Multiplex Reverse Transcription-PCR Screening for EML4-ALK Fusion Transcripts

Purpose: EML4-ALK is a fusion-type protein tyrosine kinase that is generated by inv(2)(p21p23) in the genome of non–small cell lung cancer (NSCLC). To allow sensitive detection of EML4-ALK fusion transcripts, we have now developed a multiplex reverse transcription-PCR (RT-PCR) system that captures all in-frame fusions between the two genes.

Experimental Design: Primers were designed to detect all possible in-frame fusions of EML4 to exon 20 of ALK, and a single-tube multiplex RT-PCR assay was done with total RNA from 656 solid tumors of the lung (n = 364) and 10 other organs.

Results: From consecutive lung adenocarcinoma cases (n = 253), we identified 11 specimens (4.35%) positive for fusion transcripts, 9 of which were positive for the previously identified variants 1, 2, and 3. The remaining two specimens harbored novel transcript isoforms in which exon 14 (variant 4) or exon 2 (variant 5) of EML4 was connected to exon 20 of ALK. No fusion transcripts were detected for other types of lung cancer (n = 111) or for tumors from 10 other organs (n = 292). Genomic rearrangements responsible for the fusion events in NSCLC cells were confirmed by genomic PCR analysis and fluorescence in situ hybridization. The novel isoforms of EML4-ALK manifested marked oncogenic activity, and they yielded a pattern of cytoplasmic staining with fine granular foci in immunohistochemical analysis of NSCLC specimens.

Conclusions: These data reinforce the importance of accurate diagnosis of EML4-ALK–positive tumors for the optimization of treatment strategies.

Chromosome rearrangement is a major mechanism giving rise to transforming potential in human cancers, especially in hematologic malignancies (1). A balanced translocation between chromosomes 9 and 22, for instance, generates an activated protein tyrosine kinase, BCR-ABL, that plays an essential role in the pathogenesis of chronic myeloid leukemia (2). The gene for another protein tyrosine kinase, ALK, is fused to those for NPM1 or other partner proteins in anaplastic lymphoma and soft tissue tumors, resulting in an increase in the kinase activity of ALK (3).

Mitelman et al. have suggested that chromosome translocations, in addition to being common in hematologic malignancies, are not rare in epithelial tumors (4, 5). These researchers also proposed that the genetic mechanisms underlying oncogenesis might not differ fundamentally between hematologic and epithelial malignancies, and that the current apparent difference in the frequency of chromosomal translocations between these two types of cancer is likely to disappear with the advent of new and more powerful investigative tools.

Consistent with this notion, recurrent chromosome rearrangements involving genes for ETS transcriptional factors have been identified in many cases of prostate cancer and may contribute to the hypersensitivity of prostate cancer cells to androgens (6, 7). In addition, we recently discovered another recurrent chromosome translocation in non–small cell lung cancer (NSCLC; ref. 8), a major cause of cancer deaths in humans. A small inversion within the short arm of chromosome 2, inv(2)(p21p23), was found to be present in <10% of NSCLC cases and to give rise to a novel fusion-type tyrosine kinase, EML4-ALK, that exhibited marked transforming activity in vitro (8). Transgenic mice that specifically express EML4-ALK in lung epithelial cells were also found to develop hundreds of adenocarcinoma nodules in both lungs at only a few weeks after birth, and such nodules disappeared rapidly in response to oral administration of a specific inhibitor of the catalytic activity of ALK.⁵

These data thus indicate that EML4-ALK plays a pivotal role in malignant transformation in lung cancer, and they suggest that chemical compounds that inhibit the tyrosine kinase activity of EML4-ALK may provide an effective treatment for EML4-ALK–positive lung cancer. The selection of suitable drugs for individuals with lung cancer will thus require accurate determination of the absence or presence of the EML4-ALK fusion gene in biopsy specimens.

