• Open Access

Non-solid oncogenes in solid tumors: EML4–ALK fusion genes in lung cancer

Authors


E-mail: hmano@jichi.ac.jp

Abstract

It is generally accepted that recurrent chromosome translocations play a major role in the molecular pathogenesis of hematological malignancies but not of solid tumors. However, chromosome translocations involving the e26 transformation-specific sequence transcription factor loci have been demonstrated recently in many prostate cancer cases. Furthermore, through a functional screening with retroviral cDNA expression libraries, we have discovered the fusion-type protein tyrosine kinase echinoderm microtubule-associated protein like-4 (EML4)–anaplastic lymphoma kinase (ALK) in non-small cell lung cancer (NSCLC) specimens. A recurrent chromosome translocation, inv(2)(p21p23), in NSCLC generates fused mRNA encoding the amino-terminal half of EML4 ligated to the intracellular region of the receptor-type protein tyrosine kinase ALK. EML4–ALK oligomerizes constitutively in cells through the coiled coil domain within the EML4 region, and becomes activated to exert a marked oncogenicity both in vitro and in vivo. Break and fusion points within the EML4 locus may diverge in NSCLC cells to generate various isoforms of EML4–ALK, which may constitute ~5% of NSCLC cases, at least in the Asian ethnic group. In the present review I summarize how detection of EML4–ALK cDNA may become a sensitive diagnostic means for NSCLC cases that are positive for the fusion gene, and discuss whether suppression of ALK enzymatic activity could be an effective treatment strategy against this intractable disorder. (Cancer Sci 2008; 99: 2349–2355)

Chromosome translocation is the most prevalent form of somatic changes in the cancer genome, occupying nearly three-quarters of all genetic change analyzed in cancer cells.(1) Such translocations may lead to the generation of novel fusion genes at the ligation points of chromosomes, or may juxtapose growth-promoting genes to aberrant promoter or enhancer fragments, resulting in dysregulated expression of the genes. In either case, such fusion genes or wild-type genes with altered expression may participate directly in the malignant transformation of cells that harbor chromosome translocations.

An archetypal example of such tumor-related translocations is t(9;22), which gives rise to the breakpoint cluster regio (BCR)-Abelson murine leukemia viral oncogene homolog 1 (ABL1) fusion gene in chronic myeloid leukemia (CML) and acute lymphoblastic leukemia (ALL).(2) Ligation to BCR constitutively elevates the protein tyrosine kinase (PTK) activity of ABL1, and forced expression of BCR–ABL1 in the hematopoietic system induces CML and ALL in mice,(3,4) proving that BCR–ABL1 plays a pivotal role in the pathogenesis of such leukemias. Also, molecular detection of BCR–ABL1 and the development of compounds to suppress BCR–ABL1 enzymatic activity have significantly changed the way we diagnose and treat individuals with CML. Because reverse transcription (RT)–polymerase chain reaction (PCR) can detect BCR–ABL1 fusion transcripts in almost 100% of individuals with CML, even among those without the characteristic t(9;22) translocation, molecular detection of BCR–ABL1 has become a standard technique used to diagnose CML. Given the very high sensitivity of PCR, such a strategy is also effective to follow up the tumor burden of leukemia in the patients.(5)

Further, the chemical compound STI571, which suppresses ABL1 kinase activity, has substantially prolonged the survival of patients at the chronic phase of CML, and achieved a higher probability of complete cytogenetic response among them compared to that with the previous treatment regimens.(6,7) Therefore, if translocation-mediated fusion genes encode activated enzymes with direct oncogenic potential, targeting such enzymes could provide a feasible approach to treat individuals harboring the corresponding fusion genes.

However, these fusion-type oncogenes have been reported frequently only in hematological malignancies, and not in solid tumors (especially in epithelial tumors).(8) It has been therefore widely assumed that balanced chromosome cytogenetic aberrations (and the resulting fusion genes) may be rare in the latter conditions. As shown in Table 1, for instance, the incidence of new cases with solid tumors in the USA is ~11 times larger than for those with hematological malignancies.(9) However, the number of recurrent balanced cytogenetic aberrations (RBA) in solid tumors (n = 125) is only a quarter of that in hematological malignancies (n = 495) worldwide,(10) suggesting that RBA are indeed characteristic of hematological malignancies and that these two tumor types may occur through distinct transformation mechanisms.

