEGFR signaling with an adverse effect profile similar to those of other VEGF inhibitors.⁶⁰ In a randomized discontinuation study of brivanib in 396 patients with five types of advanced solid tumors, no partial responses were seen in 42 patients with previously treated NSCLC.⁶¹ A second generation of FGFR inhibitors with greater specificity (and fewer toxicities) is in development.⁶²

DDR2 Mutations

DDR2 is a membrane-bound receptor tyrosine kinase that binds to collagen and can regulate proliferation and migration. Mutations in the DDR2 gene were recently identified in squamous cell lung cancer through a sequencing screen of 201 genes with a potential role in human cancer (including the entire tyrosine kinome).²² Investigators used an iterative approach for this analysis: In an initial screen, 20 tumors were sequenced, and mutations were identified in six tyrosine kinase genes, including DDR2. In a secondary screen of 48 tumors and cell lines, sequencing of just the six candidate tyrosine kinase genes was performed, and four DDR2 mutations were found. Lastly, a validation cohort of 222 specimens underwent sequencing of DDR2 alone, and mutations were found in five patient cases (2.2%). Supporting the oncogenic role of DDR2, the investigators showed that DDR2-mutant cell lines had impaired proliferation through shRNA knockdown of DDR2 expression. It should be noted that a prior genotyping effort found no DDR2 mutations in 54 squamous cell lung cancers,⁶³ but these investigators only tested for a single mutation (R105S), which was not one of the mutations identified in the larger screening study.

Although DDR2 has not been a major focus of drug development efforts, a recent analysis showed that ABL kinase inhibitors such as imatinib, nilotinib, and dasatinib have activity against DDR2.⁶⁴ In the previously mentioned study, it was found that dasatinib inhibited the growth of DDR2-mutant lung cancer cell lines and caused regressions of DDR2-mutant mouse xenographs. Unfortunately, dasatinib has not been found to be especially active against NSCLC, with only one partial response seen in a phase II study of 34 patients,⁶⁵ although only six of the patients in this study had squamous cell carcinoma. Intriguingly, a patient with squamous cell lung cancer responding to dasatinib and erlotinib in a separate phase II study⁶⁶ had tissue available for DDR2 sequencing, and a point mutation in the DDR2 kinase domain was identified.²²

There is optimism that DDR2 mutations will prove to be the first targetable mutations in squamous cell lung carcinoma; however, there remain a few unresolved questions. For example, although the initial screening study identified 11 total mutations in DDR2,²² these mutations occurred at 10 different codons (with only five mutations in the kinase domain) rather than there being a few recurring mutations. Whether all of these variants are oncogenic and sensitive to dasatinib will need to be determined. Additionally, an important toxicity of dasatinib therapy is the development of pleural effusions, a challenging adverse effect for many patients with lung cancer, who already have limited pulmonary reserve. To clarify the feasibility of this approach, two phase II trials of dasatinib in squamous cell lung carcinoma are under way (NCT01491633, NCT01514864).

RET Rearrangements

RET is a receptor tyrosine kinase with a well-established role in human cancer: Missense mutations in RET are common in sporadic medullary thyroid cancer, and RET rearrangements can be identified in a subset of papillary thyroid cancers.⁶⁷ RET rearrangements were recently identified in a subset of lung adenocarcinomas by several groups using different screening strategies. Two groups performed whole-transcriptome sequencing of RNA extracted from lung adenocarcinoma specimens, and each identified a patient case with an unexpected fusion between KIF5B and RET, two genes separated by more than 10 Mb on chromosome 10.^68,69 This same fusion gene was identified by a third group through use of targeted next-generation sequencing of 2,574 exons and 37 introns from 152 genes with importance in cancer pathways.¹⁷ Introns from RET were included in the sequencing effort because of the role of RET rearrangements in thyroid cancer, leading to the identification of KIF5B-RET in one of 24 NSCLC specimens studied. Lastly, a fourth group identified the KIF5B-RET fusion through a FISH screen for new fusions involving KIF5B,¹⁶ given their previous identification of a rare ALK rearrangement involving KIF5B.⁷⁰ Using a subsequent FISH screen for new RET fusions, this fourth effort additionally identified an alternate fusion partner for RET, the CCDC6 gene. It was additionally confirmed that KIF5B-RET was oncogenic when expressed in cell lines or implanted into xenographs.^16,17,69 In the largest prevalence analysis, combined RET FISH and real-time polymerase chain reaction (RT-PCR) identified 13 tumors (1.2%) with RET rearrangements among 1,119 lung adenocarcinomas.¹⁶ In an alternate prevalence analysis of 159 lung adenocarcinomas from never-smokers or former light smokers, who were wild type for know oncogenic alterations (in EGFR, KRAS, ALK, HER2, BRAF, and ROS1), 10 RET rearrangements (6.3%) were identified.¹⁷

