Abstract
There is a pressing need for methods to define the functional relevance of genetic alterations identified by next-generation sequencing of cancer specimens. We developed new approaches to efficiently construct full-length cDNA libraries from small amounts of total RNA, screen for transforming and resistance phenotypes, and deconvolute by next-generation sequencing. Using this platform, we screened a panel of cDNA libraries from primary specimens and cell lines in cytokine-dependent murine Ba/F3 cells. We demonstrate that cDNA library-based screening can efficiently identify DNA and RNA alterations that confer either cytokine-independent proliferation or resistance to targeted inhibitors, including RNA alterations and intergenic fusions. Using barcoded next-generation sequencing, we simultaneously deconvoluted cytokine-independent clones recovered after transduction of 21 cDNA libraries. This approach identified multiple gain-of-function alleles, including KRAS G12D, NRAS Q61K and an activating splice variant of ERBB2. This approach has broad applicability for identifying transcripts that confer proliferation, resistance and other phenotypes in vitro and potentially in vivo.
Introduction
Thousands of somatic variants have recently been catalogued across multiple cancer types. However, the functional relevance of nearly all of these variant alleles remains undefined and only a small minority are likely to contribute to malignant phenotypes. Thus, approaches are broadly needed to distinguish which of the genetic variants present in both coding and noncoding RNA have functional relevance, and may therefore represent therapeutic targets [1].
Inhibition of oncogene function can induce cell death and tumor regression across a variety of cancers but is almost invariably associated with the selection of resistant clones. Multiple genetic and epigenetic mechanisms can mediate resistance to targeted inhibitors, including mutation of the therapeutic target, activation of downstream signaling pathways and upregulation of orthogonal pathways. Similar to the identification of oncogenes, delineating the multiple possible mechanisms of resistance to targeted inhibitors can be labor intensive and has primarily involved biased approaches that only interrogate protein-coding genes [2].
The development of methods for cloning full-length cDNA [3] and constructing cDNA libraries [4] in the early 1980s led to a series of phenotype-based screens that broadly interrogated the transcriptome. For example, the Rennick laboratory identified a cDNA for human GM-CSF by transfecting cDNA libraries derived from activated human T-cells into COS-7 monkey cells [5]. The Nolan laboratory used Ba/F3 cells to isolate clones that expressed selectable markers like hCD2 or hIL3R after retroviral transduction of cDNA libraries [6]. Ba/F3 cells are murine hematopoietic cells that are absolutely dependent on IL3 for survival. Expression of some oncogene transcripts can substitute for IL3 signaling, allowing for the selection of monoclonal populations that proliferate in the absence of IL3.
cDNA libraries have also been used to identify new oncogenes from tumor cell lines and primary samples. In 1994, Chan et al. identified a novel RAS superfamily member by screening an ovarian cancer cell line cDNA library in 3T3 cells. Nearly fifteen years later, the Mano laboratory used essentially the same protocol to identify the EML4-ALK fusion from a lung cancer sample [7]. We previously discovered a mutant allele of the CRLF2 cytokine receptor by screening cDNA libraries generated from primary acute lymphoblastic leukemias (ALL), which multiple groups concurrently identified as an important oncogene in poor-risk ALL [8], [9].
Screens of cDNA can identify alterations present within exonic DNA as well as splice alterations, RNA editing events and noncoding RNAs. However, standard approaches for generating and screening high-quality cDNA libraries have multiple shortcomings. First, up to 5 µg of mRNA may be required, which precludes the use of most primary tumor specimens. Second, first-strand synthesis may prematurely abort and thereby generate truncated cDNA that are not present within the sample. Third, the isolation and analysis of individual clones with the desired phenotype can be very cumbersome. Finally, target cells that have a background rate of spontaneous transformation (e.g. 3T3) can generate multiple false positive clones, further reducing the efficiency of cDNA library screening.