Table 1. Number of recurrent balanced cytogenetic aberrations (RBA) in cancer
Cancer typeNo. RBAAnnual no. new cases in USA
  • Calculated from data reported previously.(9,10)

Leukemia336   44 240
Lymphoma159   71 380
Solid tumors1251 329 300

However, such a notion has been challenged recently by Mitelman et al. who have demonstrated that the number of fusion genes may simply be a function of the number of cases with an abnormal karyotype in both hematological malignancies and solid tumors.(11,12) A correlation between the number of patients with an abnormal karyotype and that of fusion genes is constant throughout all types of cancers (R2 = 0.82, P < 0.001). It is therefore possible that infrequent reports of fusion genes in solid tumors (especially in epithelial tumors) may have been attributable to technical difficulties in obtaining clear karyotyping data or to the complex chromosome rearrangements in solid tumors.

If this is the case, many more fusion-type oncogenes may await discovery in solid tumors. Indeed, evidence in support of this prediction has been provided recently both by our discovery of the fusion-type PTK echinoderm microtubule-associated protein like-4 (EML4)–anaplastic lymphoma kinase (ALK), associated with lung cancer,(13–15) and by the detection of recurrent e26 transformation-specific sequence (ETS) fusion genes in prostate cancer.(16–18)

How to screen for oncogenes

Given the marked therapeutic efficacy of STI571 in CML, chemical inhibitors of epidermal growth factor receptor (EGFR) in lung cancer with activated EGFR,(19) and specific antibodies to HER2 in breast cancers with amplification of the HER2 locus,(20) it is necessary to identify pivotal oncogenes in every cancer and to target these ‘Achilles’ heels’(21) for developing effective treatment strategies. We therefore tried to establish a functional screening system for transforming genes among a wide variety of cancer specimens.

The focus formation assay with 3T3 or RAT1 fibroblasts(22) has been used extensively to screen for oncogenes from clinical specimens. In such screening, genomic DNA is extracted from samples and transfected into the recipient fibroblasts. Because protein products of oncogenes can interfere with contact inhibition in fibroblasts, cell clones that have received oncogenes may keep growing even after the cells become confluent in culture. Such piled-up foci of cell clones (transformed foci) can be readily identified by visual inspection and subjected to the recovery of incorporated oncogenes. Application of such technologies has indeed succeeded in the isolation of a variety of transforming genes, such as mutated RAS family genes, activated RAF family genes, and a number of activated PTK genes.(23)

However, we have noticed that this type of screening system has a strong tendency to isolate the same sets of genes among different types of cancer (i.e. activated RAS family proteins and guanine nucleotide exchange factors). This could be due to an intrinsic property of the assay system, which isolates genes overriding growth inhibition mediated by cell-to-cell contact in fibroblasts. Another reason may be related to promoter specificity (Fig. 1a). In the screening systems where genomic DNA is used for transfection, any oncogene is controlled transcriptionally in the recipient cells by its own promoter and enhancer fragments. Therefore, if a promoter fragment of a given oncogene is active only in a tissue-specific manner (hematopoietic cell-specific, for instance), that gene would not be transcribed in fibroblasts, and thus could not be captured in the assay. Therefore, a genomic DNA-mediated screening system can only identify oncogenes with promoter fragments that are active in the recipient cells.

Figure 1.

Development of retroviral cDNA expression libraries for oncogene screening. (a) Although oncogenes controlled by a constitutive promoter or enhancer are expressed when introduced into fibroblasts, those controlled by a tissue-specific promoter-enhancer (e.g. specific to the hematopoietic system) are not transcribed in fibroblasts and will therefore not be detected by such functional screening. (b) To overcome this limitation, we synthesized cDNA from small quantities of clinical specimens, inserted them into a retroviral plasmid, and generated recombinant retroviral libraries. Theoretically, any type of dividing cell can be infected with such libraries, with the incorporated cDNA being expressed at a high level in the recipient cells under the control of the viral long terminal repeat (LTR).