Given that EML4 and ALK map in opposite orientations within the short arm of chromosome 2, reverse transcription-PCR (RT-PCR) analysis with primers designed to amplify the fusion points of EML4-ALK transcripts would not be expected to yield specific products from normal cells or cancer cells without inv(2)(p21p23). Such analysis should thus provide a highly reliable and sensitive means to detect EML4-ALK in clinical specimens. Given that sputum has been shown to be a suitable specimen for such molecular diagnosis of EML4-ALK positivity (8), detection of EML4-ALK–positive cells by RT-PCR analysis of sputa may be effective for the identification of lung cancer at early clinical stages. The accurate diagnosis of EML4-ALK–positive tumors, however, will require that all isoforms of EML4-ALK are detected.

The fusion of intron 13 or 20 of EML4 to intron 19 of ALK gives rise to variant 1 or 2 of EML4-ALK, respectively (8). We have recently discovered another isoform (variant 3) of EML4-ALK in which intron 6 of EML4 is ligated to intron 19 of ALK (9). Theoretically, in addition to such fusion of exons 6, 13, and 20 of EML4, an in-frame fusion to exon 20 of ALK can occur with exons 2, 18, or 21 of EML4. Given that the amino-terminal coiled-coil domain of EML4 is responsible for the dimerization and constitutive activation of EML4-ALK (8) 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 establish a highly sensitive and accurate PCR-based screening system for EML4-ALK–positive cancer, we have now developed a high-throughput multiplex RT-PCR assay for the detection of all potential EML4-ALK in-frame fusion transcripts. Among a consecutive series of lung adenocarcinoma specimens (n = 253) as well as other solid tumor samples (n = 403), we have now identified a total of 11 lung adenocarcinoma specimens positive for EML4-ALK, two of which harbor previously unidentified fusion mRNAs.

Clinical samples and RNA extraction. This study was done with clinical samples from 253 lung adenocarcinomas, 90 other NSCLCs (71 squamous cell carcinomas, 7 adenosquamous carcinomas, 7 large cell carcinomas, 2 pleomorphic carcinomas, and 3 large cell endocrine carcinomas), 21 small cell lung carcinomas, 50 breast carcinomas, 46 renal cell carcinomas, 48 colon carcinomas, 13 prostate carcinomas, 29 urothelial carcinomas, 33 gastric carcinomas, 10 uterine carcinomas, 9 hepatocellular carcinomas, 8 pancreatic carcinomas, and 46 malignant fibrous histiocytomas. All specimens were collected with the approval of the ethical committee at the Cancer Institute Hospital (Tokyo, Japan) and with the informed consent of individuals undergoing surgery from May 1995 to July 2003. The NSCLC cases were consecutive and spanned a period of 19 mo. Histologic diagnosis of NSCLC was made according to the WHO classification (10). All lesions were grossly dissected, rapidly frozen in liquid nitrogen, and stored at −80°C until RNA extraction with an RNeasy Mini Kit (Qiagen). RNA quality and the absence of contamination with genomic DNA were verified by formaldehyde-agarose gel electrophoresis.

Multiplex RT-PCR analysis and nucleotide sequencing. Total RNA was subjected to RT with random primers and SuperScript III reverse transcriptase (Invitrogen). For detection of EML4-ALK fusion cDNAs, multiplex PCR analysis was done with AmpliTaq Gold DNA polymerase (Applied Biosystems), the forward primers EML4 72F (5′-GTCAGCTCTTGAGTCACGAGTT-3′) and Fusion-RT-S (5′-GTGCAGTGTTTAGCATTCTTGGGG-3′), and the reverse primer ALK 3078RR (5′-ATCCAGTTCGTCCTGTTCAGAGC-3′). The GAPDH cDNA was amplified by PCR with the primers 5′-GTCAGTGGTGGACCTGACCT-3′ and 5′-TGAGCTTGACAAAGTGGTCG-3′. For amplification of EML4-ALK fusion cDNAs, the samples were incubated at 94°C for 10 min and then subjected to 35 cycles of denaturation at 94°C for 1 min, annealing at 64°C for 1 min, and polymerization at 72°C for 1 min. For amplification of GAPDH cDNA, the samples were subjected to 35 cycles of 94°C for 1 min, 58°C for 30 s, and 72°C for 30 s. Virtual gel electrophoresis of multiplex RT-PCR products was done with a 2100 Bioanalyzer (Agilent Technologies).

