Research ArticleNeuroblastoma

Targeted Expression of Mutated ALK Induces Neuroblastoma in Transgenic Mice

See allHide authors and affiliations

Science Translational Medicine  04 Jul 2012:
Vol. 4, Issue 141, pp. 141ra91
DOI: 10.1126/scitranslmed.3003967

Abstract

Activating anaplastic lymphoma kinase (ALK) mutations were recently detected in most familial and 10% of sporadic neuroblastomas. However, the role of mutated ALK in tumorigenesis remains elusive. We demonstrate that targeted expression of the most frequent and aggressive variant, ALKF1174L, is tumorigenic in mice. Tumors resembled human neuroblastomas in morphology, metastasis pattern, gene expression, and the presence of neurosecretory vesicles as well as synaptic structures. This ALK-driven neuroblastoma mouse model precisely recapitulated the genetic spectrum of the disease. Chromosomal aberrations were syntenic to those in human neuroblastoma, including 17q gain and MYCN oncogene amplification. Targeted ALKF1174L and MYCN coexpression revealed a strong synergism in inducing neuroblastoma with minimal chromosomal aberrations, suggesting that fewer secondary hits are required for tumor induction if both oncoproteins are targeted. Treatment of ALKF1174L transgenic mice with the ALK inhibitor TAE-684 induced complete tumor regression, indicating that tumor cells were addicted to ALKF1174L activity. We conclude that an activating mutation within the ALK kinase domain is sufficient to induce neuroblastoma development, and ALK inhibitors show promise for treating human neuroblastomas harboring ALK mutations.

Introduction

Neuroblastoma, which is the most common extracranial childhood tumor, originates from neural crest progenitor cells (1). Most neuroblastomas develop sporadically and have a broad range of severity that correlates with the presence of specific genetic alterations (1). Structural genomic aberrations, including gain of chromosome 17, are correlated with adverse outcome. Amplification of the MYCN oncogene occurs in about 20% of all neuroblastomas and defines the most fatal subset. Oncogenic activation of anaplastic lymphoma kinase (ALK), a gene normally expressed in the developing nervous system, by translocation and formation of fusion genes was first described in T cell lymphomas but is also a common feature of several other malignancies, including non–small cell lung cancer (25). Therapy with the ALK inhibitor crizotinib is effective against these tumors (6). Recently, point mutations activating the ALK tyrosine kinase domain were detected in most familial and 10% of sporadic neuroblastomas (710).

Of the identified ALK-activating mutations in neuroblastoma, ALKF1174L is the most frequent and most aggressive mutation in terms of transforming capacity (711). ALKF1174L has only been detected in sporadic neuroblastomas and has been reported as a treatment-related secondary mutation during crizotinib treatment (12). However, the role of mutated ALK in tumorigenesis remains elusive. We generated transgenic mice with a conditional allele of ALK harboring the F1174L mutation to assess whether expression of ALK, constitutively activated by a mutation in its kinase domain, is capable of initiating oncogenesis in vivo.

Results

ALKF1174L expression, which enhances neural crest stem cell proliferation, was used to create transgenic mice

To analyze the transforming capacity of wild-type ALK and the two most common mutations, ALKR1275Q or ALKF1174L, we first transfected murine neural crest stem cells with either ALK form (13). Although wild-type ALK and ALKR1275Q induced low-level proliferation, ALKF1174L increased cell viability of transfected cells more than twofold compared to wild-type ALK (fig. S1). ALKF1174L was selected for our transgenic mouse model on the basis of this ability to strongly increase viability and induce proliferation. A conditional expression allele was introduced to generate mice with tissue-specific ALKF1174L expression. No transgene expression occurred in the presence of a stop cassette, whose deletion with Cre recombinase resulted in expression of both the ALKF1174L transgene and a luciferase reporter gene (Fig. 1A). Tissue-specific Cre expression directed transgene expression to the tissue of interest. Of 63 founder mice obtained upon pronucleus injection of the linearized plasmid, 10 transgenic founders were detected by polymerase chain reaction (PCR) genotyping. Treatment of tail fibroblasts from all founder mice with recombinant Cre protein in primary culture revealed strong luciferase activity in only one transgenic founder (fig. S2). This transgenic line was selected for further experiments and backcrossed with C57Bl/6 mice for one generation. PCR genotyping of the progeny confirmed transgene inheritance (fig. S3). Strong luciferase activity was detected in tail fibroblasts derived from progeny after in vitro Cre protein treatment, indicating stable transgene expression.

Fig. 1

ALKF1174L transgenic mice develop neural crest–derived tumors. (A) Graphical representation of the vector construct. Numbers depict primer locations for genotyping (1 and 2) and PCR validation of green fluorescent protein (GFP)–pA cassette removal (3 and 4). (B) Kaplan-Meier analysis of transgenic mice with the endpoint defined as detection of palpable tumors. Mice were double-transgenic for ALKF1174L and DBHiCre or TH-IRES-Cre. (C) Bioluminescent imaging of two representative mice (luciferase activity: low, blue; high, red). (D) Autopsies of mice carrying palpable tumors. (I) Primary tumor arising from the left adrenal displacing the left kidney caudally (tu, tumor; ki, kidney). (II) Liver from a mouse with a large retroperitoneal tumor, showing multiple metastatic nodules. (III) Thoracic cavity of a mouse with a large retroperitoneal tumor and metastatic lesion or second primary tumor in the left upper thorax (tu, second primary tumor; he, heart; lu, lung). (IV) Primary tumor arising from cranial paravertebral ganglion.