To overcome this limitation, it would be desirable to express every oncogene using an exogenous promoter fragment that allows abundant expression in any type of assay cell. For this purpose, we have developed a method to construct retrovirus-based cDNA expression libraries,(13,24–29) which can express any incorporated cDNA using a strong promoter fragment, long-terminal repeat (LTR), of the retroviral genome (Fig. 1b). Our system is so sensitive that we can generate libraries from small quantities of clinical specimens (such as <1 × 105 cells). Further, given the high infection efficiency of retrovirus to dividing cells, any type of functional assay can be conducted in any proliferating cell with retroviral libraries.(30,31)

Discovery of a fusion-type PTK, EML4–ALK

Lung cancer remains the leading cause of cancer death, with an estimated ~1.3 million deaths worldwide each year.(32) Although activated EGFR has been identified in non-small cell lung cancer (NSCLC), the major subtype of lung cancer, and specific inhibitors against EGFR provide effective treatment modalities, this type of genetic mutation is found preferentially in non-smokers, young women, and the Asian ethnic group.(33) For other NSCLC patients who are not eligible for the anti-EGFR treatments, there are currently few effective treatments to improve their outcome, unless cancer cells are completely removed by surgery.(34)

We have therefore chosen NSCLC as the target of our retroviral screening system. First, among our consecutive panel of NSCLC specimens, we examined the presence of known transforming genes in lung cancer; that is, mutated KRAS and mutated EGFR. To raise a retroviral library, from the specimens negative for either mutation we chose a sample of lung adenocarcinoma resected from a 62-year-old man with a smoking history. A total of >1.4 × 106 independent retroviral clones (with a mean cDNA size of 1.81 kb) were obtained from the specimen, and were used to infect 3T3 cells for the focus formation assay.

Dozens of transformed foci were readily identified in the assay, from which retroviral insert cDNA was rescued by PCR. Surprisingly, nucleotide sequences of the 5′ and 3′ parts of one cDNA corresponded to two different genes; one for microtubule-associated EML4,(35) and the other for the receptor-type PTK ALK.(36) Nucleotide sequencing of the cDNA revealed that the cDNA was a fusion between exons 1–13 of EML4 and exons 20–29 of ALK (transcript ID ENST00000389048 in the Ensembl database; http://www.ensembl.org/index.html), thus encoding a fusion-type PTK between the amino-terminal half of EML4 and the intracellular region of ALK (Fig. 2).(13)

Figure 2.

Anaplastic lymphoma kinase (ALK) fusion proteins. Chromosome rearrangements result in the generation of fusion genes that encode an oligomerization domain (including the coiled coil domain) of nucleophosmin (NPM), tropomyosin (TPM3) or echinoderm microtubule-associated protein like-4 (EML4) fused to the intracellular region of the receptor-type protein tyrosine kinase ALK. WD, WD-repeat domain.

Anaplastic lymphoma kinase was originally identified in anaplastic large cell lymphoma with t(2;5), as a fusion protein to nucleophosmin (NPM).(37,38) NPM–ALK plays an essential role in the lymphomagenesis of this subtype, and is a promising target for therapeutic compounds,(39,40) as is the case for BCR–ABL. In addition to NPM–ALK, ALK kinase may be fused, albeit at a lower frequency, to different partner proteins in the lymphoma through various chromosome translocations, giving rise to TRK-fused gene (TFG)–ALK, 5-aminoimidazole-4-carboximide ribonucleotide formyltransferase/IMP cyclohydrolase (ATIC)–ALK, clathrin, heavy chain (CLTC)–ALK, and others.(41) Subsequently, a non-hematological neoplasm, inflammatory myofibroblastic tumor (IMT) was also shown to harbor ALK fusion proteins such as TPM3–ALK, tropomyosin (TPM)4–ALK, and CLTC–ALK.(41) However, any recurrent translocation involving the ALK locus had not been reported for epithelial tumors before our discovery of EML4–ALK.

Interestingly, the ALK part of our EML4–ALK cDNA starts from exon 20 of ALK, which is also the fusion point in the vast majority of the other ALK fusion cDNA molecules, suggesting the presence of a common fragile locus within intron 19 of the ALK gene (Fig. 2). Both the EML4 and ALK genes are mapped closely to the short arm of human chromosome 2 in opposite directions (Fig. 3a). Therefore, a chromosome segment encompassing the EML4 and ALK loci has to become inverted to produce EML4–ALK cDNA. We have indeed succeeded to amplify by PCR a genome fragment from a NSCLC specimen that contained the fusion point between the EML4 and ALK genes.(13) In this adenocarcinoma, the EML4 gene was disrupted at a position 3.6 kb downstream of exon 13, and inverted to become ligated to a position ~300 bp upstream of ALK exon 20, proving the presence of inv(2)(p21p23) in the cancer cells.

Figure 3.

