EML4-ALK induces cellular senescence in mortal normal human cells and promotes anchorage-independent growth in immortalized normal human cells


 Background: Chromosomal inversions involving anaplastic lymphoma kinase (ALK) and echinoderm microtubule associated protein like 4 (EML4) generate a fusion protein EML4-ALK in non-small cell lung cancer (NSCLC). The understanding of EML4-ALK function can be improved by a functional study using normal human cells.Methods: Here we for the first time conduct such study to examine the effects of EML4-ALK on cell proliferation, cellular senescence, DNA damage, gene expression profiles and transformed phenotypes.Results: The lentiviral expression of EML4-ALK in mortal, normal human fibroblasts caused, through its constitutive ALK kinase activity, an early induction of cellular senescence with senescence-associated b-galactosidase activity, upregulation of p16INK4A and p21WAF1, and accumulated DNA damage. In contrast, when EML4-ALK was expressed in normal human fibroblasts immortalized by telomerase reverse transcriptase (hTERT), which is activated in the vast majority of NSCLC, the cells showed accelerated proliferation and acquired anchorage-independent growth ability in soft-agar medium, without accumulated DNA damage, chromosome aberration, nor p53 mutation. EML4-ALK induced the phosphorylation of STAT3 in both mortal and immortalized cells, but RNA sequencing analysis suggested that the different STAT3-regulated signaling pathways contributed to the different phenotypic outcomes in these cells. While EML4-ALK also induced anchorage-independent growth in hTERT-immortalized human bronchial epithelial cells in vitro, the expression of EML4-ALK alone did not cause detectable in vivo tumorigenicity in immunodeficient mice.Conclusions: Our data indicate that the immortalized state is critical for EML4-ALK to manifest its in vitro transforming activity in human cells. This study provides the isogenic pairs of human cells with and without EML4-ALK expression.

EML4-ALK induces cellular senescence in mortal normal human cells and promotes anchorageindependent growth in immortalized normal human cells and immortalized normal human cells and provide mechanistic insight into its role during human carcinogenesis.

Methods
Cells and cell culture CRL-2097, BJ and MRC-5 and NIH/3T3 were obtained from ATCC (Manassas, VA, USA) and maintained in DMEM supplemented with 10% FBS. HBET1 [11] and their derived cells were maintained in LHC-9 medium (Thermo Fisher Scienti c, Waltham, MA USA) supplemented with 2 mM L-glutamine. H3122 was obtained from the NCI Repository of Tumor Cell Lines (Frederick, MD, USA) and maintained in RPMI 1640 medium supplemented with 10% FBS. For tetracycline (Tet)-inducible gene expression, doxycycline (Dox, at 1 μg/ml) was added. The number of population doubling levels (PDL) achieved between passages was determined by log 2 (number of cells obtained/number of cells inoculated) [12]. Crizotinib was purchased from Selleck Chemicals (Houston, TX, USA) and used at a concentration of 25 nM. SA-β-gal staining was performed using the kit purchased from Cell Signaling Technology (Danvers, MA, USA).
The coding sequences in all newly constructed vectors were fully sequenced for con rmation.
Anchorage-independent colony formation in soft-agar medium Cells (1,000 cells/well) were seeded on 12-well plates in 1 ml of medium containing 0.35% agarose (SeaPlaque low-melting-temperature agarose, Lonza Biosciences, Alpharetta, GA, USA) over 1-ml volume of base layer consisting of the same culture medium and 0.5% agarose. Colonies of approximately 50 mm in diameter or larger were counted on day 21.
Telomere length measurement by quantitative PCR (qPCR) Telomere length was measured by the monochrome multiplex qPCR (MMQPCR) method as described previously [19]. Relative telomere length was express as a ratio of the quantity of telomeric DNA (T) normalized to the quantity of multiple copy sequence DNA (mcs), yielding T/M ratio. Intra-and interassay coe cient of variations (CVs) were 4.1% and 10.7%, respectively. G-band karyotyping and spectral karyotyping (SKY) Treatment with Colcemid (10 mg/ml; KaryoMax, Invitrogen, Carlsbad, CA, USA), hypotonic treatment (0.075 M KCl), xation with methanol/acetic acid (3:1), slide preparation and G-banding were performed as previously described [20] and the images were captured and analyzed with the HiBand system (Applied Spectral Imaging, Carlsbad, CA, USA). For spectral karyotyping (SKY), the slides were processed using the 24-color Human SKY Paint kit (Applied Spectral Imaging) according to the manufacturer's protocol. Spectral images of the hybridized metaphases were acquired using the HyperSpectral Imaging system (Applied Spectral Imaging) and analyzed using the HiSKY v.7.2 acquisition software (Applied Spectral Imaging). All cell lines were analyzed by G-band karyotyping. hTERT-CRL-2097, hTERT-CRL-2097+EML4-ALK, its derived soft-agar clones #1 and #2, and HBET1 were also analyzed by SKY.