Direction of ALKF1174L expression to the neural crest induces tumor formation

To direct ALKF1174L expression to the neural crest and its progenitors, we crossbred progeny of ALKF1174L transgenic mice with DBHiCre (14) or TH-IRES-Cre (15) mice. ALKF1174L was mainly expressed in neural crest derivatives and the few other tissues expressing Dbh or Th, including the central nervous system structures locus coeruleus and substantia nigra (fig. S4). No morphological changes were observed in these tissues in macroscopic and histological examinations (fig. S4). Of 12 double-transgenic mice, 5 developed palpable tumors between 130 and 351 days of age (Fig. 1B and table S1). No tumors developed in control mice harboring the ALK transgene without the Cre recombinase gene. This observation supports the hypothesis that the transgene, and not a tumor suppressor gene disrupted during transgene integration, drives tumorigenesis in this model. Mice transgenic for the ALKF1174L transgene and DBHiCre or TH-IRES-Cre that did not develop tumors expressed ALK in their adrenals to varying degrees, and some adrenals appeared hypertrophic upon autopsy (fig. S5). Flox-out was incomplete in adrenals from these mice compared with tumor tissue (fig. S5 and see below). In the five mice developing tumors, the tumor formed in the neck of mouse 4 and the abdomen of the other four mice (mouse 1 to 3 and 5, see also table S1). In vivo bioluminescence imaging (BLI) detected strong luciferase activity in all tumors from ALKF1174L transgenic mice (Fig. 1C), demonstrating transgenic construct expression. Macroscopically visible tumors were present in mouse 1 to 4 at autopsy (Fig. 1D, mouse 5 was treated with ALK inhibitor). The neck tumor arose from paravertebral cervical ganglia, matching the location of 2% of human neuroblastomas(1). The retroperitoneal tumors in mouse 1 to 3 originated from the adrenal gland, reflecting the most frequent localization of human neuroblastomas (>50%). Extensive liver metastases, present in about 50% of neuroblastoma patients at diagnosis, were also present in mouse 1 and 3 (Figs. 1D, II, and 2A, IX to XI). Mouse 1 had an additional mediastinal lesion (Fig. 1D, III), a clinical location present in about 10% of patients. No bone marrow metastases were detected (fig. S6), but the pattern of disease in the ALKF1174L transgenic mice largely recapitulates the clinical spectrum of human neuroblastoma.

Fig. 2

Analysis of ALKF1174L-induced tumors confirmed molecular and histological characteristics of neuroblastoma. Representative figures from mouse 1 to 3 and 5 are shown (RNA and protein of sufficient quality were unavailable for mouse 4). (A) Histology. Scale bar, 100 μm (if not otherwise indicated). (I) Hematoxylin and eosin (H&E) staining shows predominantly small round blue cells. (II) Electron micrographs showing neuronal structures, including neurosecretory vesicles (red arrows) and structures resembling synaptic membrane formation (blue arrows). (III) Immunohistochemistry confirmed Th expression in the neuroblastic tumor cell cytoplasm (brown), but not peripheral nerve (N, internal negative control). (IV) Negative control for Th specificity (secondary antibody only). (V) Synaptophysin (Syn) expression. (VI) Negative control for Syn specificity (secondary antibody only). (VII) NSE expression. (VIII) Negative control for NSE specificity (secondary antibody only). (IX) H&E staining of a liver showing multiple metastatic nodules. (X and XI) Liver metastases stain positively for Th. Tumor cells invading the liver sinusoidal veins (black arrows). (XII) Tumors are diffusely infiltrated with CD45-positive immune cells. (B) RT-qPCR confirmed Dbh, Th, and Phox2b expression in tumors but not control tissues. Expression was normalized to normal adrenal glands. (C) RT-qPCR for human ALKF1174L confirmed its exclusive expression in tumor tissue. (D) Western blotting confirmed both ALK expression and activation (phosphorylation) in tumor tissue. The 200- to 220-kD form of ALK is a doublet, with the smaller band being the intracellular form. ALK is also present as a 140-kD form. Western blots from mouse 1 to 3 tumors were separated to better display different intensities to fit the dynamic range. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (E) Predominantly cytoplasmic localization of ALK (green).