Diversity in the fusion points between echinoderm microtubule-associated protein like-4 (EML4) and anaplastic lymphoma kinase (ALK). (a) EML4 and ALK map in opposite orientations to the short arm of human chromosome 2 (shown in the genome browser of the University of California, Santa Cruz; http://genome.ucsc.edu/cgi-bin/hggateway). Intron 13 of EML4 is ligated to intron 19 of ALK through a chromosome rearrangement, inv(2)(p21p23), generating the EML4–ALK (variant 1) fusion gene. Primers flanking the fusion point can be used for the molecular detection of EML4–ALK-positive tumors by polymerase chain reaction. Arrows indicate the direction of transcription. (b) Exon boundaries of EML4 for possible in-frame fusion to exon 20 of ALK are shown as vertical bars together with the exon numbers at the corresponding positions in the EML4 protein. Reverse tramscription–polymerase chain reaction screening has identified variants (V) 1, 2, 3, and 5 of EML4–ALK, in which exons 13, 20, 6, and 2, respectively, of EML4 cDNA are fused to exon 20 (e20) of ALK cDNA. Unexpectedly, another in-frame fusion was identified in variant 4 cDNA, in which exon 14 of EML4 was fused via an 11-bp sequence of unknown origin to the nucleotide at position 50 of ALK exon 20. CC, coiled coil domain; HELP, hydrophobic EMAL-like protein domain; WD, WD-repeat domain.

Therefore, despite the previous notion that epithelial tumors seldom carry fusion-type oncogenes, we discovered an example of a fusion-type PTK with marked oncogenic activity in lung cancer, generated through a chromosome translocation, as is the case for BCR–ABL1, NPM–ALK, and translocation, ETS, leukemia (TEL)–Janus kinase 2 (JAK2) in hematological malignancies.(42) Because mutated EGFR and KRAS have been found recurrently in NSCLC cells, it is of clinical relevance whether EML4–ALK coexists in cancer cells with active EGFR or KRAS. Interestingly, the presence of EML4–ALK seems to be mutually exclusive to that of EGFR or KRAS mutations in NSCLC,(13,15,43,44) albeit with some exceptions.(45) Therefore, it is likely that EML4–ALK-positive lung cancer forms a subgroup among NSCLC, distinct from that positive for mutated EGFR or KRAS.

Molecular detection of EML4–ALK-positive NSCLC

One of the main reasons for the poor prognosis in lung cancer is the lack of sensitive detection methods that can capture tumor cells at early clinical stages (where tumors may be surgically removed). Although pathological examination of sputa and other clinical specimens is used routinely for the diagnosis of lung cancer, reliable detection with such systems usually requires that cancer cells occupy at least a small percentage of the total cells in these specimens. Therefore, patients diagnosed with this technique as having lung cancer are often at advanced clinical stages already.

In contrast, EML4–ALK-positive cells may be detected in a very sensitive way. As EML4 and ALK are mapped to chromosome 2p in opposite directions in normal cells, a set of PCR primers (one at exon 13 of EML4 and the other at exon 20 of ALK; Fig. 3a) will not generate any specific PCR products from cDNA of normal cells or of cancer without inv(2)(p21p23). Therefore, RT-PCR of the cDNA (or PCR of the genome fusion points) should become a highly sensitive yet reliable detection method for EML4–ALK-positive tumors. Given the high sensitivity of PCR, it is even expected that one cancer cell out of 105–106 normal cells in sputa may be detected, which would significantly help to identify individuals with lung cancer at early resectable stages. Soda et al. indeed succeeded in capturing 10 cells/mL of EML4–ALK-positive cells in sputum by RT-PCR.(13) It would therefore be of great importance to test the idea that such RT-PCR-based detection with sputa may be useful as a general screening method for early stages of NSCLC (among individuals with chronic cough or sputa, for instance).

Once detected with such screening systems, individuals positive for EML4–ALK may undergo surgical resection of tumors or receive chemotherapies with compounds that specifically suppress ALK activity. Just like the case of BCR–ABL1 in CML, EML4–ALK detection will likely play a pivotal role in the diagnosis of NSCLC positive for the fusion gene. In this regard, it is mandatory that every EML4–ALK-positive tumor be identified accurately by the diagnostic PCR system. There is a caveat, however, that the break and fusion points within the EML4 and ALK loci may be more divergent than previously appreciated.