Several commercially available TKIs have activity against the RET kinase. Vandetanib, initially developed as an EGFR inhibitor, has now been US Food and Drug Administration (FDA) approved for the treatment of advanced medullary thyroid carcinoma based on an improvement in PFS and response rate over placebo.⁷¹ Vandetanib has been tested in several cell lines transformed by KIF5B-RET and was found to inhibit RET phosphorylation and suppress cell growth.^16,17,69 Sunitinib and sorafenib, two VEGFR inhibitors with activity against the RET kinase, have also been found to inhibit the growth of KIF5B-RET–transformed cell lines.¹⁷ Cabozantinib is a new kinase inhibitor with activity against medullary thyroid cancer that also has activity against KIF5B-RET–transformed cell lines.⁷² Each of these multikinase inhibitors has been found to have some degree of activity in NSCLC previously, but it remains to be determined whether sensitivity to these drugs in patients with lung cancer is the result of underlying RET rearrangements.

ONCOGENIC ALTERATIONS IDENTIFIED THROUGH FUNCTIONAL GENOMIC SCREENING STRATEGIES

Functional screening looks beyond the structure of the cancer genome to identify genes or proteins specifically associated with oncogenicity. ALK rearrangements were first identified in this fashion.⁷³ Using a retrovirus, coding DNA sequences were generated from RNA extracted from a lung adenocarcinoma specimen. These DNA sequences were transfected into cells that would only grow if the DNA sequence were sufficiently oncogenic and coded for a protein that could sustain cellular growth. Through study of one of these functionally selected DNA sequences, a fusion protein of EML4 and ALK was identified and subsequently studied. Because of their greater specificity for the identification of oncogenic alterations, there are a number of such functional screening assays now being used, one of which led to the identification of ROS1 rearrangements in NSCLC.

ROS1 Rearrangements

ROS1 is a receptor tyrosine kinase with homology to the insulin receptor and signals down the MAPK signaling cascade through phosphorylation of RAS. The ROS1 gene was first identified as an oncogene in NSCLC through a phosphoproteomic screen, leveraging the fact that oncogenic kinases should be highly phosphorylated proteins.⁷⁴ Studying 41 NSCLC cell lines and 150 tumors, investigators examined which kinases had elevated activity when compared with the average activity of that kinase across all specimens. Several expected oncogenes were identified among the highly phosphorylated kinases, including EGFR, HER2, and MET. They additionally identified that ALK, ROS1, and PDGFRα were highly phosphorylated, both in some cell lines and some tumor specimens. Studying RNA transcripts from these specimens using RT-PCR, several unexpected fusion proteins were identified, including two ALK fusions and two ROS1 fusions (SLC34A2-ROS1 and CD74-ROS1). Whereas ROS1 rearrangements had previously been identified in glioblastoma,⁷⁵ this was the first description of ROS1 rearrangements in NSCLC.

The prevalence of ROS1 rearrangements was recently studied in a multi-institutional series of 1,073 patients with NSCLC using a break-apart FISH assay.¹⁸ Of the 694 adenocarcinomas studied, 18 tumors (2.6%) were positive for ROS1 rearrangements by FISH, although translocations could be confirmed by RT-PCR in only six patients. Rearrangements were found to be more common in never-smokers and Asian patients and were associated with younger age. In a subsequent FISH analysis of 1,116 patients with lung adenocarcinoma, 13 (1.2%) had rearrangements, and 11 of these were confirmed by RT-PCR.¹⁶ This second study identified three new fusion partners for ROS1: TPM3, SDC4, and EZR.