We developed new methods for efficiently generating high-quality libraries of full-length cDNA using ≤1 µg total RNA. We screened a panel of primary specimen and cell line libraries in Ba/F3 cells to identify transcripts with either DNA or RNA alterations that could substitute for IL3 signaling. To deconvolute the integrated cDNA from pools of clones, we developed methods for next-generation sequencing and bioinformatic analysis that greatly improve the efficiency of cDNA library-based screening. These approaches can be applied to screen both primary and immortalized samples for a variety of phenotypes.
Results and Discussion
To generate high-quality, full-length cDNA libraries, we created a ‘Full length-att’ protocol that hybridizes the In-Fusion SMARTer Directional cDNA Library Construction Kit (Clontech) with recombinase reactions (Invitrogen) for shuttling between vectors (Figure S1). We perform quality control of these libraries in three steps. First, libraries are transformed into E. coli to confirm the presence of ≥107 clones per library. Second, restriction enzyme digestion of each library is performed to demonstrate clonal heterogeneity (Figure S1). Third, we developed quantitative PCR (qPCR) approaches using primers that amplify the 5′ region of housekeeping genes to confirm that similar frequencies of long (e.g. REV3L, 10719 bp) and short (e.g. GAPDH: 1401 bp) full-length cDNAs are present (Figure S1). We generate libraries in medium-throughput fashion (i.e., batches of 12) and screen for IL3-independent growth and drug resistance phenotypes in Ba/F3 cells, as previously performed [10].
To validate this approach, we screened cDNA libraries generated from the lung cancer cell lines H3122 and H2228 (Figure 1), which both harbor EML4-ALK fusions. All IL3-independent Ba/F3 clones recovered after transduction of both libraries harbored full-length EML4-ALK and were sensitive to the ALK inhibitor crizotinib (LD99<1 µM). Multiple independent EML4-ALK transcripts were recovered from both libraries, based on the presence of different lengths of 5′ untranslated region (UTR), which results from variable elongation past the translation start site during first-strand cDNA synthesis (Figure 1A). Two distinct EML4-ALK fusion transcripts were identified from H2228, which differed based on alternate splicing of exon 6 (Figure 1A). Thus, this approach can efficiently recover multiple gain-of-function transcripts from the same tumor specimen.
To identify an unknown translocation fusion partner, we screened a cDNA library isolated from a biopsy of a primary anaplastic large cell lymphoma (ALCL) that expressed the ALK oncoprotein but did not have the common NPM-ALK rearrangement (Figure 2). We recovered multiple Ba/F3 clones that all contained and expressed a full-length ATIC-ALK fusion transcript (Figure 2B–C), which was previously described in a small subset of ALCLs [11]. We isolated 96 IL3-independent Ba/F3 clones after transduction of this library and tested the effects of the specific ALK inhibitor TAE684 on proliferation. Of the 96 clones, only clone D3 was unaffected by 1 µM TAE684 (Figure 2D). Sequencing of the integrated cDNA within D3 identified a full-length STAT5B transcript that harbored a K70N mutation in the N-terminal region (Figure 2E). Thus, clone D3 had acquired ALK-independent proliferation in the absence of IL3.
We also utilized cDNA library screening to recover alleles that confer resistance to targeted inhibitors. First, we generated the PC9-ER cell line by culturing PC9 lung cancer cells, which express the transforming EGFR delE745-A750 allele, in the presence of escalating doses of the EGFR inhibitor erlotinib (Figure 3A). As previously described, this resulted in subclonal outgrowth of PC9 cells harboring the gatekeeper mutation T790M, which can be difficult to discern by Sanger sequencing (Figure 3B) [12]. Screening of a cDNA library generated from PC9-ER in Ba/F3 cells resulted in multiple full-length clones that harbored EGFR with both delE745-A750 and T790M (Figure 3C) and were sensitive to the mutant-selective EGFR inhibitor WZ4002 (LD99<1 µM) [13].