Tumorigenicity in vivo
Female NOD.SCID/Ncr mice were obtained from Charles River Laboratories (#560; Germantown, MD, USA), and were maintained under speci c pathogen-free conditions and the animals had free access to feedstuff and water under 12:12 light/dark cycles, 22 and 30 -70 % humidity. These mice at 6 to 10 weeks of age were subcutaneously injected with cells (5 ´ 10 6 per ank) in 50% Matrigel (no. 354248, Corning, NY, USA) at both anks: one side with EML4-ALK-expressing cells or their derived soft-agar clones; and the other side with vector control cells. For injection of these cells, mice were anesthetized using iso urane in accordance with NCI Animal Care and Use Committee (ACUC) guidelines. As a positive control, NIH/3T3 cells expressing EML4-ALK were injected at both anks of two mice (2 ´ 10 6 per ank). Tumor size and body weight were measured twice a week until 8 weeks or tumor size excess 20 mm. To ameliorate the suffering of mice observed throughout experimental studies, mice were euthanized by CO 2 inhalation. After 8 weeks of administration, the mice were sacrificed and photographed, and the tumors were removed and weighed. This experiment was approved by the NCI ACUC (LHC-012-D).
RNA isolation, reverse transcription (RT)-PCR and Sanger sequencing of the entire coding region of p53 mRNA Total RNA isolation, reverse transcription, and 1st strand cDNA synthesis with random hexamers were carried out as previously described [12,15]. PCR ampli cation of the entire coding region of p53 mRNA was carried out using the Platinum Taq DNA Polymerase High Fidelity (Thermo Fisher Scienti c) with the primers 5'-ATG GAG GAG CCG CAG TCA-3' and 5'-GTC AGT CTG AGT CAG GCC CTT C-3'. The ampli ed PCR product was sequenced using the BigDye Terminator v1. Diego, CA, USA) and were sequenced with paired-end reads of 150 bp on HiSeq3000/4000 sequencer (Illumina) to obtain at least 30 million read pairs per sample. Reads were trimmed for adaptors and lowquality bases using Trimmomatic [21] and were aligned with the reference genome (Human-hg38) and the annotated transcripts (Gencode_v24) using STAR (https://github.com/alexdobin/STAR). The gene expression quanti cation analysis was performed using STAR/RSEM tools (http://deweylab.github.io/RSEM/) to obtain raw read counts and normalized read counts (RPKM) for each gene.

Data mining of publicly available lung adenocarcinoma datasets
For the TCGA dataset [23], mRNA expression from RNA Seq V2 RSEM and tumor characteristics were downloaded from cBioPortal for 510 patients with lung adenocarcinoma. Log2-transformed RSEM values for hTERT were used for the analyses.
For the microarray dataset GSE31210 [24], data and sample characteristics were downloaded from the Gene Expression Omnibus for 226 patients with lung adenocarcinoma using the R package GEOquery.
The raw intensity values were processed and normalized with Robust Multi-Array Average (RMA) using the R package oligo. Affymetrix IDs were mapped to HUGO Gene Nomenclature Committee IDs. Resulting RMA expression values for hTERT were used for the analyses.

