ALKoma: A Cancer Subtype with a Shared Target

Anaplastic lymphoma kinase (ALK) is a protein tyrosine kinase (PTK) that possesses a single transmembrane domain and consists of 1,620 amino acid residues in humans (Fig. 1; refs. 1, 2). The extracellular region of ALK contains 2 MAM (meprin, A5 protein, and protein phosphatase μ) domains (3) and a putative ligand-binding domain. ALK is relatively isolated in the phylogenetic tree of the PTK superfamily, with its kinase domain sharing the highest sequence identity with that of the leukocyte receptor tyrosine kinase (LTK; 79% identity) and that of the proto-oncoprotein ROS1 (50% identity).

In the mouse, expression of ALK is prominent in the brain and peripheral nervous system of developing embryos, but it decreases rapidly after birth. In the adult human, ALK is expressed at a low level in the central nervous system but is not detected in other tissues (4). Given these expression profiles, ALK has been thought to play an important role in the development or orchestration of the nervous system. Surprisingly, however, homozygous deletion of Alk in mice does not give rise to any apparent anomalies in morphology, internal organs, or fertility, suggesting that the gene is not essential for mouse development (5). A more detailed analysis of Alk knockout mice revealed an increase in the number of hippocampal progenitor cells in the brain and an enhanced performance in a test of novel object recognition. Another strain of such mice manifested increased ethanol consumption compared with heterozygous controls (6). These observations thus suggest that ALK may contribute to behavioral control in adult mice.

The structure of ALK, together with its expression pattern, suggests that it may function as a cell surface receptor for specific ligands that may regulate the proliferation or differentiation of neural cells. In the fruit fly, jelly belly (jeb) binds and activates an ortholog of ALK (7), but a mammalian ortholog of jeb has not been identified. The growth factors pleiotrophin and midkine have been proposed as candidate ligands for mammalian ALK (8, 9), but this notion remains controversial (10), leaving the biologic relevance of pleiotrophin and midkine unclear.

ALK first emerged in the field of oncology in 1994 as a result of the identification of its fusion to nucleophosmin (NPM1) in anaplastic large cell lymphoma (ALCL; Figs. 1 and 2; refs. 11, 12). The NPM1 portion mediates constitutive dimerization and consequent activation of the NPM1-ALK fusion protein (13). In addition to NPM1-ALK, various other ALK fusion proteins have now been identified in ALCL, including TFG-ALK, ATIC-ALK, and CLTC-ALK. The next disease linked to ALK was inflammatory myofibroblastic tumor (IMT), with increased expression of ALK and a rearranged ALK locus (chromosome band 2p23) being detected in a subset of IMT cases. A more detailed analysis identified TPM3-ALK and TPM4-ALK as fusion genes in IMT (Figs. 1 and 2; ref. 14). Further screening of IMT specimens revealed additional ALK fusions including RANBP2-ALK and CARS-ALK. Interestingly, some of these fusion genes, such as TPM3/4-ALK and CLTC-ALK, have been identified in both ALCLs and IMTs. The presence of TPM3/4-ALK was also reported in esophageal squamous cell carcinoma (15) and renal cancer (16).

In 2007, interest in ALK and the therapeutic potential of its specific inhibitors was boosted in response to the discovery of another fusion gene, EML4-ALK, this time in non–small cell lung carcinoma (NSCLC; ref. 17). Such interest was further increased the next year by the identification of activating point mutations in ALK in cases of neuroblastoma (18–21). In this review, I focus on activating genetic changes of ALK relevant to human cancer. Other aspects of ALK alterations (such as overexpression) in cancer have been elegantly reviewed elsewhere (22, 23).

Chronic myeloid leukemia (CML) is characterized by the presence of a fusion-type PTK, BCR-ABL1, that is generated as a result of a balanced chromosomal translocation, t(9;22). The remarkable therapeutic efficacy of the ABL1 inhibitor imatinib in individuals with this condition (24) suggested that targeting of the essential growth drivers in different types of cancer is a promising treatment strategy. To identify such growth drivers in clinical specimens, we developed a sensitive functional screening system based on retroviral cDNA expression libraries. The application of this approach to a lung adenocarcinoma specimen resulted in the discovery of EML4-ALK as a fusion-type oncogene (Figs. 1 and 2; ref. 17).