Soda et al. first discovered two variants of EML4–ALK: exon 13 of EML4 fused to exon 20 of ALK in variant 1, and exon 20 of ELM4 fused to exon 20 of ALK in variant 2.(13) In addition, we and Pasi A. Jänne and colleagues (Dana-Farber Cancer Institute) have recently identified two more variants (variants 3a and 3b, which connect exons 6a and 6b, respectively, of EML4 to exon 20 of ALK) (Fig. 3b).(14,45) Further variants that connect various exons of EML4 to ALK are being identified by a number of groups worldwide.(44–46)

In addition to exons 6 (variant 3), 13 (variant 1), and 20 (variant 2) of EML4, an in-frame fusion to exon 20 of ALK can occur with exon 2, 18, or 21 of EML4 (Fig. 3b). Given that the amino-terminal coiled coil domain of EML4 is responsible for the oligomerization of EML4–ALK (see below) and that exon 2 of EML4 encodes the entire coiled coil domain, all of these possible fusion genes would encode EML4–ALK proteins containing the coiled coil domain and therefore likely produce oncogenic EML4–ALK kinases. To screen for all variants (both known and unknown) of EML4–ALK and to estimate the frequency of such oncogenes in human cancers, Takeuchi et al. have developed a single-tube multiplex RT-PCR system that captures all possible in-frame fusions between EML4 and ALK.(46) From screening of lung adenocarcinoma specimens (n = 253), they have identified a total of 11 samples (4.35%) carrying variants 1, 2, 3, or unknown isoforms (referred to as variants 4 and 5) of EML4–ALK.

Unexpectedly, in one of the new isoforms (variant 4), exon 14 of EML4 is connected to an unknown sequence of 11 bp, and further fused to a nucleotide at position 50 of exon 20 of ALK. Although exon 14 of EML4 is not expected to produce an in-frame fusion to exon 20 of ALK, insertion of the unknown 11-bp sequence and its ligation to a position within the ALK exon allows an in-frame connection between the two genes.

Additionally, exon 2 of EML4 is fused to exon 20 of ALK or to a nucleotide 117 bp upstream of exon 20 of ALK, giving rise to variants 5a and 5b of EML4–ALK, respectively. Takeuchi et al. further successfully isolated full-length cDNA for variants 4, 5a, and 5b of EML4–ALK, and confirmed the transforming potential of all isoforms.(46) Takeuchi et al. have also screened for EML4–ALK cDNA, with the same multiplex RT-PCR technique, among other solid tumors (n = 403) including squamous cell carcinoma (n = 71) and small cell carcinoma (n = 21) of the lung. Interestingly, none of these tumors were positive for the fusion cDNA, indicating specificity of the EML4–ALK oncogene to lung cancer (especially adenocarcinoma).

Similarly, Wong et al. have tried to identify all possible in-frame fusions between EML4 and ALK among a panel of NSCLC specimens (n = 240), discovering 13 cases (5.42%) positive for variants 1, 2, 3, and an unknown isoform of EML4–ALK.(44) Notably, variant 3 was the most frequent isoform (n = 8) in their Chinese cohort. Based on these data, the proportion of EML4–ALK-positive tumors in NSCLC seems to be ~5% in the Asian ethnic group, and may be lower in the others.(44,46,47)

It should be noted that all subtypes of EML4–ALK have not always been assayed in the published screenings and, further, that there may still be other variants not yet discovered. Therefore, to estimate the true prevalence of EML4–ALK-positive tumors within a given ethnic group, it is necessary to examine, in large cohorts, all possible in-frame fusions between EML4 and ALK among the subjects. Additionally, given the increasing number of EML4–ALK variants, I personally hope that researchers may, in the near future, develop a more reasonable and uniform nomenclature system for such variants (E13; A20 for variant 1, and E6a; A20 for variant 3a, for instance) than the current one (variants 1, 2, 3, etc).

With regard to other diagnostic tools for EML4–ALK-positive tumors, immunohistochemical detection of EML4–ALK proteins in specimens obtained by biopsy or surgical resection would be a convenient screening system in clinics. In anaplastic large cell lymphoma, such screening with antibodies to the intracellular region of ALK has been used routinely to detect lymphoma positive for NPM–ALK.(48) Unfortunately, however, it is often difficult to stain EML4–ALK with such antibodies in NSCLC that are positive for EML4–ALK mRNA (K. Takeuchi and K. Inamura, personal communication). This discrepancy may be due to: (i) the weaker promoter activity of the EML4 gene (which drives the expression of EML4–ALK) compared to that of NPM (which drives the expression of NPM–ALK), or (ii) a lower stability of the EML4–ALK protein than that of NPM–ALK. Further improvements in the sensitivity of immunohistochemical detection of EML4–ALK would be desirable to apply such systems to routine pathological screenings.