EML4-ALK causes the early induction of cellular senescence in normal, mortal human broblasts
We transduced normal human broblasts CRL-2097, which are mortal, or have a limited replicative lifespan, with a lentiviral vector expressing the Tet repressor (generating CRL-2097/TR), and then with either a lentiviral vector encoding EML4-ALK or the control vector (Fig. 1a). While the control cells with no EML4-ALK expression reached approximately 11 or 12 population doubling levels (PDL) before proliferation arrest, the EML4-ALK-expressing cells ceased to proliferate earlier at PDL 6 after lentiviral transduction (Fig. 1b). In addition to the cells shown in (a) (observed through day 88 or 40), those retrovirally transduced with oncogenic Ras (H-RasV12) and its control vector (pBabe) were examined (Supplementary Fig. S1A-B; observed through day 20). Both the control and EML4-ALK-expressing cells, upon proliferation arrest, were similarly positive for senescence-associated b-galactosidase (SA-β-gal) (Fig. 1c). This early induction of cellular senescence by EML4-ALK was reminiscent of oncogenic Rasinduced senescence [13], although H-RasV12 induced senescence earlier after fewer PDL in this CRL-2097 strain (Fig. 1b, Supplementary Fig. S1A and S1B).
The p16 INK4A protein expression was increased during the replicative lifespan in both EML4-ALKexpressing and control cells, although its levels at senescence were lower than observed in H-RasV12induced senescence. An increase in p21 WAF1 protein expression was also observed similarly in the EML4-ALK-expressing and control cells, and to a lesser degree in H-RasV12-induced senescence. These ndings suggest that the similar levels of upregulation of p16 INK4A and p21 WAF1 occur with the EML4-ALKinduced senescence and natural replicative senescence, although the former undergoes fewer PDL than the latter to achieve those expression levels.
The EML4-ALK-induced senescence is associated with accumulated DNA damage A phosphorylated H2AX (g-H2AX) [14] was detected by immuno uorescence staining in CRL-2097/TR broblasts with EML4-ALK expression (EML4-ALK-Dox+ in Fig. 1b) and with the control vector (Vector in Fig. 1b) when approaching senescent proliferation arrest (at PDL 6 and PDL 12, respectively) ( Fig. 1e-f). The early induction of cellular senescence in the EML4-ALK-expressing cells was associated with signi cantly more accumulation of g-H2AX foci ( Fig. 1e-f), suggesting that EML4-ALK leads to an accelerated rate of persistent DNA damage. The level of p53 protein phosphorylated at serine 15, another indicator of DNA damage [27,28], was increased in both control and EML4-ALK-expressing cells upon senescence ( Supplementary Fig. S2A).
A quantitative PCR-based measurement of telomere length showed that the vector control cells underwent progressive telomere shortening through their replicative senescence at PDL 12 ( Supplementary Fig. S2B). The EML4-ALK-expressing cells had the telomere length shorter than that of the original cells before lentiviral transduction, similar to that of the control cells at a comparable PDL (PDL 7) and longer than that of the replicatively senescent control cells at PDL 12 ( Supplementary Fig.  S2B). These data suggest that normal, mortal human broblasts with and without EML4-ALK expression undergo telomere shortening at similar rates per PDL, and that the accelerated accumulation of DNA damage in the EML4-ALK-expressing cells may be of non-telomeric origin [14].
The EML4-ALK-induced senescence depends on its ALK kinase activity The EML4-ALK-expressing CRL-2097/TR with Dox addition were treated with an ALK tyrosine kinase inhibitor (TKI) Crizotinib at 25 nM (a concentration close to the reported IC 50 value [29]) and were monitored for cell proliferation. The treatment with Crizotinib inhibited the autophosphorylation of EML4-ALK while not affecting the expression level of EML4-ALK, con rming its TKI activity against the ALK kinase activity (Fig. 2a). The expression of EML4-ALK without Crizotinib reproducibly induced the proliferation arrest at PDL 6 ( Fig. 2b). In contrast, the Crizotinib-treated cells bypassed this early proliferation arrest and underwent approximately 4 more PDL, similarly to the control cells without EML4-ALK expression (Fig. 2b).
Another normal human broblast strain MRC-5 (without a Tet repressor) was transduced with the lentiviral vector encoding wild-type EML4-ALK or a kinase-dead mutant of EML4-ALK (K589M) [4], along with the control vector. The K589M mutant of EML4-ALK was constitutively expressed at a higher level than the wild-type counterpart but was not autophosphorylated (Fig. 2c). Since the MRC-5 broblasts used in this experiment were closer to natural replicative senescence than the CRL-2097 broblasts used above, the control vector-transduced cells underwent at most 5 PDL before they ceased to proliferate (Fig.  2d). The expression of wild-type EML4-ALK, again in these broblasts approaching natural replicative senescence, caused earlier induction of proliferation arrest (Fig. 2d) with SA-b-gal (Fig. 2e), which was similar to oncogenic Ras-induced senescence ( Supplementary Fig. S1C-D). Importantly, the K589M mutant-expressing cells did not show early proliferation arrest and behaved like the control vectortransduced cells (Fig. 2d-e), further supporting that the ALK kinase activity mediates the early induction of cellular senescence by EML4-ALK.
EML4-ALK accelerates cell proliferation and promotes anchorage-independent growth in hTERTimmortalized normal human broblasts Although EML4-ALK-positive NSCLC rarely have other accompanying genetic alterations [5], they still have a mechanism to maintain telomere length and function, in most cases via hTERT activation [30]. Consistent with the previous ndings [31], ALK fusion-positive NSCLC tumor tissues, as well as negative ones, were con rmed to express hTERT (Supplementary Fig. S3A and B). We thus hypothesized that EML4-ALK might cooperate with the expression of hTERT, leading to telomerase activation and cell immortalization, to cause cellular transformation in normal human cells. To test this hypothesis, hTERT was retrovirally transduced into CRL-2097 to establish an hTERT-immortalized normal human cell line (hTERT-CRL-2097), which maintained elongated telomeres ( Supplementary Fig. S4A). This cell line had normal karyotype (Supplementary Fig. S4B), maintained normal p16 INK4A response to oncogenic Ras ( Supplementary Fig. S4C) and retained wild-type TP53 (below in Fig. 3e), thus not coincident with the changes frequently associated with human cell immortalization. These hTERT-CRL-2097 cells were transduced with the wild-type EML4-ALK vector or the control vector, and the constitutive expression and autophosphorylation of the EML4-ALK protein was con rmed (Fig. 3a). Unlike in mortal CRL-2097, the expression of EML4-ALK in this immortalized cell line did not induce senescent proliferation arrest, but instead resulted in accelerated cell proliferation (Fig. 3b). The hTERT-CRL-2097 cells, with or without EML4-ALK expression, accumulated no or little DNA damage (g-H2AX foci in Fig. 3c and Supplementary  Fig. S5), consistent with no induction of cellular senescence and in contrast to DNA damage accumulation in mortal CRL-2097 broblasts (the leftmost bar in Fig. 3c; and Fig. 1e above). Furthermore, the anchorage-independent formation of cell colonies in soft-agar medium was enhanced in the EML4-ALK-expressing hTERT-CRL-2097 cells, which was inhibited by treatment with the ALK TKI Crizotinib (Fig.   3d).
In another hTERT-immortalized normal human broblast cell line, hTERT-BJ ( Supplementary Fig. S6A), the expression of wild-type EML4-ALK, but not of the K589M mutant, accelerated cell proliferation ( Supplementary Fig. S6B) and promoted anchorage-independent growth in soft agar ( Supplementary Fig.  S6C). The ability of wild-type EML4-ALK to promote anchorage-independent growth was again abrogated by treatment with the ALK TKI Crizotinib (Supplementary Fig. S6C). From our reproducible results in two lines of hTERT-immortalized normal human broblasts, we conclude that EML4-ALK has in vitro transformation activity in these cells through its constitutive ALK kinase activity.
The EML4-ALK-induced anchorage-independent growth occurs without chromosome aberrations, without loss or mutation of the TP53 gene, and with normal p16 INK4A response We performed karyotype analysis in hTERT-CRL-2097 cells transduced with the control vector, those expressing EML4-ALK and six of their derived clones isolated from soft-agar culture. All of these cells examined, as well as the original hTERT-CRL-2097 cells (as mentioned above and Supplementary Fig.  S4B), maintained normal male karyotype 46,XY ( Fig. 3e and Supplementary Fig. S7A-B). All cell lines listed in Fig. 3e had normal male karyotype 46, XY by G-banding (at least 10 well-spread metaphases per line were examined; an example is shown in Supplementary Fig. S7A). hTERT-CRL-2097, hTERT-CRL-2097+EML4-ALK and soft-agar clones #1 and #2 were also con rmed by SKY to have 46, XY (again, at least 10 metaphases per line were examined; an example is shown in Supplementary Fig. S7B). By direct sequencing of the RT-PCR products, the entire coding region of TP53 was shown to be wild-type without a clonal homozygous or heterozygous mutation. The polymorphic codon 72 of TP53 was heterozygous (Pro/Arg) in the original hTERT-CRL-2097 and remained heterozygous (Pro/Arg) in hTERT-CRL-2097 expressing EML4-ALK and their derived soft-agar clones (examples are shown in Supplementary Fig.  S7C), indicating that no chromosome instability nor loss of heterozygosity (LOH) occurred at TP53 during EML4-ALK-induced acquisition of anchorage-independent growth.
Although loss or mutation of the TP53 gene is frequently associated with cell transformation and carcinogenesis [32], all of the EML4-ALK-expressing cells and their derived soft-agar clones had the wildtype TP53 sequence without any homozygous or heterozygous mutation (Fig. 3e). The original hTERT-CRL-2097 cells showed Pro/Arg heterozygosity at the polymorphic codon 72, which was maintained in the EML4-ALK-expressing cells and all the soft-agar clones ( Fig. 3e and Supplementary Fig. S7C), indicating that there was no loss of a TP53 allele in these cells. The p16 INK4A pathway, which is frequently impaired during cell transformation and carcinogenesis [26,33], was suggested to remain intact during the EML4-ALK-mediated cell transformation by the nding that oncogenic Ras-induced p16 INK4A upregulation [13] was observed in the EML4-ALK-expressing cells (Supplementary Fig. S7D).