Transduction of Ba/F3 cells at a high multiplicity of infection (MOI) allows for screening of >107 transcripts in medium-throughput fashion, but the IL3-independent Ba/F3 clones typically harbor 5–15 distinct cDNA inserts (Figure 1). To functionally deconvolute inserts from IL3-independent Ba/F3 clones, we developed methods for insert identification by next-generation sequencing. To test this strategy, we generated 21 cDNA libraries from primary specimens and cell lines and transduced each library into 5×106 Ba/F3 cells. After withdrawal of IL3, proliferating clones were subjected to PCR amplification of integrated cDNA, followed by barcoding by library, pooling and sequencing on the SOLiD 4 platform. Approximately 18 Gbp of sequence data was mapped to the human reference genome (hg18). As genes with functional relevance are typically present in multiple clones with different 5′ UTRs (Figure 1), we developed a bioinformatic pipeline for the detection of attB fusion sequences and precisely mapped the genomic coordinates of the integrated cDNA.
Among the 21 libraries were two positive controls, which were known to harbor activating RAS family alleles. Library transduction followed by PCR amplification of integrated cDNA from IL3-independent clones and next-generation sequencing successfully identified both KRAS G12V from the small cell lung carcinoma line SHP-77 and NRAS Q61K from the melanoma cell line SK-MEL-23 (Figure 4). From the latter, 12 independent fusions of attB sequence with the 5′ UTR of NRAS, representing full-length transcripts, were recovered in amplified fragments (Figure 4C).
NRAS was also the most abundant transcript in amplified fragments from the SK-MEL-23 library, with a fragments per kilobase of exon per million fragments mapped (FPKM) value of 61516. The next 7 most abundantly recovered genes (PKM2, TKT, ACTG1, RPL11, AP4M1, OAZ2, IMPDH2) are all housekeeping genes and were present at FPKM values between 10165–14711. From amplified fragments recovered from the SHP-77 library, KRAS was the 5th most abundant transcript (FPKM 30645), behind the non-coding transcript GAS5 (FPKM 56640) and the genes UBE2V2, CHMP5, RPL18 and RPL19 (FPKM 36338–48842). No variants were present in the coding sequence of any of these 4 genes. Thus, the identification of mutated transcripts expressed at high FPKM can highlight driver alterations with this method.
Next-generation sequencing-based deconvolution identified a splice variant of ERBB2 that lacks exon 16 (ERBB2Δ16) from the breast carcinoma line HCC1569 (Figure 5). Expression of ERBB2Δ16 results in resistance to the ERBB2-specific monoclonal antibody trastuzumab and constitutive activation of SRC [14]. ERBB2Δ16 was also recovered from the melanoma cell lines MeWo and 70 W (Figure 5C). Transcripts isolated from each line had different 5′ UTR sequences, indicating full-length transcripts and confirming that they derived from independent lines (Figure 5C). Although ERBB2Δ16 was detectable in MeWo and 70 W cells, it was expressed at levels 10–1000-fold lower than in the trastuzumab-resistant breast cancer cell lines HCC1569 and BT474 [14] by qPCR (Figure 5D, 5E). ERBB2Δ16 was not among the top 50 transcripts recovered by next-generation sequencing (based on FPKM). Finally, neither melanoma MeWo nor 70 W was sensitive to the EGFR/ERBB2 inhibitor lapatinib (Figure 5B). This suggests that ERBB2Δ16 is not a primary driver of proliferation in MeWo and 70 W cells but could potentially compensate upon targeted inhibition of other pathways.
In conclusion, cDNA library screening with deconvolution by next-generation sequencing can efficiently identify fusions, intragenic alterations and splice variants from primary specimens and cell lines in medium throughput. These approaches are amenable to screens both in vitro and potentially in vivo, either to agnostically interrogate cDNA libraries or to identify individual transcripts within selected pools for proliferation, resistance and other phenotypes.
Methods
Ethics statement
All specimens were acquired from patients at the Dana-Farber Cancer Institute. This study was approved by the Dana-Farber Cancer Institute Institutional Review Board (DFCI IRB #01-026). No research was conducted outside of the United States. All participants provided written consent under the aforementioned protocol.
