Alternative transcription initiation leads to expression of a novel ALK isoform in cancer

Abstract

Activation of oncogenes by mechanisms other than genetic aberrations such as mutations, translocations, or amplifications is largely undefined. Here we report a novel isoform of the anaplastic lymphoma kinase (ALK) that is expressed in ∼11% of melanomas and sporadically in other human cancer types, but not in normal tissues. The novel ALK transcript initiates from a de novo alternative transcription initiation (ATI) site in ALK intron 19, and was termed ALK(ATI). In ALK(ATI)-expressing tumours, the ATI site is enriched for H3K4me3 and RNA polymerase II, chromatin marks characteristic of active transcription initiation sites. ALK(ATI) is expressed from both ALK alleles, and no recurrent genetic aberrations are found at the ALK locus, indicating that the transcriptional activation is independent of genetic aberrations at the ALK locus. The ALK(ATI) transcript encodes three proteins with molecular weights of 61.1, 60.8 and 58.7 kilodaltons, consisting primarily of the intracellular tyrosine kinase domain. ALK(ATI) stimulates multiple oncogenic signalling pathways, drives growth-factor-independent cell proliferation in vitro, and promotes tumorigenesis in vivo in mouse models. ALK inhibitors can suppress the kinase activity of ALK(ATI), suggesting that patients with ALK(ATI)-expressing tumours may benefit from ALK inhibitors. Our findings suggest a novel mechanism of oncogene activation in cancer through de novo alternative transcription initiation.

Extended Data Figure 1. Comparison of the RNA-seq profiles of various ALK transcripts

RNA-seq data are displayed in the Integrative Genomics Viewer (IGV). The grey bars/arrows indicate the sequencing reads. The blue lines connect sequencing reads that are aligned over the splice site of joining exons. a, The ALK^ATI transcript shows expression of ALK exons 20–29 and of ~ 400 bp in intron 19 (blue shaded area). No expression of exon 1–19 or intronic areas, other than in intron 19, are observed. The detailed view illustrates that the sequencing reads align continuously between exon 20 and intron 19 indicating uninterrupted transcription. The 5′-UTR of ALK^ATI (intron 19) and exon 20–29 are expressed at comparable levels. b, The full-length wild-type ALK transcript shows expression of all ALK exons and only very little expression of the introns. The detailed view displays that the sequencing reads align sharply to the exons, but not to the intron 19 region, which is present in ALK^ATI (blue shaded area). c, The ALK fusion transcript of a non-small cell lung cancer with an EML4–ALK translocation shows expression of ALK exons 20–29, and little expression of exons 1–19 and all introns. The detailed view illustrates that the transcription starts mainly at exon 20 due to a preserved splice site. Only few reads are aligned to the intron 19 region (blue shaded area). The green-labelled reads highlight chimaeric read pairs indicating the EML4–ALK translocation.