ALKF1174L-driven murine tumors resemble human neuroblastomas

Tumors consisted almost entirely of poorly differentiated cells with large nuclei and sparse neuropil, similar to human neuroblastomas (Fig. 2A, I). Synaptic membranes and neurosecretory vesicles were identifiable in tumor cell electron micrographs (Fig. 2A, II), ultrastructures present only in cells of neuronal lineage and confirming the tumors as neuroblastomas. All tumors expressed mRNA and protein for the neuroblastoma markers Th, Dbh, neuron-specific enolase (NSE), synaptophysin, Ncam1, and Phox2b (Fig. 2, A and B, and fig. S7). No tumor expressed CD45, a common leukocyte marker and marker for lymphoma, the tumor most histologically similar to neuroblastoma. A sparse infiltration with CD45-positive cells was observed, indicative of an immune response and regularly observed in human neuroblastomas (Fig. 2A, XII). The macroscopic appearance, sites of primary tumors and metastases, histology, electron microscopy, and marker gene/protein expression unambiguously confirmed tumors in our mouse model as neuroblastomas.

The ALKF1174L transgene is highly expressed and activated in the murine neuroblastomas

Transgene activation by Cre recombinase was restricted to tumor tissue and not detected in normal tissues not expressing Cre recombinase (fig. S8). A PCR by-product indicative of the nonactivated transgene suggested the presence of normal cells in tumors. Cells of the tumor stroma, including infiltrating leukocytes, are the likely source of this by-product. Real-time reverse transcription–quantitative PCR (RT-qPCR) specific for human ALKF1174L and Western blotting confirmed exclusive expression of the transgene in tumors (Fig. 2, C and D). Although ALK mRNA expression was comparable among tumors, ALK protein expression was higher in tumor 1, which also harbored a MYCN amplification. Phosphospecific ALK immunoblotting (Fig. 2D) demonstrated that ALK expression was concomitant with activation in the tumors. The effect of activated ALKF1174L expression on downstream gene expression was analyzed using transcriptional expression profiles obtained from normal murine adrenal medulla, ALK-driven neuroblastomas, and MYCN-driven neuroblastomas from TH-NMYC transgenic mice (16). Unsupervised hierarchical clustering using the 1% of genes with the highest SDs in expression across all samples revealed distinct clustering of normal adrenal medulla and ALK- and MYCN-driven tumors (fig. S9A). An ALK pathway activity score was calculated for all samples using the expression profiles. This score was significantly higher in tumors exclusively driven by the ALKF1174L transgene than in the normal murine adrenal medulla or MYCN-driven tumors, demonstrating that the ALK pathway is strongly activated in tumors arising in transgenic mice with targeted ALKF1174L expression (fig. S9B). Thus, expression of oncogenic ALKF1174L was restricted to tumor tissues and strongly activated the ALK pathway in the ALKF1174L transgenic mouse model.

Murine neuroblastomas recapitulate genomic aberrations of human neuroblastomas

Most human neuroblastomas harbor a complex and specific pattern of chromosomal aberrations (17). The relatively long and variable times to tumor development and incomplete tumor penetrance in our mouse model indicate a requirement for secondary genetic events for ALK-driven murine neuroblastomagenesis. Whole-genome analysis of these murine neuroblastomas detected segmental aberrations in three of four tumors (mouse 1, 3, and 4), reminiscent of aggressive human neuroblastomas (17) (Fig. 3A and table S2). The tumor from mouse 2 had numerical chromosomal aberrations, also common in human neuroblastomas and associated with a less aggressive clinical course. Neuroblastomas from mouse 1 to 3 displayed gain of the entire chromosome 3, and from mouse 4 almost the entire chromosome 3 (Fig. 3A). This chromosome is partially syntenic to human chromosome 1q, a region often gained in human neuroblastomas (17). A partial gain of chromosome 11q was detected in the tumor from mouse 4 (Fig. 3B), corresponding to a region on chromosome 17 gained in 50% of all human neuroblastomas. This aberration has been correlated with adverse outcome, although its biological relevance remains unclear. Tumors from mouse 2 and 3 had losses of murine chromosomes 4 and 5, harboring regions syntenic to human chromosome 1p36, a common region of loss of heterozygosity (LOH) in human neuroblastoma that correlates with adverse outcome and is suspected to harbor one or more tumor suppressor genes. Strikingly, the tumor from mouse 1 harbored a high-level MYCN amplification (Fig. 3, C to E), which occurs in 20% of human neuroblastomas, defines the most aggressive tumor subset, and was the first oncogene amplification used for therapy stratification of patients (1). The spectrum of chromosomal aberrations in ALK-driven murine tumors recapitulates the diverse spectrum of aberrations observed in human neuroblastomas.