Cell lines and media
Ba/F3 cells were obtained from RIKEN. SHP-77, HCC1569 and BT-474 were obtained from ATCC. H3122, H2228 and PC9 were previously reported [15]. MeWo, 70 W and SK-MEL-23 were previously reported [16], [17] and were a kind gift of Dr. David Fisher (Massachusetts General Hospital). All cells were cultured according to published conditions. Immunohistochemistry for ALK was performed as previously described [18].
RNA isolation and full-length-att cDNA library construction
For construction of full-length cDNA libraries, total RNA was purified using TRIzol reagent (invitrogen) or the RNeasy Mini Kit (Qiagen) followed by DNase I treatment according to the manufacturer's instructions. ≤1 ug total RNA was used for cDNA library construction using a “Full length-att” approach that hybridizes the In-Fusion SMARTer Directional cDNA Library Construction Kit (Clontech) with recombinase reactions (Invitrogen) for shuttling between vectors. The In-Fusion SMARTer Directional cDNA Library methodology was performed with the following modifications. PCR primers were designed to incorporate attB1 recombinase recognition sites during amplification of the SMARTer first-strand cDNA:
attB1_forward: 5′-GGGGACAACTTTGTACAAAAAAGTTGGAAGCAGTGGTATCAACGCAGAG-3′
attB2_reverse: 5′-CCGCACAACTTTGTACAAGAAAGTTGGGTCGGGGTACGATGAGACACCA-3′
Long distance-PCR was carried out with PrimeSTAR HS DNA polymerase (Takara) using the following PCR conditions: 95°C for 1 min, 15 cycles of 98°C for 10 sec, 55°C for 5 sec and 68°C for 10 min. After purification according to the manufacturer's instructions, 300 ng of ds cDNA was cloned into 300 ng of pDONR222 vector with Gateway BP Clonase II enzyme mix (Invitrogen). The retroviral Gateway destination vector pMSCVpuroATT was constructed by inserting Reading Frame Cassette A into the retroviral expression vector pMSCVpuro (Clontech) using the Gateway Vector Conversion System (Invitrogen). Libraries were transferred from the pDONR222 backbone to pMSCVpuroATT using Gateway LR Clonase (Invitrogen). After transformation into ElectroMAX DH10BT1 Phage Resistant E. coli (Invitrogen), each pMSCVpuroATT library was cultured at 30°C for 16 hour and purified with thePureLink HQ Midi Plasmid Purification Kit (invitrogen), as described [10].
For libraries constructed with the CloneMiner II cDNA construction kit (Invitrogen), high-quality mRNA was isolated from total RNA with the FastTrack2.0 mRNA Isolation Kit (Invitrogen) according to the manufacturer's instructions. Up to 5 µg mRNA was used according to the manufacturer's instructions. Recombination of attB-flanked cDNA was performed into the attP-containing pDONR222 vector to create a high titer entry library. This library was then transferred into pMSCVpuroATT, as above.
Library quality control
All libraries were confirmed to contain ≥107 clones by transformation into E.coli and heterogeneity was confirmed by BsrGI digest (Figure S1). Adequate representation of long full-length cDNA was confirmed within cDNA libraries using qPCR with primers that amplified the 5′ region of long (TFRC: 5241 bp, POLQ: 8787 bp, REV3L: 10719 bp) and short (GAPDH: 1401 bp, ACTB: 1852 bp) housekeeping genes.