STAT3 is activated by EML4-ALK in both mortal and immortalized normal human broblasts
We examined the activation status of three major factors downstream of EML4-ALK (i.e., STAT3, Akt and Erk1/2) by western blot (Fig. 4a and 4b). Mortal normal human broblasts CRL-2097/TR with and without EML4-ALK expression both showed increased levels of phosphorylated Akt when they became senescent, while the levels of phosphorylated Erk1/2 did not show a consistent change associated with EML4-ALK expression or increased PDL levels (Fig. 4a). Notably, a striking induction of STAT3 phosphorylation was observed in the EML4-ALK-expressing cells at PDL 4, which was decreased upon senescent proliferation arrest (PDL 6) but still at a higher level than that in the control cells at senescence (PDL 11) (Fig. 4a). Also in hTERT-immortalized CRL-2097 broblasts, the expression of EML4-ALK resulted in remarkable induction of phosphorylated STAT3, while no increase in phosphorylated Akt and a slight increase in phosphorylated Erk1/2 were associated with EML4-ALK expression (Fig. 4b). These ndings suggest that STAT3 functions as a major downstream effector of EML4-ALK in both mortal and hTERT-immortalized normal human broblasts, consistent with its involvement in both cellular senescence and transformation [34][35][36][37]. MiR-21, a microRNA induced by STAT3 [38], was also upregulated by EML4-ALK in both mortal and hTERT-immortalized broblasts (Supplementary Figure S8). Although a decrease in phosphorylated Src was associated with cellular senescence in the presence or absence of EML4-ALK (Fig. 4a), the expression of EML4-ALK did not affect the phosphorylation level of Src ( Fig. 4a and 4b), which is known to mediate acquired resistance to ALK TKIs [39].