Fig. 3

Tumors harbor chromosomal aberrations syntenic to characteristic aberrations of human neuroblastoma. (A) Mouse karyotype overview of all genomic imbalances detected in the four murine neuroblastomas (gains, green bars; losses, red bars). (B) Tumor/control ratio plot for mouse chromosome 11 in tumor 4 shows partial chromosome 11q gain corresponding to gain of almost entire human chromosome 17 (0 to 16.9 Mb and 18.6 to 78.7 Mb). (C) Partial ratio plot for the mouse chromosome 12 region encompassing the Mycn amplicon in tumor 1. (D) Mycn copy number assessment (qPCR) in all four tumors (Tu) and matched controls (ctrl) (C57Bl/6 mouse DNA = normal calibrator sample). Lines below (B) to (E) represent murine cytobands and megabase position (top), murine coding genes (middle), and color-coded human syntenic regions (bottom). (E) Fluorescence in situ hybridization on tumor from mouse 1 independently confirmed Mycn amplification. Normal cells contain two Mycn copies (yellow signals). Tumor cells contain many copies of Mycn (yellow signals) as double minutes. (F) Kaplan-Meier analysis showing that ALK and MYCN synergistically accelerated neuroblastoma formation (red curve). 129B6F1: F1 background from 129x1/SvJ × C57Bl/6 mouse cross. Endpoint defined as detection of palpable tumors. (G) Number of chromosomal aberrations detected by aCGH was significantly reduced in tumors from TH-NMYC;ALKF1174L;DBHiCre triple-transgenic mice (mouse 8 to 10). No aberrations were detected in the tumor from mouse 9. (H) The number of structural or numerical chromosomal aberrations in human tumors was plotted against the presence of MYCN amplification or ALK mutation/amplification or both.

ALKF1174L and MYCN synergize in inducing neuroblastoma

The ALKF1174L mutation has been associated with MYCN amplification in human neuroblastomas (11). Concordantly, we observed spontaneous MYCN amplification in the most widely metastasized murine tumor (mouse 1). Perhaps ALK and MYCN not only are associated but also functionally synergize to induce neuroblastoma. To address this question, we crossbred ALKF1174L;DBHiCre mice with the TH-NMYC transgenic mouse model, which expresses MYCN in the neural crest driven by a rat Th promoter in the 129x1/SvJ strain (16). In the 129x1/SvJ strain background, about 50% of mice (28 of 60) expressing the TH-NMYC transgene alone developed neuroblastomas at 100 days of age (Fig. 3F, gray curve). Tumor incidence is lower in TH-NMYC mice, with longer times to tumor formation in the C57Bl/6 strain background (16). Crossbreeding TH-NMYC mice (129x1/SvJ strain) with ALKF1174L;DBHiCre mice (C57Bl/6 J strain) generated triple-transgenic mice (TH-NMYC;ALKF1174L;DBHiCre) and littermate controls in the same genetically defined F1 background. Compared to the original 129x1/SvJ strain background, tumor incidence and time to tumor formation in mice expressing the TH-NMYC transgene alone decreased markedly in the F1 strain background, of which none had developed tumors by 100 days (Fig. 3F, blue curve) and only one developed a tumor during long-term follow-up. Expressing both the ALKF1174L and the TH-NMYC transgenes in the neural crest markedly accelerated neuroblastoma formation in comparison to MYCN overexpression alone, with all triple-transgenic mice developing tumors within 48 days (Fig. 3F, red curve, and table S1). Tumor incidence was significantly higher and time to tumor formation significantly shorter than in mice expressing only the TH-NMYC transgene in the F1 background or TH-NMYC mice in the 129x1/SvJ strain background, which is more susceptible than the F1 background to neuroblastoma development. The ALK activity score for tumors from ALKF1174L;DBHiCre or TH-NMYC transgenic mice was significantly higher than in normal adrenals but lower than tumors from ALKF1174L;DBHiCre transgenic mice (fig. S9B). Cluster analysis revealed a MYCN and an ALK cluster in tumors from TH-NMYC;ALKF1174L;DBHiCre mice (fig. S9A). Gene expression rank contributes to the ALK activity score. The presence of both clusters might result in an ALK activity score lower than that of tumors driven exclusively by ALK. Using immunoblotting, we assessed ALK and MYCN protein expression and phosphorylation of ALK, extracellular signal–regulated kinase 1 and 2 (ERK1/2), and MYCN Ser62 (S62) (fig. S10). Relative quantification of ALK expression and comparison to blot 2E did not support the idea that MYCN might stabilize ALK (fig. S11). However, MYCN protein levels were slightly elevated in TH-NMYC;ALKF1174L;DBHiCre tumors, and this was accompanied by strong MYCN S62 phosphorylation.

Array comparative genomic hybridization (aCGH) revealed a significant reduction in the number of genomic aberrations in tumors from triple-transgenic mice compared to tumors from TH-NMYC (18) or ALKF1174L;DBHiCre mice (Fig. 3G). Strikingly, human neuroblastomas harboring both ALK mutation and MYCN amplification also had significantly fewer chromosomal aberrations than tumors with either MYCN amplification or ALK mutation alone (Fig. 3H). This reduction of aberrations in human and murine tumors driven by MYCN and ALK, together with the very short time to tumor formation observed in the triple-transgenic mice, indicates that activating both MYCN and ALK pathways may be sufficient to drive neuroblastomagenesis. Our experiments demonstrate a strong synergism between ALK and MYCN in initiating murine neuroblastoma.