GAPDH-F: 5′-GTCAGCCGCATCTTCTTTTG-3′; 5′-ACGACCAAATCCGTTGACTC-3′
ACTB: 5′-GATGCAGAAGGAGATCACTGC-3′; 5′-TGATCCACATCTGCTGGAAG-3′
TFRC #1: 5′-AGTGATTGTCAGAGCAGGGAAA-3′; 5′-CCCAGATGAGCATGTCCAAA-3′
TFRC #2: 5′-GAGTGTGAGAGACTGGCAGGAA-3′; 5′-CGGTGAAGTCTGTGCTGTCC-3′
POLQ #1: 5′-GGCAGCACCTCTCCATCAA-3′; 5′-TCATCCACAACCACCATTCC-3′
POLQ #2: 5′-TGTGGCTTCCTGGTTGAATG-3′; 5′-GGGCTCAAATTCCCTCACAA-3′
REV3L #1: 5′-AAGTGGATGCTGTAGCTGCTGA-3′; 5′-GCCGTTGCTTTTCATCTTCC-3′
REV3L #2: 5′-TCTGGCTGCTGTCAAGTTCC-3′; 5′-TCATCTTGTTCCCACCGAAA-3′
Retroviral cDNA library screening
293T cells (7×105 cells in 2 mL of Dulbecco's Modified Eagle Medium supplemented with 10% fetal calf serum) were inoculated into a single well of a 6-well plate. After 24 hours, 3 µg of pMSCVpuroATT cDNA library and 1 µg of pEcoPack vector were co-transfected into 293T cells with 10 µL of Lipofectamine2000 transfection reagent (invitrogen) and 500 µL of Opti-MEM I Reduced Serum Media (Invitrogen). On the following day, 2.5 mL of fresh medium was added to the cells. After an additional day, the supernatant was passed through a 0.45 µm filter. Before the infection, sub-confluent Ba/F3 cells were diluted 100-fold and then cultured for 3 days in log-growth phase using RPMI-1640 medium supplemented with 10% FCS, 0.5 ng/mL of mouse IL3 (Prospec), penicillin and streptomycin. 2.5 mL of viral supernatant was applied to 5×106 Ba/F3 cells in 2.5 mL, along with 16 µg/mL of polybrene, 10 mM of HEPES-HCL and 0.5 ng/mL of mouse IL3 in a single well of a 6-well plate. The cells were centrifuged at 2500 rpm for 90 min at 37°C. Infected cells were washed once with media containing IL3 on the next day and 1 µg/mL of puromycin was added to the cells 2 days after infection. 48 hours later (4 days after infection), the cells were washed 3 times and 3–4×104 cells were plated into a 96-well plate with puromycin containing media without mouse IL3. After 10–28 days, visible colonies were isolated. Alternatively, cells were washed 96 hours after infection and transferred to a flask containing media without IL3. Proliferating cells were visible 21–28 days after infection. In our previous screening [10], infection involved up to 3–6×107 Ba/F3 cells at the same density in multiple plates.
Identification of integrated cDNA
To determine integrated cDNA sequences from IL3-independent Ba/F3 clones, genomic DNA was purified using the QIAamp DNA Blood Mini Kit (Qiagen) and used as template for a step-down PCR with primers flanking the inserted cDNA sequences:
att-f: 5′-TCCTCCCTTTATCCAGCCCTCACTCCTTCTCTAGG-3′
att-r: 5′-CTAAAGCGCATGCTCCAGACTGCCTTGGGAAAAGC-3′
PCR was conducted with KOD-FX DNA polymerase (Toyobo) using the following PCR conditions: 94°C for 4 min, 5 cycles of 98°C for 10 sec and 74°C for 8 min, 5 cycles of 98°C for 10 sec and 72°C for 8 min, 5 cycles of 98°C for 10 sec and 70°C for 8 min, 15 cycles of 98°C for 10 sec and 68°C for 8 min, and 68°C for 10 min. For the isolation of individual integrants, 25 µL of PCR product was separated by agarose gel electrophoresis and bands were purified from the gel using the QIAquick Gel Extraction Kit (Qiagen) and sequenced from the 5′ end by the Sanger method (Dana-Farber Cancer Institute Molecular Biology Core Facility).
For deep sequencing, PCR products amplified from independent Ba/F3 transformants within the same cDNA library were mixed at equimolar amounts (1–10 µL) and purified using the QIAquick PCR Purification Kit (Qiagen).