Extended Data Figure 2. Identification of the ALK^ATI transcript

a, Northern blot of wild-type ALK-expressing neuroblastoma cell lines (SK-N-DZ and SK-N-BE2), EML4–ALK-expressing lung cancer cell lines (H3122, variant (V) 1 and H2228, variant (V) 3), ALK^ATI-expressing melanoma, one anaplastic thyroid carcinoma (ATC-28), and negative controls (melanoma cell lines). Except for the negative controls, each lane shows two bands: the lower B-band matches the shorter canonical (RefSeq) ALK transcript ending at ~ chr2:29,415,640; the upper A-band corresponds to a transcript with a 1.8 kb longer 3′-UTR ending at ~chr2:29,413,840. Two ALK^ATI-expressing melanomas, MM-284 and MM-74, show only weak signals because less than 1 μg RNA was available; for all other samples 5–10 μg RNA were used. See Supplementary Fig. 1 for uncropped blots. b, RNA-seq data displayed in IGV. The Sashimi plot illustrates the shorter B and the longer A ALK transcripts by the sharp drop of sequencing reads in the 3′-UTR at chr2:29,415,640 for the B and at chr2:29,413,840 for the A transcript. c, IGV view of the 5′-RACE-cDNA fragments obtained by massively parallel sequencing. More than 95% of the sequencing reads (grey arrows) start within the main ATI site of 25 bp (hg19 chr2:29,446,744–29,446,768). d, Sanger sequencing of the cloned 5′-RACE-cDNA fragments confirms the continuous transcription starting in ALK intron 19 and extending to exons 20 and 21. e, The ALK^ATI transcript consists of ~ 400 bp upstream of exon 20 and of ALK exons 20–29. The transcriptional initiation site was defined as the first base pair at which more than 5% of the transcripts were initiated (chr2:29,446,766). Other major transcription initiation sites are marked in red, the 5′- and 3′-UTRs in dark blue, the coding DNA sequence (CDS) in black, and the first and last base of each exon in light blue. The translation is initiated at 3 start codons (ATGs; bold and underlined): first ATG, hg19 chr2:29,446,360–29,446,362; second ATG, (+ 7–9); and third ATG (+ 61–3). f, The amino acid sequence of ALK^ATI. The translation is initiated at 1 of 3 start codons. The corresponding 3 methionines (bold and underlined) result in 3 different proteins, 61.08 kDa (552 amino acids), 60.82 kDa (550 amino acids), and 58.71 kDa (532 amino acids). The kinase domain is highlighted in red. The lysine in the ATP binding domain is marked bold and underlined, and was mutated to methionine (referring to wild-type ALK: p.K1150M) in the kinase-dead ALK^ATI-KD.

Extended Data Figure 3. RNApol II and H3K4me3 are enriched at the ATI site of ALK^ATI-expressing tumour samples

a, b, ChIP–qPCR of H3K4me3 (a) and RNApol II (b) at the ATI site demonstrating enrichment of both marks in the ALK^ATI-expressing human tumour samples, but not in the negative controls, including a lung cancer cell line with EML4–ALK translocation (H3122) and a melanoma cell line (SKMEL-524). Error bars show mean ± s.e.m.; n = 3 technical replicates.

Extended Data Figure 4. ALK^ATI is transcribed from a genomically intact ALKlocus

a, Interphase FISH with ALK flanking probes demonstrates juxtaposed green and orange signals indicating no ALK rearrangement in MM-15. Scale bar, 10 μm. b, Interphase FISH in MM-74 shows 3 green/orange fusion signals in the majority of nuclei indicating a trisomy 2, but no ALK rearrangement. Scale bar, 10 μm. c, The top panel shows the genome-wide array CGH profile of MM-15 with numerous chromosomal gains and losses across the entire genome. The chromosomes are aligned along the x axis. The blue line illustrates the relative copy number (log₂ ratio) and the blue bars highlight copy number gains and losses. The middle panel illustrates the relative copy number (blue line) of chromosome 2. Distal to the ALK locus, a loss on the short (p) arm of chromosome 2 is indicated. The lower panel illustrates the relative copy number across the ALK locus. The red and green squares represent the log₂ ratio of individual array CGH probes (green, positive log₂ ratio; red, negative log₂ ratio). No disruption or selective gains or losses are found at the ALK locus. d, The genome-wide array CGH profile of MM-74 shows numerous chromosomal gains and losses across the entire genome in the top panel. The middle panel displays a relative copy number gain of the entire chromosome 2, which is in line with the trisomy of chromosome 2 as indicated by FISH. The lower panel also displays trisomy of chromosome 2, but indicates no focal gains and losses at the ALK locus.

Extended Data Figure 5. Targeted sequencing and whole-genome sequencing reveals no recurrent genomic aberrations at the ALK locus

a, Ultra-deep sequencing data of the ALK locus are displayed in IGV. The genomic region around intron 19 reveals several single nucleotide variations (SNVs). However, the vast majority of SNVs at the ALK locus are also found in the general population as they are detected in the pool of normal DNA, which was used as the control (pooled normal, bottom panel). Numerous SNVs are also documented in the Single Nucleotide Polymorphism database (dbSNP; http://www.ncbi.nlm.nih.gov/SNP/). No genomic aberrations were found at the transcription initiation site of ALK^ATI. Supplementary Table 2 shows the detected SNVs and indels at the ALK locus after filtering out SNPs documented in the dbSNP database. None of the genomic aberrations was found in more than one case, indicating that the expression of ALK^ATI is probably not caused by alterations of the DNA nucleotide sequence. b, c, Circos plots of the whole-genome sequencing data of MM-15 (b) and ATC-28 (c) illustrating numerous SNV and structural aberrations. Supplementary Table 3 lists the detected single nucleotide polymorphisms, and Supplementary Table 4 the detected structural aberrations. No recurrent genomic aberrations were found at the ALK locus.