TAE-684 causes complete regression of ALK-driven murine neuroblastomas

The ALKF1174L mutation has been shown to confer relative (at least 10-fold) resistance to the clinically used ALK inhibitor crizotinib (12, 19). Whether other ALK inhibitors may be more effective is clinically relevant. Two mice with small palpable abdominal tumors detectable with BLI and without detectable metastasis were treated orally with TAE-684 once daily. Tumors were no longer detectable by palpation, BLI, or at autopsy after 14 days, demonstrating that TAE-684 is highly effective against ALK-driven murine neuroblastomas and can cause complete tumor regression (Fig. 4A). To confirm TAE-684 efficacy at the histological level, and to prove that we were treating neuroblastomas, we treated another mouse with a large abdominal tumor twice daily with TAE-684 for only 120 hours. BLI revealed a gradual and significant fourfold decrease in luciferase activity during treatment and a small tumor upon autopsy (Fig. 4B). Histology of the remaining tumor confirmed it as a neuroblastoma with extensive necrosis and signs of differentiation, similar to human neuroblastomas after chemotherapy, demonstrating that this tumor was in the process of regression (Fig. 4C). As tumors in the ALKF1174L;DBHiCre transgenic mice are driven by a single strong oncogene, the strong response to TAE-684 treatment is most likely due to oncogene addiction to ALKF1174L. The effectiveness of inhibiting only one driving oncogene in a tumor driven by ALKF1174L and MYCN is a clinically important question, because tumors harboring the ALKF1174L mutation and MYCN amplification define the most aggressive neuroblastomas. To investigate this, we treated TH-NMYC;ALKF1174L;DBHiCre mice with either TAE-684 or control after tumors became palpable and detectable by BLI. No metastases were detected in any animal at the beginning of treatment. After 14 days of treatment, control mice had significantly larger tumors than mice treated with TAE-684, which had small but detectable tumors (Fig. 4D and fig. S11). These experiments provide proof of principle that ALK inhibitory therapy using TAE-684 is effective against murine neuroblastomas initiated and driven by ALKF1174L alone as well as by both ALKF1174L and MYCN.

Fig. 4

TAE-684 induced complete regression of neuroblastomas in ALKF1174L transgenic mice. (A) TAE-684 treatment (10 mg/kg by oral gavage) was initiated when small tumors were palpable and detectable by BLI, and continued for 14 days. No tumor was palpable or detectable by BLI or at autopsy (only mouse 5 underwent autopsy) after treatment. (B) BLI before treatment and after 60 and 120 hours of treatment of large tumor in mouse 7 treated with TAE-684 (10 mg/kg, twice daily). Relative luminescence levels were measured for a defined region of interest (ROI; relative units). (C) Histology of the large tumor derived from mouse 7 after 120 hours of TAE-684 treatment. (D) Mice triple-transgenic for TH-NMYC;ALKF1174L;DBHiCre were treated for 14 days with TAE-684 or control, after which tumors were harvested and weighed.

TAE-684 reduces viability of human ALKF1174L-mutated neuroblastoma cells

The SH-SY5Y human neuroblastoma cell line, which harbors the ALKF1174L mutation, was treated with TAE-684 or crizotinib (Fig. 5A). TAE-684 treatment significantly reduced SH-SY5Y cell viability compared to controls or cultures treated with similar doses of crizotinib (Fig. 5A). Higher crizotinib doses further inhibited cell viability, suggesting that crizotinib resistance conferred by the ALKF1174L mutation can be overcome by increasing dosage (19). TAE-684 treatment of established tumors derived from SH-SY5Y cells subcutaneously xenografted into nu/nu mice induced near-complete regression, whereas tumors continued to grow in control or crizotinib-treated mice until euthanasia was necessary (Fig. 5, B and C). Although near-complete tumor regression was observed in TAE-684–treated mice, tumors relapsed after TAE-684 treatment ceased. Histological analysis of tumors harvested after only short-term TAE-684 treatment revealed extensive necrosis, more apoptotic cells, and fewer Ki-67–positive cells compared with tumors from control or crizotinib-treated mice (Fig. 5D). Phosphorylation of ALK and the downstream signaling elements, AKT and ERK1/2, was suppressed exclusively in TAE-684–treated tumors. We conclude that TAE-684 is effective against human tumor cells harboring the ALKF1174L mutation, as predicted by our transgenic mouse model.

Fig. 5

TAE-684 reduces viability of human ALKF1174L-mutated neuroblastoma cells. (A) SH-SY5Y neuroblastoma cells were treated with either TAE-684, crizotinib, or carrier solution (control) in culture, and cell viability was determined using the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay. (B) Mice were subcutaneously inoculated with 2 × 107 SH-SY5Y neuroblastoma cells and treated after palpable tumors appeared. Treatment day 1 was 15 days after tumor cell inoculation, and tumor volumes at this time were 128 ± 18 mm3 in the control group, 129 ± 29 mm3 in mice treated with crizotinib, and 161 ± 40 mm3 in mice treated with TAE-684 (entire range, 100 to 195 mm3). Mice were treated daily with TAE-684, crizotinib, or carrier solution (control), and tumor size was measured with a digital caliper. Mean tumor size is displayed until time of euthanasia of the first mouse in each group (treatment day 12 for control and crizotinib groups, treatment day 28 for TAE-684 group). (C) Kaplan-Meier plot for mice harboring SH-SY5Y xenografts treated with TAE-684, crizotinib, or control, using tumor size exceeding 2000 mm3 as the endpoint. (D) Representative images of tumor histology and immunohistochemistry after TAE-684, crizotinib, or control treatment. Cleaved caspase-3 is a marker for apoptosis. Ki-67 is a marker for proliferation. (E) Phosphorylation of ALK, AKT, and ERK1/2 in tumors treated with TAE-684, crizotinib, or control was analyzed on Western blots.