The primers used for direct amplification of the fusion points of individual cDNAs were 5′-AGGAGAGAACTCAGCGACACTACC-3′ and 5′-TCCACGCTCAAAAGTGCCAAGTCC-3′ for variant 4 and 5′-GCTTTCCCCGCAAGATGGACGG-3′ and 5′-AGCTTGCTCAGCTTGTACTCAGGG-3′ for variant 5. Full-length cDNAs for EML4-ALK variants were amplified with PrimeSTAR DNA polymerase (Takara Bio) and the primers 5′-ACTCTGTCGGTCCGCTGAATGAAG-3′ and 5′-CCACGGTCTTAGGGATCCCAAGG-3′.

Fluorescence in situ hybridization analysis. Surgically resected lung cancer tissue was fixed in 20% formalin, embedded in paraffin, sectioned at a thickness of 4 μm, and placed on glass slides. The unstained sections were processed with a Histology FISH Accessory Kit (Dako), subjected to hybridization with fluorescently labeled bacterial artificial chromosome clone probes for EML4 and ALK (GSP Laboratory) or for genomic regions upstream and downstream of the ALK break point (Dako), stained with 4,6-diamidino-2-phenylindole, and examined with a fluorescence microscope (BX51; Olympus).

Immunohistochemical analysis. Unstained paraffin-embedded sections were depleted of paraffin with xylene, rehydrated with a graded series of ethanol solutions, and then subjected to heat-induced antigen retrieval with Target Retrieval Solution pH 9.0 (Dako) before immunohistochemical staining with a mouse monoclonal antibody to ALK (ALK1, Dako) at a dilution of 1:20. Immune complexes were detected with the use of an EnVision+DAB system (Dako) with minor modifications.⁶

Transforming potential of EML4-ALK proteins. Protein analysis of EML4-ALK variants was done as described previously (8). In brief, the EML4-ALK variant 4, 5a, or 5b cDNAs were fused with an oligonucleotide encoding the FLAG epitope tag and inserted into the retroviral expression plasmid pMXS (11). The resulting plasmids and similar pMXS-based expression plasmids for EML4-ALK variant 1, variant 1(K589M), variant 2, variant 3a, and variant 3b were individually introduced into HEK293 cells. Lysates of the transfected cells were subjected to immunoprecipitation with antibodies to FLAG, and the resulting precipitates were subjected either to immunoblot analysis with the same antibodies or to an in vitro kinase assay with the YFF peptide (12). Mouse 3T3 fibroblasts were also infected with recombinant retroviruses for each of the EML4-ALK variants or wild-type ALK and were then cultured for 12 d for a focus formation assay. The same set of 3T3 cells was injected s.c. into nu/nu mice, and tumor formation was examined after 20 d.

Multiplex RT-PCR screening for EML4-ALK fusion transcripts in lung adenocarcinoma. As described above, exons 2, 6, 13, 18, 20, and 21 of EML4 may participate in an in-frame fusion to exon 20 of ALK (Fig. 1A). To identify all possible EML4-ALK fusion cDNAs in a single-tube experiment, we designed a mixture of two sense primers (one targeted to exon 2 and the other to exon 13 of EML4) and a single antisense primer (targeted to exon 20 of ALK) and did multiplex RT-PCR with these primers and total cDNA preparations from tumor specimens. The exon 2 primer for EML4 would be expected to generate a PCR product of 458 bp with the exon 2 (EML4)-exon 20 (ALK) fusion cDNA or of 917 bp with the exon 6-exon 20 fusion cDNA (variant 3). In addition, the exon 13 primer for EML4 would be expected to generate PCR products of 432, 999, 1,185, or 1,284 bp with the exon 13-exon 20 (variant 1), exon 18-exon 20, exon 20-exon 20 (variant 2), and exon 21-exon 20 fusion cDNAs, respectively.