Extended Data Figure 6. Local chromatin context at the alternative transcription initiation (ATI) site

a, UCSC Genome Browser view of the ATI site. The RepeatMasker track shows transposable elements at the ATI region, including a long-terminal repeat (LTR) in intron 19 (LTR16B2) and a long interspersed element (LINE) in intron 18. The ENCODE tracks reveal a DNase I hypersensitivity cluster and H3K4me1 enrichment, but no H3K27ac enrichment. b, The methylation status of the ALK locus was assessed by custom capture of the ALK locus, followed by bisulfite treatment and next-generation sequencing. Bisulfite sequencing results of H3122 (top) and MM-15 (bottom) are displayed in the CG-bisulfite mode of IGV. The red colour denotes ‘C’ (cytosine) corresponding to methylated cytosine, which is preserved during the bisulphite reaction. The blue colour denotes ‘T’ (thymine) corresponding to unmethylated cytosine, which is converted to uracil in the bisulfite reaction, and subsequently amplified to thymine during PCR. c, Methylation level at CpG sites in ALK^ATI-expressing tumour samples (MM-15 and ATC-28) and non-ALK^ATI-expressing control cells (H3122, a lung cancer cell line with EML4–ALK expression and SKMEL-28, a melanoma cell line without ALK^ATI expression). d, Comparison of the methylation status of CpG sites adjacent to the ATI site in ALK^ATI-expressing tumour samples (MM-15 and ATC-28) and non-ALK^ATI-expressing control cells (H3122 and SKMEL-28). The regions flanking LTR16B2 have significantly lower CpG methylation levels in ALK^ATI-expressing samples than controls; red dots indicate a statistically significant difference (P < 0.05; Mann–Whitney test) between ALK^ATI-expressing and non-expressing samples. Black dots indicate no statistically significant difference. e, ChIP-seq profile of H3K27ac at the ALK^ATI locus. The 17 blue profiles were retrieved from ENCODE, the 5 red profiles are original data from our lab. Only the 3 melanoma samples (MM-15, SKMEL-28, and SKMEL-524; bottom), but not the 19 non-melanoma cell lines, show H3K27ac enrichment at the ATI site. f, ChIP–qPCR validation for the H3K27ac enrichment at the ATI site in 6 melanoma cell lines. Error bars show mean ± s.e.m.; n = 3 technical replicates. g, Luciferase reporter assay of LTR16B2 in melanoma cell lines (red) and lung cancer cell lines expressing EML4–ALK (green). Error bars show mean ± s.d.;n = 9 (3 biological replicates combined from 3 independent experiments).

Extended Data Figure 7. ALK^ATI is active in vitro, shows nuclear and cytoplasmic localization by immunohistochemistry, and induces tumorigenesis