EML4 and ALK loci both map to the short arm of human chromosome 2 in opposite orientations and are separated by a distance of approximately 12 Mbp. A small inversion, inv(2)(p21p23), affecting both loci gives rise to the fusion gene. EML4-ALK was the first recurrent fusion-type oncogene in NSCLC, and, together with ETS fusions in prostate cancer (25), its existence argues against the previous notion that oncogenesis mediated by chromosome rearrangement is relatively specific to hematologic malignancies and sarcomas (rather than epithelial tumors).

EML4-ALK encodes a protein consisting of an amino-terminal portion of EML4 fused to the intracellular portion of ALK. EML4-ALK undergoes constitutive dimerization mediated through the coiled-coil domain of the EML4 portion and thereby acquires transforming ability in a manner dependent on the associated upregulation of PTK activity. Indeed, several cell lines positive for EML4-ALK were found to be dependent on ALK catalytic activity for their proliferation (26, 27). Furthermore, transgenic mice that express EML4-ALK in lung type II alveolar cells manifest hundreds of adenocarcinoma nodules in both lungs soon after birth, and treatment of the animals with an ALK inhibitor results in the rapid clearance of these nodules (28). Such successful treatment of EML4-ALK transgenic mice with an ALK inhibitor was subsequently confirmed independently (29). These data suggested that EML4-ALK is the essential growth driver for NSCLC positive for this fusion kinase, and that targeting of ALK activity may therefore prove therapeutically effective in the clinic.

EML4-ALK is present in 3% to 6% of NSCLC cases (30–32). Of note, the clinical characteristics of patients with NSCLCs positive for EML-ALK are similar to those of such individuals who harbor activating mutations in the EGF receptor gene (EGFR): Both groups of patients thus tend to manifest an adenocarcinoma histology and to be non- or light smokers. However, the presence of EML4-ALK and that of EGFR or KRAS mutations are mutually exclusive, albeit with rare exceptions (33). Furthermore, whereas NSCLC with EGFR mutations is more prevalent in Asian populations than in Caucasians, no such ethnic differences have been reported for NSCLCs with EML4-ALK. With regard to pathologic characteristics, EML4-ALK–positive NSCLC has been found to contain signet-ring cells or to manifest a mucinous cribriform pattern (34–36), but other pathologic subtypes of NSCLCs also may harbor EML4-ALK. EML4-ALK may be infrequently present in other types of human cancers (16, 37).

The original EML4-ALK we identified resulted from ligation of intron 13 of EML4 to intron 19 of ALK (17), but many other variants of the fusion gene have since been described. The vast majority of ALK translocations, including EML4-ALK, NPM1-ALK, and TPM3-ALK, involve intron 19 of ALK. Theoretically, ligation of EML4 introns 1, 2, 6, 13, 18, 20, or 21 to intron 19 of ALK should generate an in-frame fusion at the mRNA level. Furthermore, given that exon 2 of EML4 encodes the coiled-coil domain essential for EML4-ALK activation, all such fusions with the exception of that of EML4 intron 1 would be expected to generate an oncogenic kinase. Large-scale screening of clinical specimens has revealed that the original EML4-ALK variant (referred to as E13;A20 or variant 1) together with a variant in which intron 6 of EML4 is fused to intron 19 of ALK (referred to as E6;A20 or variant 3; ref. 26) account for more than 90% of all EML4-ALK–positive cases of NSCLCs in Japan (M. Soda, personal communication), but many rare variants have also been identified (17, 26, 27, 30, 31, 34, 37, 38).