Transforming activity of EML4–ALK

How does fusion to EML4 induce a marked transforming potential in ALK? A number of fusion-type PTK carry an oligomerization motif within the fusion partner regions, which thereby leads to dimerization and autophosphorylation of the corresponding kinase domain.(49,50) Consistent with this notion, the NPM region in the NPM–ALK protein was shown to be essential in the oligomerization and transforming potential of this fusion kinase (Fig. 2).(51) Similarly, TPM3–ALK and TPM4–ALK found in IMT and EML4–ALK in NSCLC all carry a coiled coil domain within the fusion partners, which may act as an oligomerization motif. Indeed, EML4–ALK homodimerizes in cells, but internal deletion of the basic domain of EML4 (which contains the coiled coil domain) severely hampers such physical association. Accordingly, this mutant form of EML4–ALK loses its marked tumor-formation activity in vivo, and has decreased tyrosine kinase activity in vitro.(13) It should be noted, however, that truncation of subdomains other than the coiled coil domain of EML4–ALK also affects its transforming potential, suggesting that self-oligomerization is not the only mechanism to induce oncogenic potential in EML4–ALK.(13)

As wild-type ALK is a PTK with a single transmembrane region, it is presumed that the in vivo function of ALK is that of a cell surface receptor for specific ligands (probably growth factors). Unfortunately, however, such ligands have not been isolated in mammals. In Drosophila melanogaster, a protein homologous to human and mouse ALK is expressed in visceral mesoderm in the embryo, and malfunctions in Alk lead to visceral mesoderm defects in early embryogenesis, which resembles the phenotype of dysfunction in a secreted protein, jelly belly (Jeb). Upon binding to Jeb, Alk becomes activated to trigger the Ras–mitogen-activated protein kinase cascade and transcriptional activation of a subset of genes,(52) suggesting that Jeb is the ligand of Alk in fly.

In mammals, however, pleiotrophin(53) and midkine(54) are proposed ligands of ALK and are expressed specifically in brain and spinal cord.(55,56) Pleiotrophin may bind ALK at a low dissociation constant and induce tyrosine phosphorylation on ALK as well as putative downstream effector molecules.(53) However, other cellular receptors for pleiotrophin have also been identified and, hence, it is not yet clear if the observed effects of pleiotrophin are mediated mainly through ALK.(57,58)

Although it is theoretically possible that the extracellular region of ALK may act as its own ligand, the Jeb–Alk interaction in fly suggests that ALK likely functions as a cellular receptor for specific ligands in mammalian cells as well. Presumably, upon ligand binding (and only upon such binding), ALK becomes oligomerized and activated to trigger a transient growth signal in cells (Fig. 4). However, EML4–ALK and other ALK fusion proteins are constitutively oligomerized through the binding motif within the corresponding fusion partners, and activated to maintain a persistent mitogenic signal that finally leads to tumor formation.

Figure 4.

Activation mechanisms for anaplastic lymphoma kinase (ALK) and echinoderm microtubule-associated protein like-4 (EML4)–ALK. Wild-type ALK is thought to undergo transient homodimerization in response to binding of a specific ligand, resulting in its activation and mitogenic signal transduction. In contrast, EML4–ALK is constitutively oligomerized via the coiled coil domain (CC) of EML4, resulting in persistent mitogenic signaling that eventually leads to malignant transformation.

A pivotal role of ALK fusion proteins in malignant transformation has been demonstrated clearly in the case of NPM–ALK. Retroviral transduction of NPM–ALK mRNA in bone marrow cells induces lymphoma-like disorders in mice,(59,60) and transgenic mice with Vav promoter-driven expression of NPM–ALK results in the generation of lymphoma.(61) Furthermore, injected lymphoma cells positive for NPM–ALK were effectively eradicated from mice by treatment with a specific inhibitor against ALK,(40) suggesting that the activated kinase potential of NPM–ALK is central to the lymphomagenesis.