a, In vitro kinase assay. The indicated ALK variants were stably expressed in NIH-3T3 cells, immunoprecipitated, and assayed for tyrosine kinase activity. After the enzymatic reaction, the immunoprecipitated material was used for immunoblots to assess the amount of ALK protein used in the kinase assay. Error bars, mean ± s.d.; n = 4 technical replicates. b, Melanoma (MM-15) expressing ALK^ATI shows cytoplasmic and nuclear localization of ALK by immunohistochemistry. Melanocytic tumour expressing a TPM3–ALK translocation shows a cytoplasmic localization of the ALK fusion protein. Fibroblasts, epithelial cells, and reactive lymphocytes serve as internal negative controls. Scale bars, 100 μm. c, Flow cytometry analysis for green fluorescent protein (GFP) co-expressed from the same ALK-expression vector. Cells were cultured in IL-3-supplemented medium until day 0 (blue curve) and the number of GFP-positive cells was assessed. The number of GFP-positive ALK-expressing cells was assessed again 14 days after IL-3 withdrawal (red curve). d, Immunoblots of explanted NIH-3T3 tumour grafts expressing the indicated ALK isoforms. ALK^ATI was expressed at similar protein levels as in two ALK^ATI-expressing clinical human tumour samples. e, Growth curves of tumour grafts of melan-a cells stably expressing the indicated ALK isoforms in cohorts of 4–5 mice each with bilateral grafts. Error bars, mean ± s.e.m.; n = 8 tumours for ALK^F1174L, n = 10 tumours for all other experimental groups; see also Source Data associated with this figure. f, Immunoblots of explanted melan-a tumour grafts expressing the indicated ALK variants compared to ALK^ATI-expressing human tumour samples. g, Flow cytometry analysis of the GFP signal in NIH-3T3 cells stably expressing low (ALK^ATI-low) or high levels of ALK^ATI (ALK^ATI-high) before grafting into SCID mice. h, Immunoblot of t-ALK in ALK^ATI-low and ALK^ATI-high cells, confirming differential expression of ALK^ATI. See Supplementary Fig. 1 for uncropped blots for a, d, f and h. i, Growth curvesof tumour grafts of ALK^ATI-low and ALK^ATI-high cells. Error bars, mean ± s.e.m.; n = 10 tumours; see also Source Data associated with this figure.

Extended Data Figure 8. Concentration-dependent ALK inhibition in ALK^ATI-, wild-type ALK-, ALK^F1174L-, and EML4–ALK-expressing Ba/F3 cells

a, b, Cell viability assay of Ba/F3 cells, either in the presence or absence of IL-3 (1 ng ml⁻¹), expressing the indicated ALK isoforms and treated with the indicated doses of ALK inhibitors ceritinib (a) and TAE-684 (b). Cell viability was measured after 72 h of drug treatment. Error bars, mean ± s.e.m.; n = 3 biological replicates. c–e, Representative immunoblots of Ba/F3 cells stably expressing wild-type ALK (c), ALK^F1174L (d), or EML4–ALK (e) and treated with increasing concentration of crizotinib for 2 h. See Supplementary Fig. 1 for uncropped blots.

Extended Data Figure 9. Expression of wild-type ALK, ALK^F1174L, and EML4–ALK confers sensitivity to the ALK inhibitor crizotinib in vivo

a–c, Bioluminescence of luciferase-labelled NIH-3T3 grafted tumours expressing wild-type ALK (a), ALK^F1174L (b), or EML4–ALK (c) in SCID mice treated with either vehicle or crizotinib. Error bars, mean ± s.e.m.; n = 8 tumours; see also Source Data associated with this figure. d–f, H&E staining and immunohistochemistry of explanted tumours expressing wild-type ALK (d), ALK^F1174L (e), or EML4–ALK (f) 48 h after the first crizotinib treatment. Scale bar, 50 μm. g, MSK-IMPACT assay reveals copy number alterations and loss of CDKN2A and PTEN in melanoma metastasis MM-382, but no mutations. The log₂ ratio was calculated across all targeted regions by comparing the coverage in tumour versus normal. h, FISH for ALK shows no rearrangement; the 3 juxtaposed green/orange signals indicate a trisomy 2. Scale bar, 1 μm. i, The four FISH signals for MET and centromere 7 indicate a tetrasomy 7, but no MET amplification. Scale bar, 1μm.

Figure 1. Alternative transcription initiation (ATI) results in a novel ALK transcript

a, Distribution of RNA-seq reads of ALK variant transcripts: ALK^ATI RNA-seq reads align to both ALK intron 19 and exons 20–29; full-length, wild-type ALK (ALK^WT) RNA-seq reads align to all ALK exons, but not to the introns; translocated ALK RNA-seq reads align only to ALK exons 20–29. b, Mapping of the ATI sites of ALK^ATI to a 25 bp region in ALK intron 19 (hg19 ch2:29,446,768–29,446,744; blue shaded area). c, ChIP-seq profile of H3K4me3 at the ATI site. d, Quantitative mRNA profiling of different ALK variants using Nanostring nCounter: 2 wild-type ALK-expressing neuroblastoma cell lines (SK-N-BE2 and SK-N-DZ), 2 EML4–ALK-expressing lung cancer cell lines (H3122 and H2228), 9 ALK^ATI-expressing tumours (8 melanomas (MM) and 1 anaplastic thyroid carcinoma, ATC-28). e, Similar SNV frequencies in DNA-seq, RNA-seq, and ChIP-seq (H3K4me3) data indicate that ALK^ATI is biallelically expressed.