ROS1 has been a recognized oncogene in glioblastoma for many years,⁷⁵ but selective ROS1 inhibitors have not yet been developed clinically. However, it has been identified that cells carrying ROS1 rearrangements have sensitivity to ALK inhibitors. Using an automated platform, investigators screened 602 cell lines for sensitivity to TAE684, a selective ALK inhibitor, and found that 10 cell lines had greater than 50% cell kill after 72 hours of treatment.⁷⁶ Although most of these cell lines harbored ALK alterations, one cell line (HCC-78) had previously been found to harbor a ROS1 translocation.⁷⁴ After confirming that this cell line was similarly sensitive to crizotinib, investigators treated a patient with ROS1-rearranged lung adenocarcinoma with crizotinib, and he had a near complete response.¹⁸ Fourteen patients with ROS1-rearranged lung cancers have now been treated in an expansion cohort of the ongoing phase I trial of crizotinib, and nine (64%) had a confirmed response.⁷⁷

ONCOGENIC ALTERATIONS IDENTIFIED THROUGH STUDY OF CANCERS SENSITIVE TO TARGETED THERAPY

One final strategy for oncogene discovery is the strategy initially used to identify EGFR mutations in NSCLC. With this strategy, a therapy with known targets is administered to an unselected patient population, and subsequent analysis of tissue from responders can potentially identify an oncogenic alteration underlying sensitivity. This is an extremely attractive strategy for oncogene discovery because it is inherently tied to a treatment option. However, this strategy can be quite challenging—years after FDA approval, there is still no biomarker that clearly predicts for bevacizumab sensitivity.⁷⁸ In the case of EGFR-sensitizing mutations, this oncogene was easier to identify for several reasons: first, EGFR was known to be commonly overexpressed and was a suspected oncogene; second, EGFR mutations are relatively common compared with other alterations; and third, EGFR mutations can be identified with a relatively simple technology (ie, Sanger sequencing).

Going forward, several important challenges will be encountered when using prospective clinical trials to identify oncogenic alterations underlying sensitivity. Most important is tissue adequacy, both in terms of quantity and quality. A core biopsy will be needed for most genomic analysis, but those requiring more than a core biopsy may be difficult to perform consistently in a cohort of patients with advanced NSCLC. Some analysis techniques may require fresh/frozen tissue, such that a fresh biopsy will need to be obtained on study entry. After the molecular characteristics of the tumors have been determined, the analysis will require that patients be separated into sensitive and resistant groups in some meaningful fashion. How to identify the sensitive phenotype using conventional computed tomography imaging is somewhat unclear; some have found that minor response may correlate more closely with presence of an underlying driver mutation than RECIST partial response.⁷⁹

DISCUSSION

In conclusion, steady progress is being made in identifying targetable oncogenes in lung adenocarcinoma and squamous cell carcinoma. Although many of these oncogenic alterations are uncommon, rationally targeted therapies can have a profound effect on outcomes in these subpopulations of NSCLC. At present, we recommend testing for these genomic alterations if feasible to determine clinical trial eligibility for appropriate patients.

A key challenge in the coming years will be developing efficient assays for identifying these genomic alterations such that appropriate targeted therapies can be delivered. Because sequential testing for mutations in each oncogene may be impractical, personalized cancer care is likely to eventually require multigene assays such as those reviewed elsewhere in this series. To facilitate the identification of new oncogenes underlying drug sensitivity, the greatest progress will be made by ensuring that prospective trials of targeted therapies are accompanied by painstaking collection of tumor tissue.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Although all authors completed the disclosure declaration, the following author(s) and/or an author's immediate family member(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.

Employment or Leadership Position: None Consultant or Advisory Role: Geoffrey R. Oxnard, Genentech (C), Boehringer Ingelheim (C); Pasi A. Jänne, Boehringer Ingelheim (C), Roche (C), Genentech (C), Abbott Laboratories (C), AstraZeneca (C), Pfizer (U), Quintiles (C), sanofi-aventis (C) Stock Ownership: None Honoraria: None Research Funding: None Expert Testimony: None Other Remuneration: Pasi A. Jänne, Labcorp

AUTHOR CONTRIBUTIONS

Manuscript writing: All authors

Final approval of manuscript: All authors