Importantly, EML4 exons theoretically unable to undergo in-frame fusion to exon 20 of ALK may be involved in the generation of oncogenic EML4-ALK proteins. For instance, an EML4-ALK cDNA in which exon 14 of EML4 is fused in-frame to the 13th nucleotide of ALK exon 20 has been identified (30). In addition, another variant in which the nucleotide located 19 bp upstream of the 3′ end of EML4 exon 15 is fused to that located 20 bp downstream of the 5′ end of ALK exon 20 was described (27). It is therefore important that clinics adopt diagnostic techniques, such as those based on multiplex reverse transcription PCR (RT-PCR), that are able to capture all these variants of EML4-ALK reliably.

Currently, no single technique may be suitable for the analysis of all specimen types. Immunohistochemistry and FISH, for example, are applicable to formalin-fixed, paraffin-embedded (FFPE) tissue specimens (32, 34, 35) but may not be suitable for the analysis of sputum, pleural effusion, bronchial lavage fluid, or frozen tissue. Conversely, the latter specimen types are readily examined by RT-PCR (30), whereas the former ones may not be suitable for this technique. I thus propose that diagnostic tools for the detection of EML4-ALK should be selected on the basis of the available specimen types. FISH and immunohistochemistry should be applied to FFPE tissue samples, whereas multiplex RT-PCR is appropriate for the other specimen types.

The first ALK inhibitor to enter clinical trials was crizotinib, which is also known as PF-02341066 and is actually a dual inhibitor for both ALK and MET kinases (Table 1; ref. 39). At the time of the discovery of EML4-ALK, crizotinib had already entered a phase I trial that mainly targeted digestive tract cancers positive for MET amplification. After the report of EML4-ALK, the phase I trial with crizotinib was expanded to include tumors positive for ALK rearrangement, and the drug soon proved as therapeutically efficacious for such tumors.

The response rate for crizotinib in patients with ALK–rearranged NSCLCs in the trial was shown to be 57%, with a disease control rate of up to 90% (32). Furthermore, the median overall survival for crizotinib treatment was not determinable within the follow-up period of 18 months (40). Comparison of crizotinib treatment with a historical control was further striking. Whereas the median overall survival was not achieved for patients who received crizotinib as a second- or third-line treatment, it was only 6 months for the ALK-rearranged control patients who received conventional chemotherapy in the second- or third-line setting.

While crizotinib can inhibit MET kinase activity and MET may become amplified in NSCLCs (41), there was no MET amplification present in the above EML4-ALK–positive cohort (32). Similarly, while ROS1 tyrosine kinase is sensitive to crizotinib (42), a recent large-scale screening of gene fusions among NSCLC (n = 1,529) revealed a complete mutual exclusiveness between ALK and ROS1 fusions (43). Furthermore, the mechanism for the insensitivity of ALK fusion–positive tumors to crizotinib has been mostly secondary mutations within the kinase domain of EML4-ALK (see below). These data evidence that crizotinib exerts its marked therapeutic efficacy in NSCLCs through specific suppression of EML4-ALK activity.

These observations clearly supported the clinical use of ALK inhibitors for the treatment of EML4-ALK–positive NSCLC. Indeed, on August 26, 2011, the U.S. Food and Drug Administration approved crizotinib as a treatment for ALK-rearranged NSCLCs. From the time of our first report of the identification of EML4-ALK in 2007, it took only 4 years for the first ALK inhibitor to be approved for use in the clinic, which is a record for cancer drug development (44).

Following the discovery of EML4-ALK, efforts have expanded to detect novel ALK fusions in epithelial tumors. Development of a sensitive immunohistochemical technique (intercalated antibody–enhanced polymer method, or iAEP) to stain ALK proteins in FFPE specimens led to the identification of several NSCLC samples that were positive with this approach but negative for EML4-ALK with multiplex RT-PCR (34). Further investigation of these specimens revealed the presence of another fusion of ALK, KIF5B-ALK. Similar to EML4-ALK, KIF5B contains dimerization motifs that play an essential role in the oncogenic activity of KIF5B-ALK and there are now known to be several variants of KIF5B-ALK (45). Togashi and colleagues reported still another ALK fusion, KLC1-ALK, in NSCLCs (46).