Figure 2. The ALK^ATI transcript encodes three ALK proteins mainly containing the kinase domain

a, Illustration of ALK protein isoforms. MAM, meprin, A-5 protein, and receptor protein-tyrosine phosphatase mu; Gly, glycine-rich region. b, Immunoblots of total (t) and phosphorylated (p) ALK in two ALK-expressing neuroblastoma cell lines, two EML4–ALK variant-expressing lung cancer cell lines, three ALK^ATI-expressing tumours, and a negative control melanoma cell line. c, Immunoblots of 293T cells with transient expression of ALK^ATI, in which the three predicted start codons were mutated from ATG to AAG, individually or in combination as indicated. KD, kinase-dead. d, Co-immunoprecipitation (IP) and immunoblots (IB) in 293T cells expressing V5-tagged ALK^ATI (V5–ALK^ATI), HA-tagged ALK^ATI (HA–ALK^ATI), or both. e, ALK immunofluorescence in NIH-3T3 cells expressing the indicated ALK isoforms. Scale bar, 25 mM. f, Haematoxylin and eosin (H&E) staining and ALK immunohistochemistry in ALK^ATI-expressing human tumour samples. Scale bar, 50 μm. See Supplementary Fig. 1 for uncropped blots from b–d.

Figure 3. ALK^ATI promotes growth-factor-independent proliferation in vitro and tumorigenesis in vivo

a, Growth curves of Ba/F3 cells stably expressing the indicated ALK isoforms in the absence of IL-3. Error bars, mean ± s.d.; n = 8, pooled data from 2 experiments with 4 technical replicates each. b, ALK immunoblots of previously transformed (capable of IL-3- independent growth) Ba/F3 cells with exogenous expression of the indicated ALK isoforms, and of tumours with endogenous ALK^ATI expression. See Supplementary Fig. 1 for uncropped blots. c, Tumour growth of NIH-3T3 cells stably expressing the indicated ALK isoforms. Error bars, mean ± s.e.m.; n = 10 tumours, see also Source Data associated with this figure.

Figure 4. ALK^ATI expression confers sensitivity to ALK inhibitors in vitro and in vivo

a, Dose–response curves to crizotinib of Ba/F3 cells expressing the indicated ALK isoforms in the presence or absence of IL-3. Error bars, mean ± s.e.m.; n = 3 biological replicates. b, Representative immunoblots of ALK^ATI-expressing Ba/F3 cells treated with increasing concentrations of crizotinib for 2 h. See Supplementary Fig. 1 for uncropped blots. c, Normalized tumour volume in mice implanted with NIH-3T3 cells expressing the indicated ALK isoforms and treated with vehicle (n = 8 tumours) or crizotinib (n = 10 tumours). Error bars, mean ± s.e.m.; see also Source Data. d, H&E staining and immunohistochemistry of explanted ALK^ATI-expressing tumours 48 h after first crizotinib treatment. e, Normalized bioluminescence signal of ALK^ATI-expressing, luciferase-labelled NIH-3T3 tumours treated with vehicle or crizotinib. Error bars, mean ± s.e.m.; n = 8 tumours; see also Source Data associated with this figure. f, Quantitative mRNA ALK profiling of a metastatic melanoma (MM-382) compared to wild-type ALK, EML4–ALK, or ALK^ATI using Nanostring nCounter. g, H&E staining and ALK immunohistochemistry (inset) of MM-382. Scale bars in d and g, 50 μm. h, Computed tomography images of a subcutaneous melanoma metastasis from patient 1 (MM-382) before and after crizotinib treatment.