Virtual gel electrophoresis of the multiplex RT-PCR products (Fig. 1B) revealed that 11 samples (4.35%) were positive for EML4-ALK cDNA among a consecutive series of 253 lung adenocarcinoma specimens, including those examined in our previous studies (8, 9, 13). All of the specimens previously shown to harbor EML4-ALK (two cases with variant 1, three with variant 2, and two with variant 3) were faithfully detected with our multiplex RT-PCR system. No specific PCR products were obtained for other types of lung cancer (n = 111) or other solid tumors (n = 292). Nucleotide sequencing of the PCR products for the newly identified positive cases revealed that one specimen was positive for variant 1 and another for variant 3 of EML4-ALK, but that the remaining two specimens harbored previously unidentified variants (Fig. 1B and C). Exon 14 of EML4 was ligated to a position within exon 20 of ALK in the product from tumor ID no. 8398 (designated variant 4), whereas exon 2 of EML4 was ligated to exon 20 of ALK in the product from tumor ID no. 8993 (designated variant 5).

Structure of EML4-ALK variant 4 cDNA. To verify the presence of novel EML4-ALK variants in the cancer cells, we first did direct RT-PCR analysis for the cDNA of tumor ID no. 8398 with a new set of primers encompassing the putative fusion point of variant 4. This analysis showed the presence of the fusion cDNA (Fig. 2A). Nucleotide sequencing of the PCR product revealed that exon 14 of EML4 was fused to an unknown sequence of 11 bp, which in turn was connected to the nucleotide at position 50 of exon 20 of ALK (Fig. 2B). (We failed to detect a region of the human genome (build 36) homologous to the 11-bp connecting sequence in a BLAST search.⁷

) 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. Fusion cDNAs in which the point of connection is located within, rather that at the 5′ terminus of, exon 20 of ALK have also been described for MSN-ALK (14) and MYH9-ALK (15).

We further examined whether a full-length cDNA encoding such an unexpected EML4-ALK variant could be isolated from the cancer cells. For this purpose, we designed a sense primer targeted to the 5′ untranslated region of EML4 cDNA as well as an antisense primer targeted to the 3′ untranslated region of ALK cDNA. Direct RT-PCR analysis with this primer set yielded a single PCR product of ∼3.4 kbp with total cDNA of tumor ID no. 8398 (Supplementary Fig. S1A). Complete nucleotide sequencing of the PCR product revealed that the cDNA contained an open reading frame for 1,097 amino acids comprising residues 1 to 547 of human EML4, residues 1,075 to 1,620 of human ALK, and 4 amino acids of unknown origin between these two sequences (Supplementary Fig. S1B). The isolation of a full-length cDNA containing the 11-bp insert indicated that the variant 4 protein was likely expressed in the cancer cells.

Structure of EML4-ALK variant 5 cDNAs. We similarly investigated the presence of variant 5 mRNA in the cells of tumor ID no. 8993. Direct RT-PCR analysis to amplify the fusion point of this variant cDNA yielded two independent products of 298 and 415 bp (Fig. 2C). Nucleotide sequencing of each product revealed that the former contained exon 2 of EML4 and exon 20 of ALK, as expected, whereas in the latter, exon 2 of EML4 was connected to a position within intron 19 of ALK located 117 bp upstream of exon 20 (Fig. 2D). These fusion constructs were designated variants 5a and 5b, respectively.

Although no mRNAs or expressed sequence tags in the nucleotide sequence database were found to contain the 117-bp sequence of intron 19 of ALK, the human genome sequence surrounding the 5′ terminus of this 117-bp sequence is AG-GT (Fig. 2D), which conforms to the consensus sequence for a splicing acceptor site. To show that such a cryptic exon is indeed involved in the production of an oncogenic kinase, we attempted to detect full-length cDNAs for variants 5a and 5b from total cDNA of tumor ID no. 8993. A doublet of PCR products of ∼2.0 kbp was obtained (Supplementary Fig. S1A), and nucleotide sequencing of these products revealed that they indeed encode EML4-ALK variant 5a and 5b proteins (Supplementary Fig. S1C). Genomic PCR and fluorescence in situ hybridization (FISH) analyses further revealed that the cells of tumor ID no. 8993 harbor a single EML4-ALK fusion gene, suggesting that variant 5a and 5b mRNAs are generated by alternative splicing of the primary transcript of this single fusion gene (see below).