A few experiments also support such a pivotal role for EML4–ALK in lung cancer. Koivunen et al. have found that treatment with a specific inhibitor for ALK (TAE684) induces rapid cell death of one NSCLC cell line (NCI-H3122) in culture, which harbors variant 1 of EML4–ALK.(45) Another NSCLC cell line, NCI-H2228, was shown to be positive for variant 3 of EML4–ALK, but TAE684 treatment failed to induce drastic effects in this these cells. However, McDermott et al. did find partial inhibition of cell viability (66% reduction compared to the control experiments) in NCI-H2228 after treatment with TAE684, as well as in NCI-H3122 (75% reduction).(62) Similarly, we found marked growth suppression of NCI-H2228 with an ALK inhibitor only in a spheroid culture system (Fig. 5),(14) not in a regular in vitro culture. These data thus indicate that EML4–ALK may be the principle transforming protein in some NSCLC cells (such as NCI-H3122) that are therefore fully sensitive to ALK inhibitors. However, it is likely that at least one other potential transforming protein is present with EML4–ALK in other NSCLC (e.g. NCI-H2228), and thus inhibition of ALK enzymatic activity provides significant but incomplete growth suppression.

Figure 5.

Suppression of the growth of an echinoderm microtubule-associated protein like-4 (EML4)–anaplastic lymphoma kinase (ALK)-positive cell line with an ALK inhibitor. The non-small cell lung cancer cell line NCI-H2228, which is positive for EML4–ALK (variant 3), was maintained in spheroid culture for 2 days and was photographed after an additional 5 days of incubation in the absence (upper panel) or presence (lower panel) of 5 nmol/L 2,4-pyrimidinediamine.(14) Scale bars = 4 mm.

With regard to the in vivo role of EML4–ALK, it is important to test if EML4–ALK activity can induce lung cancer in vivo. To address this issue, Soda et al. have recently generated transgenic mice in which EML4–ALK mRNA is transcribed specifically in lung epithelial cells by the use of a promoter fragment of the surfactant protein C gene.(63) Surprisingly, all independent lines of such mice develop hundreds of adenocarcinoma nodules in both lungs at only a few weeks after birth, proving for the first time the marked transforming activity of EML4–ALK in vivo (Soda et al., unpublished data). More importantly, treatment of such mice with a chemical compound that suppresses ALK activity rapidly cleared those nodules from the lungs. Therefore, it is likely that EML4–ALK-positive lung cancer cells are at least partially dependent on their PTK activity for growth, and any means to suppress this activity would be a promising strategy for treating this intractable disorder.

Concluding remarks

The data from EML4–ALK transgenic mice clearly show the central role of this fusion kinase in lung cancer, and such mice also provide an efficient in vivo screening system for ALK inhibitors. Recently, treatment with an ALK inhibitor was shown to suppress or inhibit the growth of some neuroblastoma cell lines, in addition to NSCLC and anaplastic large cell lymphoma.(62,64) Because the EML4–ALK and NPM–ALK fusion genes are present in the latter two, such inhibitor-sensitive neuroblastoma cell lines may also possess other ALK mutants. These data suggest that tumors of any tissue origin may be treated with the same ALK inhibitors provided that they carry any one of the oncogenic ALK mutants. Therefore, ‘ALKoma’(65,66) may form a novel clinical entity as is the case for v-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog (KIT) mutant-positive tumors in acute myeloid leukemia, mastocytosis,(67,68) and gastrointestinal stromal tumor.(69)

However, given the marked diversity in the sensitivity of EML4–ALK-positive cell lines to ALK inhibitors,(45,62) identification of coexisting oncogenes in ALK mutant-positive tumors would be valuable to increase the efficacy of treatments with ALK inhibitors.

Furthermore, as various ALK fusion proteins have divergent subcellular localizations (probably dependent on the nature of fusion partner proteins),(46,70) downstream signaling pathways may vary among them. Indeed, although signal transducer and activator of transcription (STAT) proteins likely play a key role in the mitogenic signaling of NPM–ALK, such a role is unlikely for EML4–ALK(46,71) and some other ALK fusions.(70) It is therefore of great clinical relevance to decipher the profiles of oncogenes or tumor-suppressor genes and downstream proteins for each ALK fusion in each cancer subtype.

Acknowledgments

I apologize to all of the authors whose work could not be included in this manuscript owing to space constraints. I thank the members of our laboratory for their support and excellent work, and Yuichi Ishikawa and Kengo Takeuchi for helpful suggestions and discussion. This work was supported in part by a grant for Research on Human Genome Tailor-made from the Ministry of Health, Labor, and Welfare of Japan, and by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

Ancillary