Yet another novel ALK fusion, VCL-ALK, was recently identified in a tumor of unclassified renal cell carcinoma with renal medullary carcinoma (RMC) characteristics that developed in a 16-year-old boy with sickle cell trait (47). VCL-ALK was also detected in RMCs isolated from a 6-year-old boy (48). RMC mostly affects young individuals and has a poor outcome, but the discovery of VCL-ALK has raised the possibility of effective treatment with an ALK inhibitor for patients who harbor this fusion gene. Furthermore, screening of tissue microarrays of renal cell carcinoma with the iAEP technique led to the detection of single cases each positive for TPM3-ALK or EML4-ALK (E2;A20 variant; ref. 16).

Attempts by several groups to identify genes underlying neuroblastoma through different approaches (mapping of single-nucleotide polymorphisms for familial neuroblastoma and chromosome copy number analysis for sporadic neuroblastoma) resulted in the almost simultaneous identification of activating mutations within ALK (Figs. 2 and 3; refs. 18–21). About 10% of sporadic neuroblastoma cases harbor somatic nonsynonymous mutations within ALK, including K1062M, F1174L/C/I, F1245C/V/L, and R1275Q amino acid substitutions. On the other hand, a distinct but partially overlapping set of ALK mutations (T1087I, G1128A, R1275Q, and others) has been identified in familial neuroblastoma.

Importantly, these mutations do not confer equal transforming ability. The F1174L mutant, for instance, efficiently phosphorylates the signaling molecules STAT3 and AKT, but not extracellular signal–regulated kinase (ERK)1/2, whereas the R1275Q mutant efficiently phosphorylates ERK1/2 but not STAT3 (18). Knockdown experiments revealed that the growth of neuroblastoma cell lines was dependent to a greater extent on the F1174L mutant than on R1275Q (18, 20). The mutant ALK proteins thus contribute substantially to the transformation process in neuroblastoma, but the extent to which they do so varies among the mutation types.

These mutations also differentially affect the sensitivity of neuroblastoma to ALK inhibitors (49, 50), which may not be surprising given that point mutations within the kinase domain of ALK affect its 3-dimensional structure (including that of the inhibitor binding cleft) and thereby influence inhibitor binding. The F1174L mutant confers marked resistance to crizotinib in a cell-based assay (49, 50), suggesting that this amino acid substitution not only increases the enzymatic activity of ALK by changing the structure of the kinase domain but by doing so also affects the binding of ALK inhibitors.

The I1250T mutation of ALK was recently identified in an individual with neuroblastoma and was shown to abolish the activity of the enzyme (51). Each ALK mutation identified in neuroblastoma therefore needs to be assessed for how (or if) it contributes to carcinogenesis.

Nucleotide sequencing of ALK exons encoding the kinase domain in cell lines and fresh specimens of thyroid cancer (52) revealed novel missense mutations specifically in anaplastic thyroid cancer (ATC) samples (2 positive samples of 11). Each of the 2 identified amino acid substitutions (L1198F and G1201E) increased the enzymatic activity of ALK and induced the formation of transformed foci when introduced into mouse 3T3 cells, clearly indicating the activating nature of these mutations. It is thus likely that a subset of ATC and neuroblastoma cases share the same transforming gene, but it remains to be determined whether the ALK mutation profiles differ between the 2 disorders.

In addition to nonsynonymous mutations, ALK infrequently becomes amplified in neuroblastoma (21, 53). While its clinical relevance is yet to be clarified, ALK amplification frequently co-occurs with amplification of MYCN amplification, a known growth driver for this disorder, suggesting that ALK also contributes to carcinogenesis.

Recently, van Gaal and colleagues (54) have discovered frequent copy number gain of ALK in rhabdomyosarcoma accompanied with an increased level of ALK protein. In some cases, ALK copy number was even above 10. Interestingly, contrary to neuroblastoma, rhabdomyosarcoma with ALK amplification do not carry MYCN amplification. Such ALK anomaly is likely to be connected to carcinogenesis because ALK gain was associated with poor survival and the occurrence of metastases. In addition, van Gaal and colleagues also identified one case with ALK (D1225N) and 7 with ALK frameshift mutations of 43 rhabdomyosarcoma specimens, although clinical relevance of such mutations is yet to be examined.