Establishment of erlotinib resistant PC-9 cell line and inhibitor assays
PC9 cells were initially treated with 1 nM erlotinib and the concentration was increased to 3 nM, 10 nM, 100 nM and 1 µM as clones emerged over the course of approximately 2 months. To confirm resistance to erlotinib among the surviving PC9-ER cells, a dose response study was performed using the CellTiter-Glo® Luminescent Cell Viability Assay (Promega) and read by the 2104 EnVision® Multilabel Reader (PerkinElmer) after 48 hour treatment. Dose response to lapatinib was also quantified using CellTiter-Glo® after 48 hour treatment, as above.
SOLiD library construction and sequencing
A sequencing library was generated from PCR products which were amplified from transfected DNA from each of 21 specimens. An average of approximately 10 IL3-independent BaF3 clones were derived from each screened library. We estimated that cDNA inserts would average 2 kb and that each BaF3 clone contains an average of 7 inserts, for a total of approximately 3×106 bases of cDNA sequence PCR amplified from all 21 specimens. Sequencing of 1010 bases would thus generate greater than 3000-fold mean coverage which may be necessary to confidently recover larger and/or repetitive cDNA with lower PCR amplification efficiency.
The method used to generate the libraries is described in the Applied Biosystems SOLiD 4 Library Preparation guide. Briefly, for each library, the PCR products were concatatenated by ligation and then sheared. An adaptor was ligated to each end of each sheared fragment. One of the two adaptors contained a barcode sequence, which was unique for each library. Each library was amplified by 6–8 cycles of PCR. After quantification, the libraries were mixed in equimolar amounts and then subjected to emulsion PCR. The beads resulting from the emulsion were sequenced from each end - 75 bases from one end and 35 bases from the other end, on 4 lanes using an Applied Biosystems 5500 sequencer. The libraries were binned according to barcode sequence.
Mapping and counting
Sequence data were mapped to the human reference genome, hg18, using the Genomic Resequencing and Whole Transcriptome Mapping modules contained in Lifescope 2.5 (Applied Biosystems.) The Lifescope 2.5 Exon Counts module was used in conjunction with a custom Python script to generate gene-level data, indentifying those transcripts which were present at a high level in each sample.
Gateway Junction Detection
To identify the precise genomic coordinates of integrated cDNA, a pipeline for the detection of gateway fusion sequences was developed. Individual reads containing a piece of the Gateway attB sequence (5′-TTGTACAAAAAAGTTGG′-3′) were pulled from the Whole Transcriptome mapped data. Reads were selected if their color-space sequence matched the color-space sequence of the gateway, either forward, (5′-T0113110000021010-3′) or in reverse (5′-G0101200000113110-3′). A python script was developed to identify and report the precise coordinate of the gateway fusion for each selected read, taking into account forward vs. reverse strandedness, clipped mappings, and mappings crossing exon junctions. These gateway fusion coordinates were collected and reported at the gene level for each of the 21 multiplexed samples.
Variant Analyses
Single Nucleotide Variants were called using DiBayes (Applied Biosystems) and annotated with information including exonic context (using RefSeq), known polymorphisms (using DBSNP 130), amino acid consequence, and frequency and quality of reads supporting each variant. Fusions genes were detected using the output of Lifescope's Splice Finder module and annotated against RefSeq using a Python script.