Discussion

Mutations activating the ALK kinase domain have recently been identified in a subset of neuroblastomas. Here, we demonstrate that the ALKF1174L mutant is sufficient to drive neuroblastoma oncogenesis in transgenic mice. Using a conditional Cre/Lox approach, we specifically expressed ALKF1174L in the neural crest, the tissue neuroblastoma is thought to originate from. Developing tumors not only were unambiguously identified as neuroblastomas but also recapitulated the anatomical location, tumor histology, and metastasis pattern of human neuroblastomas.

Tumor incidence and time to tumor development in our ALKF1174L transgenic mouse model were also comparable to the only other mouse model for neuroblastoma, the TH-NMYC mouse, which developed tumors at 3 to 6 months of age with an incidence of about 20% in the C57Bl/6 background (16). Time to tumor formation in relation to life span for both mouse models is longer than for humans, but the absolute time until tumor development is similar in both species, possibly reflecting that a certain number of secondary hits are required before a malignant tumor develops. The similar time required for tumor development is plausible because the mutation rate is comparable in these species. Time to tumor development is significantly shortened in TH-NMYC;ALKF1174L;DBHiCre mice, which also require fewer spontaneous secondary hits, further supporting this idea. The observation that many ALKF1174L transgenic mice expressed ALKF1174L in their adrenals but did not develop tumors is also in line with this observation. ALK expression levels in normal adrenal cells varied but remained below ALKF1174L tumor cell expression, and the fraction of adrenal cells expressing ALK varied between mice and was also lower than in tumors. The long (~3 kb) stop cassette, which decreases the efficiency of Cre-mediated excision, is most likely the cause. We consider this an advantage, because our system recapitulates the situation where not all cells of a given tissue are transformed by expression of an oncogene. Normal adrenal medulla cells remain during spontaneous initiation of human neuroblastomas and might interact with the not yet fully transformed ALKF1174L-expressing cells.

One unanswered research question is whether a specific developmental window exists for neuroblastomagenesis. Neuroblastoma is believed to originate from sympathoadrenergic progenitor cells that are present only early in life. This is considered the reason why neuroblastoma does not develop during adult life. This question could be addressed experimentally in our mouse model using a tamoxifen-dependent Cre variant expressed under the control of the DBH promoter, thus allowing ALK activation at different developmental time points by administering tamoxifen to the mice.

Chromosomal aberrations syntenic to those observed in human neuroblastomas were detected in the murine tumors, indicating that similar molecular pathways drive neuroblastoma oncogenesis in our ALKF1174L transgenic mouse model and human disease. The high number of chromosomal aberrations suggests that secondary hits are required for ALKF1174L to establish the neuroblastoma. The high synteny between chromosomal regions altered in ALK-driven murine neuroblastomas and human neuroblastomas suggests that our model recapitulates the genomic makeup of human neuroblastoma, including secondary hits. Specific aberrations also occurred in TH-NMYC mouse tumors. The human chromosome 1q region, often gained in human neuroblastomas, is syntenic to murine chromosome 3. The entire or partial murine chromosome 3 was gained in all ALK-driven tumors and most MYCN-driven tumors examined, further emphasizing the functional relevance of this region for neuroblastoma formation, independent of the tumor-driving oncogene (18). Some ALK- and MYCN-driven tumors also exhibited partial gains of murine chromosome 11, specifically syntenic to (almost the entire) human chromosome 17(q), which is gained in 50% of all human neuroblastomas, emphasizing the central importance of DNA copy number gain for this chromosomal segment.

One ALK-driven murine neuroblastoma spontaneously acquired MYCN amplification. MYCN is a key oncogene in neuroblastoma tumor biology and is amplified in the most aggressive neuroblastoma subgroup. MYCN overexpression is also sufficient to drive neuroblastomagenesis in TH-NMYC transgenic mice (16). The spontaneous occurrence of MYCN amplification together with other genetic aberrations characteristic of human neuroblastomas in ALK-driven murine neuroblastomas is striking. These recapitulate the spectrum of genetic alterations and tumor biology observed in the human disease, ranging from simple numerical chromosomal aberrations, various structural aberrations, gain of human chromosome 17 syntenic regions, and MYCN amplification. The ALKF1174L mutation has been detected in all subtypes of human neuroblastomas (711). Although the TH-NMYC mouse model exclusively reflects the most aggressive, MYCN-amplified neuroblastoma subset, our ALK-driven mouse model recapitulates the entire spectrum of human neuroblastoma, broadening its application to study neuroblastomagenesis and treatment.