EML4-ALK regulates different signaling pathways in mortal and immortalized normal human broblasts
We performed RNA sequencing (RNA-seq) in duplicated samples of mortal CRL-2097/TR with and without EML4-ALK expression (Supplementary Fig. S9A; Supplementary Table S1) and hTERTimmortalized CRL-2097 with and without EML4-ALK expression (Supplementary Fig. S9B; Supplementary Table S2). The analysis of the differentially expressed genes to KEGG (Kyoto Encyclopedia of Genes and Genomes) Pathways (https://www.genome.jp/kegg/) identi ed the cytokine-cytokine receptor interaction pathway as signi cantly upregulated by EML4-ALK expression in mortal CRL-2097/TR broblasts, along with some related and overlapping pathways involving tumor necrosis factor (TNF) or viral infection (Table 1 and Supplementary Table S3). The interferon (IFN)-a/b and IFN-g signaling pathways, which have some genes in common with the above KEGG pathways, were also identi ed from Reactome Pathway Database (https://reactome.org/) as EML4-ALK-upregulated pathways in mortal CRL-2097/TR broblasts (Table 1 and Supplementary Table S3). Consistent with the EML4-ALK-induced phosphorylation of STAT3 (as above in Fig. 4), these cytokines and IFN signaling pathways include several STAT3-upregulated genes such as IL1B, CXCL8, SOCS3, IRF1 and IRF7 (Table 1 and  Supplementary Table S3). These data suggest that STAT3 mediates the effect of EML4-ALK on activating the proin ammatory cytokine and IFN signaling cascades, which coordinately induce and maintain sustained DNA damage and senescent proliferative arrest in mortal normal human cells [37,40].
In hTERT-immortalized CRL-2097 broblasts, instead of the above-mentioned cytokine and IFN signaling pathways, the complement and blood coagulation cascades signaling was identi ed as signi cantly upregulated by EML4-ALK expression (Table 1 and Supplementary Table S3). This pathway included A2M, PLAT and PLAU as STAT3-upregulated genes [41,42], suggesting that STAT3 also mediates the modulation by EML4-ALK of blood coagulation, which may have clinical implications in increased risk of disseminated intravenous coagulation in patients with EML4-ALK-positive cancer [43]. Consistently, a Japanese cohort also showed an upregulation of the blood coagulation pathway in EML4-ALK-positive lung cancer [24] (Supplementary Fig. S10). We also found that some integrin and non-integrin components of focal adhesion and extracellular matrix (ECM) interactions were downregulated by EML4-ALK in hTERT-immortalized CRL-2097 broblasts (Table 1 and Supplementary Table S3), which likely contributed to EML4-ALK-induced anchorage-independent growth via overcoming anoikis [44,45].