Supporting Information
Funding Statement
This work was supported by Stand Up to Cancer Innovative Research Grant SU2C-AACR-IRG0409 and American Cancer Society Research Scholar Grant RSG-10-181-01-RMC to DMW. NS was partially supported by Astellas Pharma Inc. OW was partially supported by the Deutsche Forschungsgemeinschaft. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
References
- 1. Salmena L, Poliseno L, Tay Y, Kats L, Pandolfi PP (2011) A ceRNA hypothesis: the Rosetta Stone of a hidden RNA language? Cell 146: 353–358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Garraway LA, Janne PA (2012) Circumventing cancer drug resistance in the era of personalized medicine. Cancer Discov 2: 214–226. [DOI] [PubMed] [Google Scholar]
- 3. Okayama H, Berg P (1982) High-efficiency cloning of full-length cDNA. Mol Cell Biol 2: 161–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Gubler U, Hoffman BJ (1983) A simple and very efficient method for generating cDNA libraries. Gene 25: 263–269. [DOI] [PubMed] [Google Scholar]
- 5. Lee F, Yokota T, Otsuka T, Gemmell L, Larson N, et al. (1985) Isolation of cDNA for a human granulocyte-macrophage colony-stimulating factor by functional expression in mammalian cells. Proc Natl Acad Sci U S A 82: 4360–4364. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Kitamura T, Onishi M, Kinoshita S, Shibuya A, Miyajima A, et al. (1995) Efficient screening of retroviral cDNA expression libraries. Proc Natl Acad Sci U S A 92: 9146–9150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Soda M, Choi YL, Enomoto M, Takada S, Yamashita Y, et al. (2007) Identification of the transforming EML4-ALK fusion gene in non-small-cell lung cancer. Nature 448: 561–566. [DOI] [PubMed] [Google Scholar]
- 8. Harvey RC, Mullighan CG, Chen IM, Wharton W, Mikhail FM, et al. (2010) Rearrangement of CRLF2 is associated with mutation of JAK kinases, alteration of IKZF1, Hispanic/Latino ethnicity, and a poor outcome in pediatric B-progenitor acute lymphoblastic leukemia. Blood 115: 5312–5321. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Hertzberg L, Vendramini E, Ganmore I, Cazzaniga G, Schmitz M, et al. (2010) Down syndrome acute lymphoblastic leukemia, a highly heterogeneous disease in which aberrant expression of CRLF2 is associated with mutated JAK2: a report from the International BFM Study Group. Blood 115: 1006–1017. [DOI] [PubMed] [Google Scholar]
- 10. Yoda A, Yoda Y, Chiaretti S, Bar-Natan M, Mani K, et al. (2010) Functional screening identifies CRLF2 in precursor B-cell acute lymphoblastic leukemia. Proc Natl Acad Sci U S A 107: 252–257. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Colleoni GW, Bridge JA, Garicochea B, Liu J, Filippa DA, et al. (2000) ATIC-ALK: A novel variant ALK gene fusion in anaplastic large cell lymphoma resulting from the recurrent cryptic chromosomal inversion, inv(2)(p23q35). Am J Pathol 156: 781–789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Rho JK, Choi YJ, Lee JK, Ryoo BY, Na II, et al. (2009) The role of MET activation in determining the sensitivity to epidermal growth factor receptor tyrosine kinase inhibitors. Mol Cancer Res 7: 1736–1743. [DOI] [PubMed] [Google Scholar]
- 13. Zhou W, Ercan D, Chen L, Yun CH, Li D, et al. (2009) Novel mutant-selective EGFR kinase inhibitors against EGFR T790M. Nature 462: 1070–1074. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Mitra D, Brumlik MJ, Okamgba SU, Zhu Y, Duplessis TT, et al. (2009) An oncogenic isoform of HER2 associated with locally disseminated breast cancer and trastuzumab resistance. Mol Cancer Ther 8: 2152–2162. [DOI] [PubMed] [Google Scholar]
- 15. Koivunen JP, Mermel C, Zejnullahu K, Murphy C, Lifshits E, et al. (2008) EML4-ALK fusion gene and efficacy of an ALK kinase inhibitor in lung cancer. Clin Cancer Res 14: 4275–4283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Houghton AN, Real FX, Davis LJ, Cordon-Cardo C, Old LJ (1987) Phenotypic heterogeneity of melanoma. Relation to the differentiation program of melanoma cells. J Exp Med 165: 812–829. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kerbel RS, Man MS (1984) Single-step selection of unique human melanoma variants displaying unusually aggressive metastatic behavior in nude athymic mice. Invasion Metastasis 4 Suppl 1: 31–43. [PubMed] [Google Scholar]
- 18. Rodig SJ, Mino-Kenudson M, Dacic S, Yeap BY, Shaw A, et al. (2009) Unique clinicopathologic features characterize ALK-rearranged lung adenocarcinoma in the western population. Clin Cancer Res 15: 5216–5223. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.