MYCN amplification has been associated with ALKF1174L mutation in human neuroblastomas in the largest study to date, comprising a meta-analysis of >700 primary neuroblastomas (11). We also observed spontaneous MYCN amplification in one ALK-driven murine neuroblastoma. On the basis of these findings, we assessed whether ALKF1174L and MYCN not only are associated but also functionally synergize to induce neuroblastoma. Expressing both the ALKF1174L and the TH-NMYC transgenes in the neural crest markedly accelerated neuroblastoma formation. Conceivably, ALK might directly or indirectly stabilize or induce MYCN or vice versa. Although we could not substantiate ALK stabilization by MYCN here and cannot determine whether MYCN induces ALK expression (20), because the transgenic mouse overexpresses ALK independently of the intrinsic ALK promoter, strong MYCN S62 phosphorylation was demonstrated in tumors arising in TH-NMYC;ALKF1174L;DBHiCre transgenic mice. ALK signals via ERK1/2, and ERK1/2 phosphorylates S62 to stabilize MYCN. This functional link between MYCN and ALKF1174L might explain the synergism in murine neuroblastomagenesis and the correlation in patients, and warrants further functional analysis. We observed a strong reduction of chromosomal aberrations in TH-NMYC;ALKF1174L;DBHiCre mouse tumors. MYCN and ALKF1174L may be sufficient to drive neuroblastomagenesis without further secondary hits, making TH-NMYC;ALKF1174L;DBHiCre transgenic mice a valuable tool to further analyze mechanisms contributing to tumorigenesis in a strictly defined more stable genomic setting.

It is tempting to speculate whether overexpression of ALKR1275Q or even wild-type ALK would also drive neuroblastoma development in mice. Our initial assessment in cultured murine neural crest progenitor cells identified ALKF1174L as having the highest transforming capacity, an observation in accordance with a published assessment in Ba/F3 cells (11). Observations in human neuroblastoma families support the notion that mutations other than ALKF1174L are able to induce neuroblastoma development in humans. However, penetrance is variable in these families. Overexpression and amplification of wild-type ALK correlate with adverse patient outcome (11), but conclusive evidence that ALK overexpression alone can induce neuroblastomagenesis in humans or mice does not yet exist. It seems counterintuitive that mutated ALK might drive neuroblastomagenesis of varying aggressiveness in both humans and mice without affecting the course of disease. All members of human neuroblastoma families carry the same ALK mutation, whereas only some develop highly aggressive MYCN-amplified neuroblastomas and others develop benign ganglioneuromas or no tumors at all. Considering that cancer develops as the result of the accumulation of multiple mutations, the initial hit is in most cases insufficient to confer the fully transformed and malignant phenotype to tumor cells. The aggressiveness of the tumor may depend on the secondary hits obtained during a multistep process of tumorigenesis. Tumor cells remain addicted to mutated ALK, the initial hit, as demonstrated by their susceptibility to ALK inhibitor. ALK appears to remain a driving oncogene in the presence of secondary hits that, in part, determine tumor aggressiveness.

Our results also indicate that ALKF1174L may be a central and nonredundant component of neuroblastoma tumor biology, presenting it as an important drug target for neuroblastoma treatment. Although crizotinib has been shown to be effective against non–small cell lung carcinomas and other tumors harboring activated ALK, the ALKF1174L mutation confers resistance to crizotinib. TAE-684 is an ALK inhibitor in the preclinical phase of development that is effective even against tumor cells harboring the ALKF1174L mutation. Oral treatment of tumor-bearing ALKF1174L transgenic mice with TAE-684 induced complete tumor regression. This is notable because it (i) demonstrates oncogene addiction of the neuroblastomas to ALKF1174L, (ii) provides strong preclinical evidence and a preclinical model system to accelerate transfer of ALK inhibitory treatment to the clinic, and (iii) provides proof of principle for the effectiveness of treating tumors harboring the ALKF1174L mutation with second-line ALK inhibitors, such as TAE-684. Even neuroblastomas driven by the two strong oncogenes, MYCN and ALKF1174L, were susceptible to TAE-684 treatment. This is clinically important because MYCN-amplified neuroblastomas harboring the ALKF1174L mutation constitute the most aggressive disease subgroup.

Here, we demonstrate that a kinase-activating ALK mutation is sufficient to drive neuroblastoma oncogenesis in mice. Our data make a case for setting the development of therapy specifically targeting mutated ALK as a high priority for neuroblastoma treatment. The ALKF1174L transgenic mouse model not only yields further insight into neuroblastoma pathogenesis but also recapitulates the clinical and molecular spectrum of human neuroblastoma, making it a valuable tool for neuroblastoma research. The recapitulation of typical human neuroblastoma genomic aberrations in our mice is noteworthy because secondary spontaneous chromosomal aberrations in transgenic mouse models rarely resemble the human cancers they model (21). Because the ALKF1174L mutation occurs in other human malignancies as a secondary mutation after crizotinib treatment and confers crizotinib resistance, our humanized neuroblastoma mouse model provides a tool for preclinical ALK inhibitory therapy development with ALK inhibitors, such as TAE-684, designed to overcome resistance conferred by the ALKF1174L mutation.