EML4-ALK also has in vitro transforming activity in hTERT-immortalized normal human bronchial epithelial cells but does not cause in vivo tumorigenicity
The transforming activity of EML4-ALK was also tested in hTERT-immortalized, normal human bronchial epithelial cells, which represent a cell type more relevant to NSCLC pathogenesis. For this purpose, we used a previously established cell line, HBET1, which has a tetraploid karyotype with no or few structurally abnormal chromosomes ( Supplementary Fig. S11), maintains elongated telomeres and does not show anchorage-independent growth or in vivo tumorigenicity [11]. The HBET1 cells constitutively expressing EML4-ALK with its autophosphorylation (Fig. 5a) showed accelerated cell proliferation (Fig.   5b) and acquired anchorage-independent growth in soft agar (Fig. 5c), as observed above in hTERTimmortalized broblasts. The increased levels of phosphorylation of STAT3 and Akt, but not of Erk1/2 or Src, were associated with EML4-ALK expression in HBET1 cells (Fig. 5a).
To examine in vivo tumorigenicity, we injected hTERT-immortalized HBET1 and CRL-2097 cells with and without EML4-ALK, along with the soft-agar clones derived from the EML4-ALK-expressing cells, subcutaneously into immunode cient NOD.SCID/Ncr mice (Fig. 5d). None of these cells were able to form a growing tumor, while EML4-ALK-expressing mouse NIH/3T3 cells as a positive control [4] consistently formed progressively growing tumors (Fig. 5d). These results suggest that the expression of EML4-ALK alone is not su cient for hTERT-immortalized normal human cells to acquire in vivo tumorigenicity.