Detection of the EML4-ALK fusion genes by FISH. To confirm the rearrangements involving the ALK locus in the specimens harboring variants 4 and 5 of EML4-ALK cDNA, we did FISH analysis with tissue sections. We first designed a FISH-based “fusion assay” for EML4 and ALK genes. Bacterial artificial chromosome fragments encompassing the entire genes were fluorescently labeled green and red, respectively. An overlapping signal for both probes was readily identified in a merged image for the tumor cells harboring variants 4 or 5 of EML4-ALK (Fig. 3A). To confirm further the breakage of the ALK locus, we did an “ALK split assay” with bacterial artificial chromosome fragments encompassing the 5′ or 3′ regions of the locus and labeled green and red, respectively. In this assay, the normal ALK locus would be expected to yield an overlapping signal, whereas a pair of separate green and red signals would indicate genomic breakage within ALK. As expected, a proportion of cells of tumor ID no. 8398 or no. 8993 in the histologic sections generated one overlapping signal and one pair of split signals (Fig. 3B), suggesting that these tumor cells each have at least one normal and at least one rearranged ALK locus.

These data, together with genomic PCR analysis (data not shown), thus indicated that the cells of each of these tumors harbor one normal chromosome 2 and a chromosome 2 with an inv(2)(p21p23) rearrangement. The other EML4-ALK cDNA-positive specimens (variants 1 to 3) in this cohort showed a similar FISH labeling profile, consistent with the presence of the corresponding EML4-ALK rearrangements (data not shown).

Detection of EML4-ALK proteins in situ. To detect EML4-ALK proteins in the cancer cells, we did immunohistochemical analysis with the ALK1 monoclonal antibody to ALK (16). The cytoplasm of tumor cells harboring EML4-ALK variant 1 (ID no. 9034), variant 4 (ID no. 8398), or variant 5 (ID no. 8993) manifested a diffuse pattern of immunoreactivity with fine granular concentrations (Fig. 3C). No normal pulmonary epithelial cells or lymphocytes in the sections of these specimens reacted with the antibody.

Transforming activity of EML4-ALK variants. We prepared expression plasmids for FLAG epitope–tagged EML4-ALK variants 1, 2, 3a, 3b, 4, 5a, and 5b, the predicted molecular sizes of which are 118,356; 146,913; 87,613; 88,874; 122,541; 71,046; and 74,867 Da, respectively. Each of these proteins, as well as a kinase-inactive mutant of EML4-ALK variant 1 (8), was expressed independently in HEK293 cells, immunoprecipitated, and subjected to immunoblot analysis with antibodies to FLAG. Each cDNA generated an EML4-ALK protein of the expected molecular size (Fig. 4A). The same immunoprecipitates were subjected to an in vitro kinase assay with the synthetic peptide YFF (12). Each variant protein (with the exception of the kinase-inactive mutant of variant 1) was shown to possess protein tyrosine kinase activity, with that of variants 3a, 3b, and 5b being most prominent (Fig. 4A).

To examine the transforming potential of the EML4-ALK variants, we transfected mouse 3T3 fibroblasts with the corresponding expression plasmids and then cultured the cells for 12 days. Transformed foci were readily detected for the cells expressing the variants of EML4-ALK but not for cells overexpressing wild-type ALK (Fig. 4B). Furthermore, s.c. injection of the transfected 3T3 cells into the shoulder of nude mice revealed that those expressing the various EML4-ALK isoforms, but not those overexpressing wild-type ALK, formed large tumors in vivo (Fig. 4B).

We have done multiplex RT-PCR analysis to detect all possible isoforms of EML4-ALK transcripts in NSCLC cells, and unexpectedly identified two novel subtypes of the fusion event. This finding was supported by detection of the corresponding fusion genes by genomic PCR and FISH analyses and by that of the encoded proteins by immunohistochemical analysis in the NSCLC cells. Together with the previously isolated variants (8, 9), we have to date identified a total of seven distinct isoforms of EML4-ALK (variants 1, 2, 3a, 3b, 4, 5a, and 5b). Given that each of these isoforms possesses marked transforming activity, they all likely play an important role in the development of NSCLC. Our failure to detect EML4-ALK cDNA in the other solid tumors (n = 313) examined suggests that EML4-ALK may be an oncogene specific to NSCLC, especially to lung adenocarcinoma.