Materials and Methods

Mouse generation, treatment, and imaging

The ALKF1174L complementary DNA was introduced into a Cre-conditional expression vector, comprising a 1.8-kb chicken β-actin promoter (22) (Fig. 1A). Transgenic mice were generated by pronucleus injection of this vector. Other transgenic mice used in this study were previously described: TH-NMYC (16), DBHiCre (14), and TH-IRES-Cre (15). On the basis of breeding schemas, all mice used for this study were hetero- or hemizygous for all alleles, including the ALKF1174L transgene.

In vivo TAE-684 or crizotinib treatment

TAE-684 or crizotinib (Axon Medchem) was resuspended in 10% 1-methyl-2-pyrrolidinone–90% PEG 300 (polyethylene glycol, molecular weight 300) (Sigma) solution. Mice received TAE-684 or crizotinib (10 mg/kg) by oral gavage once daily. Tumors were monitored by in vivo luciferase imaging.

ALK pathway activity score calculation

Tumors and normal murine adrenals were profiled on Affymetrix Murine 430 version 2 oligonucleotide microarrays according to the manufacturer’s protocol. Calculation of the ALK pathway activity score is based on the expression values of ALK-regulated genes established from eight different ALK inhibitor– and small interfering RNA–treated human neuroblastoma cell lines with an adaptation of a previously reported algorithm (23). This ALK pathway activity score was validated on a published data set of 252 human primary neuroblastomas (24) and a tetracycline-inducible human neuroblastoma cell line SK-N-AS model expressing either ALKWT, ALKF1174L, or ALKR1275Q in an ALK-negative background (24).

Statistical methods

Kaplan-Meier analyses and log-rank tests were performed with GraphPad Prism 5. Parametric data were compared with the Student’s t test. P values of <0.05 were considered to be significant. RT-qPCR data were analyzed with the qbasePLUS software 1.5.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/4/141/141ra91/DC1

Materials and Methods

Fig. S1. Transforming capacity of ALK in neural crest stem cells.

Fig. S2. Identification of founder mouse by analysis of luciferase activity in vitro.

Fig. S3. PCR genotyping of ALKF1174L transgenic mice.

Fig. S4. Histology and ALKF1174L expression in the substantia nigra.

Fig. S5. Histology of adrenals from mice not developing tumors.

Fig. S6. Bone marrow histology.

Fig. S7. RT-qPCR analysis of NSE and Ncam1 in murine neuroblastomas and normal tissues.

Fig. S8. PCR analysis of flox-out in tumors and normal tissues.

Fig. S9. Analysis of gene expression patterns in murine neuroblastomas and adrenals.

Fig. S10. Expression and phosphorylation of ALK, MYCN, and ERK1/2 in murine tumors.

Fig. S11. TAE-684 treatment of TH-NMYC;ALKF1174L;DBHiCre transgenic mice.

Table S1. Summary information for mice developing neuroblastomas.

Table S2. Chromosomal aberrations detected in tumors arising in ALK transgenic mice.

References and Notes

  1. Acknowledgments: We thank U. Rommerscheidt-Fuss, C. Golletz, S. Steiner, and N. Solomentsev for excellent technical assistance; N. Van Roy for expert assistance with fluorescence in situ hybridization analysis; K. Astrahantseff for manuscript proofreading; and H. Stephan for figure preparation. Funding: Supported by the National Genome Research Network (NGFNplus, PKN-01GS0894-6 to J.H.S., A.E., and A. Schramm), the German Cancer Aid (grant 108941 to J.H.S. and A.E.), the German Research Council (SFB 832 to L.C.H., R.T.U., and R.B.), and the Fund for Scientific Research (grants G.0198.08 and 31511809). A.E. is funded by the European Union (European Network for Cancer Research in Children and Adolescents: 7th EU Framework Programme, NoE No. 261474; Analysing and Striking the Sensitivities of Embryonal Tumours: 7th EU Framework Programme, CP No. 259348). A. Schramm is funded by the German Cancer Aid, German Ministry for Research and Education, and the German Research Foundation. F.P. and K.D.P. receive funds for Scientific Research-Flanders (Belgium) postdoctoral research fellowships. Author contributions: L.C.H., T.T., J.V., F.S., R.B., A.E., and J.H.S. designed the project; L.C.H., T.T., A. Schramm, K.D.P., C.K., B.D.W., A.O., M.P., S.L., A. Spruessel, F.P., P.M., B.M., S.K.-K., A.K., K.K., L.M., S.C., R.T.U., S.S., and J.H.S. performed the experiments and analyzed the data; J.H.S. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Affymetrix mRNA expression array data are accessible in the Gene Expression Omnibus database (accession no. GSE32386), and aCGH profiles are accessible in arrayCGHbase (http://medgen.ugent.be/arrayCGHbase/).
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article