Discussion
This study for the rst time performed the functional assays of an ALK fusion protein using normal human cells, which were mortal or immortalized. The major ndings include: i) EML4-ALK, through its constitutive ALK kinase activity, induces mortal cells to undergo early senescence ( Fig. 1 and 2); ii) EML4-ALK promotes immortalized cells to proliferate anchorage-independently ( Fig. 3 and 5); iii) Accumulation of DNA damage in mortal but not immortalized cells (Fig. 1e and 3c) is associated with senescence induction; and iv) EML4-ALK-induced activation of STAT3 (Fig. 4) regulates different signaling pathways in mortal and immortalized cells (Table 1).
In mortal normal human broblasts, the expression of EML4-ALK caused senescent proliferation arrest earlier than natural replicative senescence, but later than senescence induced by oncogenic Ras (Fig. 1b and Supplementary Fig. S1), which was previously reported to cause severer DNA damage [13,46].
Interestingly, a later phase of replicative lifespan of EML4-ALK-expressing mortal broblasts seemed to have selected against cells with its higher expression level (compare Dox+ at PDL 4 and 6 in Fig. 1a), possibly leading to an underestimation of the activity of EML4-ALK. In agreement with this notion, the inactive mutant K589M showed a higher expression level than wild-type EML4-ALK in mortal broblasts (Fig. 2c) and immortalized broblasts did not show a selection against the higher expression level during expansion (Fig. 3a).
Our ndings highlight a signi cant contrast of EML4-ALK activities between mortal and immortalized normal human cells. Since hTERT and telomerase are suggested to contribute to genome stability not only at telomeres but also at non-telomeric sites [47,48], hTERT-immortalized cells may become less vulnerable to telomeric and non-telomeric DNA damage, leading to a functional switch of EML4-ALK from senescence induction to in vitro transformation. Consistently, while the expression of EML4-ALK in mortal normal human broblasts activated the STAT3-mediated signaling to DNA damage and senescence involving proin ammatory cytokines and IFN pathways [37,40] (Table 1 and Supplementary Table S3), these pathways were not activated in EML4-ALK-expressing hTERT-immortalized broblasts, which instead showed changes in gene expression that could contribute to their anchorage-independent growth (Table 1 and Supplementary Table S3). Chromosome abnormality, loss or mutation of TP53, or impairment of p16 INK4A response was not associated with EML4-ALK-induced anchorage-independent growth ( Fig. 3e and Supplementary Fig. S7).
These data are consistent with a previous nding that no or few chromosomal changes, other than an ALK-rearranging translocation, were observed in ALK fusion-positive tumors [1] and may re ect the molecular features of ALK fusion-positive tumors infrequently accompanying other genetic changes [5].
Despite their ability of anchorage-independent growth, the EML4-ALK-expressing, hTERT-immortalized normal human cells of broblastic origin (hTERT-CRL-2097) and of bronchial epithelial origin (HBET1) did not form tumors in immunode cient mice under our experimental conditions (Fig. 5d), while mouse NIH/3T3 cells expressing EML4-ALK were highly tumorigenic [4]. This contrast further highlights the importance of this study using normal human cells and prompts us to identify a factor(s) or event(s) that cooperates with EML4-ALK for acquisition of in vivo tumorigenicity.

Conclusions
Our data indicate that the immortalized state is critical for EML4-ALK to manifest its in vitro transforming activity in human cells. Out data also suggest that STAT3 is major downstream of EML4-ALK, but the different STAT3-regulated signaling pathways contributed to the different phenotypic outcomes in normal and immortalized cells. In addition, this study has established the isogenic pairs of human cell lines with and without EML4-ALK expression. We expect that these isogenic cells will be a novel in

Consent for publication
Not applicable.

Availability of data and material
All data generated and analyzed in this study are included in this published article and its supplementary information les. Raw data and materials are available from the corresponding author upon request.

Competing interests
The authors have no con icts of interest to declare. Author's contributions AM, IH and CCH conceived and designed the study; AM, MM, JB, and IH performed the experiments. TO, HO, HT, SB, AR, MK and DL contributed data analysis. MS, AG and HM provided essential materials and expertise. AM, MM and IH wrote the manuscript. CCH was responsible for the overall project. All authors provided critical revision and approved the nal manuscript.    The HBET1 cells with EML4-ALK or the control vector were examined for anchorage-independent growth, as in Fig. 3d and S5C. Data analysis and presentation are also as in Fig. 3d and S5C. Arrows in a representative image indicate colonies that formed in soft-agar medium. Scale bars, 50 μm. ** P < 0.01. (d) Summary of tumorigenicity assay in NOD.SCID/Ncr mice. Cells were subcutaneously injected into each ank (5 106 per ank) of mice at 6-10 weeks of age, followed by observation until 8 weeks after injection. In each mouse, one ank had EML4-ALK-expressing cells or soft-agar clones, and the other ank had cells with the control vector. Whereas NIH/3T3 cells expressing EML4-ALK (2 106 per ank, at both anks of two mice) produced tumors necessitating euthanasia, no progressively growing tumors formed from any of hTERT-CRL-2097-and HBET1-derived cells.