In our multiplex RT-PCR analysis, a sense primer targeted to exon 2 of EML4 was designed to detect fusion events involving exon 2 or 6 of EML4, and PCR products of the expected sizes were indeed obtained with NSCLC specimens positive for such fusion events (variants 5 and 3, respectively). The other sense primer was targeted to exon 13 of EML4 and was designed to detect fusion events involving exon 13, 18, 20, or 21 of EML4. Given that we were able to readily amplify a specific product of 1185 bp corresponding to the fusion event involving exon 20 of EML4 (variant 2), it is likely that all possible fusions giving rise to PCR products up to this size would have been detected in our cohort. It should be noted, however, that a possible fusion between exon 21 of EML4 and exon 20 of ALK would be expected to generate a PCR product of 1,284 bp. Although the size difference between the 1,185- and 1,284-bp products is small (99 bp), it is still possible that our multiplex RT-PCR analysis failed to efficiently amplify the longer product and that there may be as-yet-undetected fusion events for EML4-ALK in our cohort.

All EML4-ALK isoforms manifested a similar subcellular distribution profile despite marked differences in the size and domain structure of the EML4 portions of these chimeric proteins. In addition, the intracellular signaling systems activated by EML4-ALK may be shared among variants 1 to 5 (Supplementary Fig. S2). The EML4 portion of variant 5 comprises only the coiled-coil domain. This domain of EML4 may therefore play an essential role not only in the dimerization and activation of EML4-ALK isoforms (8) but also in tethering EML4-ALK to specific subcellular components. The pattern of subcellular immunostaining for EML4-ALK (cytoplasmic staining with fine granular foci) was distinct from that for other ALK fusion proteins associated with other malignancies (17, 18), suggesting that the subcellular localization of ALK fusion kinases varies substantially. The first such fusion kinase to be identified, NPM-ALK, preferentially phosphorylates STAT3, which is thought to participate in mitogenic signaling by NPM-ALK (19–21). Five ALK fusion kinases (NPM-ALK, TFG-ALK, ATIC-ALK, TPM3-ALK, and CLTC-ALK) were shown to differ markedly in their abilities to transform 3T3 fibroblasts, to phosphorylate STAT3 and AKT, and to activate phosphoinositide 3-kinase (17). Furthermore, a proteomics approach to identify tyrosine-phosphorylated proteins failed to detect marked phosphorylation of STAT3 in NSCLC specimens positive for EML4-ALK (22). It is therefore likely that each ALK fusion kinase exerts its effects through fusion-specific (although possibly partially overlapping) downstream pathways. In addition, we detected slight differences in catalytic and transforming activities among the variants of EML4-ALK (Fig. 4). These differences are likely due to the different portions of EML4 present in the different variants, which may affect dimerization affinity or the recruitment of substrates.

In addition to EML4-ALK, NSCLC cells harbor other potent oncogenes such as mutant versions of EGFR or KRAS. These three oncogenes, however, were found to be mutually exclusive in our previous NSCLC cohort (8, 13), suggesting that EML4-ALK–positive NSCLC is a distinct subclass of lung cancer. Given that a selective inhibitor of the kinase activity of ALK rapidly induces cell death in EML4-ALK–positive cancer cells both in vitro (8, 9) and in vivo,⁸

determination of the presence or absence of EML4-ALK in a given tumor may in the future inform the choice of treatment strategy for NSCLC. The demonstration of the existence of multiple isoforms of EML4-ALK transcripts will necessitate optimization of the detection systems so that all isoforms are detected with a high accuracy and sensitivity.

During our revision process, a novel EML4-ALK fusion variant was reported by Koivunen et al. (Clin Cancer Res 2008;14:4275–83). They have designated it as variant 4, which is different from our variant 4 in the present study.

K. Takeuchi is a consultant providing advisory services to Dako for their antibodies.

We thank Kazuko Yokokawa, Motoyoshi Iwakoshi, Miyuki Kogure, and Tomoyo Kakita for technical assistance as well as Yuki Takano for help in preparation of the manuscript.