Some EML4-ALK–positive NSCLC tumors are insensitive to initial treatment with ALK inhibitors, whereas others acquire resistance to these drugs after initial successful treatment. Information on the molecular mechanisms underlying such resistance remains limited.

The first insight into such resistance mechanisms was provided by analysis of an individual with EML4-ALK–positive NSCLC who initially showed a partial response to crizotinib treatment but underwent a rapid relapse after 5 months. EML4-ALK cDNA was amplified by PCR from tumor specimens obtained before the onset of treatment and after relapse, and it was then subjected to deep sequencing with a next-generation sequencer. Comparison of the 2 data sets resulted in the identification of 2 nonsynonymous mutations present only in the latter specimen. Interestingly, these 2 amino acid substitutions (C1156Y and L1196M) were found to independently confer resistance to crizotinib and other ALK inhibitors (55).

Neither C1156Y nor L1196M affects the enzymatic activity of EML4-ALK. NSCLC cells harboring either mutant might therefore be positively selected in vivo only in the presence of ALK inhibitors. Furthermore, Leu¹¹⁹⁶ is the “gatekeeper” site buried deeply within the ATP-binding pocket of ALK, with the corresponding sites of EGFR (Thr⁷⁹⁰) and BCR-ABL1 (Thr³¹⁵) being the most frequently mutated residues associated with gefitinib and imatinib resistance, respectively (56).

Examination of a similar EML4-ALK–positive individual who initially responded to crizotinib but later acquired resistance revealed the presence in post-relapse EML4-ALK cDNA of a mutation resulting in an L1152R substitution in the kinase domain (57). Another mutation conferring drug resistance (F1174L) was identified in a case of RANBP2-ALK–positive IMTs (58). Of note, F1174L is also one of the frequent amino acid substitutions identified in neuroblastoma (Fig. 3). Further extensive screening of crizotinib-resistant mutations among 18 relapsed patients revealed 4 secondary mutations (1151Tins, L1196M, G1202R, and S1206Y) within the kinase domain of EML4-ALK, all of which confer resistance to ALK inhibitors (59). Such investigation further identified a high-level amplification of EML4-ALK in one patient, suggesting another mechanism (gene amplification) of drug insensitivity.

Another screening of EML4-ALK–positive tumors at a relapsed phase led to the identification of L1196M (in 2 of 14 resistant cases) and G1269A (in 2 cases) mutations accounting for the acquired drug resistance, and of gene amplification in 2 cases (60). Interestingly, in this cohort, some tumors lost EML4-ALK oncogene at the relapsed phase, but instead acquired activating EGFR or KRAS mutations. Whether such oncogenes other than EML4-ALK were present in a minor population of the original tumor or secondarily acquired during the crizotinib treatment remains elusive.

As of March 2012, 5 distinct ALK inhibitors are in clinical trials worldwide (Table 1), with some of these drugs showing inhibitory activity in vitro even with the gatekeeper mutant (61, 62). It will thus be of interest to determine whether treatment with such inhibitors results in the development of acquired resistance or not.

Since the initial discovery of NPM1-ALK in 1994, our knowledge of the role of ALK in human cancer has increased markedly. We now recognize the contribution of ALK to an unexpectedly wide range of tumors, with ALK translocations underlying lymphoma, lung carcinoma, kidney cancer, and soft tissue tumors and ALK mutations being responsible for neuronal and thyroid cancer and ALK amplification frequently occurring in rhabdomyosarcoma.

Given that all ALK fusion kinases retain an intact ATP-binding pocket (where most ALK inhibitors bind), it is likely that ALK inhibitors will be effective against any tumor type that harbors such an ALK fusion (63, 64). From the standpoint of pharmaceutical companies, the fact that a single compound can serve as a magic pill for many different types of cancer in different organs is sufficiently compelling to warrant the development of such drugs even if the individual cancer types do not have a high incidence.

For neuroblastoma and ATC, however, the efficacy of an ALK inhibitor may be substantially influenced by the type or position of the missense mutation. However, given that ALK is a transmembrane protein, the mutant proteins may be effectively targeted by antibodies (65), as is the case for other transmembrane proteins targeted by the antibodies rituximab and trastuzumab.

Genetic information was first integrated into the classification of hematologic malignancies. For instance, whereas acute myeloid leukemia (AML) had previously been divided into subgroups with distinct differentiation profiles (assessed by pathologic analyses), AML harboring RUNX1-RUNX1T1, PML-RARA, or MLL rearrangements is now defined as a distinct entity according to the current World Health Organization classification (66). Likewise, all-trans retinoic acid is highly effective only in the treatment of retinoic acid receptor alpha (encoded by RARA) fusion–positive AML, acute promyelocytic leukemia (APL). Similarly, ABL1 inhibitors are only used against BCR-ABL1–positive CML/acute lymphoblastic leukemia. It is likely that tumor cells of APL or CML are deeply addicted to the activity of PML-RARA or BCR-ABL1, respectively. Therefore, single reagents targeting individual oncoproteins have a profound therapeutic effect.

In the treatment of epithelial tumors, single EGFR inhibitors provide a high response rate to NSCLCs harboring activating EGFR mutations as well. Importantly, treatment with such EGFR inhibitors worsens the prognosis of NSCLC without EGFR mutations (67). Targeted drugs are, thus, likely to be highly effective only against tumors in which corresponding targets carry activating mutations and become “essential growth drivers” for the cancer. The definition of essential growth drivers for given oncogenes is difficult because treatment efficacy for cell lines in vitro does not always recapitulate that in clinics. For instance, while inhibitors against RAS or PIK3CA are able to suppress the growth of cancer cell lines, these compounds have failed to provide such efficacy in humans.

Therefore, significant response in patients by a monotherapy with a targeted drug should be a faithful indicator for the corresponding target to be the essential growth driver. Unfortunately, however, such clinically proven drivers are still few, but EML4-ALK is a newcomer to this growing list. Importantly, other ALK fusions, NPM1-ALK and RANBP2-ALK, are likely the essential growth drivers as well, because the crizotinib monotherapy has a profound therapeutic effect (63, 64). I propose that these tumors be collectively referred to as “ALKoma,” in which abnormal ALK plays an indispensable component in the carcinogenesis.

Interestingly, the term “ALKoma” encompasses multiple organs. NSCLCs, IMTs, and ALCLs were once regarded as completely distinct clinical entities affecting different organs. However, it is now known that a fraction of each of these cancer types shares activated ALK as the essential growth driver and such tumors can be targeted for treatment with ALK inhibitors (32, 63, 64). Because of the similar protein structure to that of EML4-ALK/NPM1-ALK/RANBP2-ALK, VCL-ALK in RMCs should be a good candidate for the next member of ALKoma. Data from ongoing clinical trials with crizotinib for neuroblastoma will also tell us whether activated ALK with nonsynonymous mutations is also a strong driver for this disorder. The term “ALKoma” has the advantage that it indicates the preferred treatment for the tumor. It is thus a good example of genetic information–based “beyond organ” cancer classification, many other examples of which may emerge in the future.

H. Mano has commercial research grant from Astellas Pharma and Illumina Inc.; Ownership Interest (including patents); and serves as a scientific advisor for Pfizer Inc., Astellas Pharma, Chugai Pharmaceutical, and Daiichi Sankyo Co., Ltd., and is an CEO of CureGene Co., Ltd.

Conception and design: H. Mano

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Mano

Writing, review, and/or revision of the manuscript: H. Mano

Study supervision: H. Mano

The author apologizes to all the authors whose work could not be included in the manuscript owing to space constraints and also thanks the members of his laboratory as well as K. Takeuchi for their dedication and support.

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 as well as by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

Received January 6, 2012; revised April 20, 2012; accepted April 20, 2012; published OnlineFirst May 21, 2012.