Aberrant signaling through cytokine receptors and their downstream signaling pathways is a major oncogenic mechanism underlying hematopoietic malignancies. To better understand how these pathways become pathologically activated and to potentially identify new drivers of hematopoietic cancers, we developed a high-throughput functional screening approach using ex vivo mutagenesis with the Sleeping Beauty transposon. We analyzed over 1,100 transposon-mutagenized pools of Ba/F3 cells, an IL3-dependent pro-B-cell line, which acquired cytokine independence and tumor-forming ability. Recurrent transposon insertions could be mapped to genes in the JAK/STAT and MAPK pathways, confirming the ability of this strategy to identify known oncogenic components of cytokine signaling pathways. In addition, recurrent insertions were identified in a large set of genes that have been found to be mutated in leukemia or associated with survival, but were not previously linked to the JAK/STAT or MAPK pathways nor shown to functionally contribute to leukemogenesis. Forced expression of these novel genes resulted in IL3-independent growth in vitro and tumorigenesis in vivo, validating this mutagenesis-based approach for identifying new genes that promote cytokine signaling and leukemogenesis. Therefore, our findings provide a broadly applicable approach for classifying functionally relevant genes in diverse malignancies and offer new insights into the impact of cytokine signaling on leukemia development. Cancer Res; 76(4); 773–86. ©2015 AACR.

Cytokine signaling is known to regulate the growth and development of hematopoietic cell populations (1–3) and it is well established that aberrant signaling through cytokine receptors and downstream signaling pathways is an underlying cause of hematopoietic malignancies (4, 5). Accordingly, there is a strong precedence for using growth factor independence as a surrogate for transforming potential in leukemia. For example, genes that confer cytokine independence in murine bone marrow–derived Ba/F3 cells, which are normally dependent on IL3 for proliferation and viability, often exhibit broader transforming activity (6, 7). In the Ba/F3 system, constitutive activation of the established oncogenes FLT3, ALK, or JAK3, or expression of TEL-JAK or EBF-PDGFRB gene fusions confers IL3-independent growth (8–11). Ba/F3 cells have also proven to be useful for validating oncogenes such as BCR-ABL and for characterizing the transforming potential of mutated kinases (12, 13). Collectively, these studies have demonstrated how the functional analysis of genes in these cells informs our understanding of mechanisms of leukemogenesis. However, a high-throughput analysis of genes that promote cytokine independence and transformation in this system has not been previously performed. Therefore, a comprehensive analysis of genes that drive growth factor independence and malignant transformation in Ba/F3 cells may reveal novel genes that promote signaling through the IL3 pathway and that participate more broadly in hematopoietic malignancies and other cancers.

Although recent genome sequencing efforts have identified numerous genetic lesions in leukemias, a large fraction of the recurrently mutated genes still require functional validation. Transposon-mediated mutagenesis has emerged as a powerful methodology for functionally annotating cancer genomes. A number of studies have utilized transposon-based insertional mutagenesis screens in mice to identify and validate a number of genes relevant to tumorigenesis, including T-cell leukemia, hepatocellular carcinoma, medulloblastoma, and nerve sheath tumors (14–21). To date, a vast majority of studies have relied upon the mobilization of mutagenic DNA transposons such as Sleeping Beauty (SB) or piggyBac in transgenic mouse models. When mobilized in the genome, these elements disrupt normal gene function by activating proto-oncogenes via a strong viral promoter, or by inactivating tumor suppressor genes via a transcriptional stop cassette. In these studies, the transposon is mobilized ubiquitously or in a tissue-specific manner at sufficient frequencies to induce tumors alone or in cooperation with an initiating event such as oncogenic activation of Myc, Braf, Npm, or genetic loss of the Tp53 or Ptch tumor suppressors (20, 22, 23).

Although in vivo transposon-mediated forward genetic screens have proven to be effective for cancer gene identification, they are also resource intensive. To facilitate the rapid and cost-effective identification of genes that regulate tumor-promoting pathways, we have developed a complementary ex vivo transposon mutagenesis approach wherein cells growing in culture are mutagenized and then screened for the acquisition of specific phenotypes in vitro or in vivo, such as growth factor independence or tumor-forming ability. Transposon insertion sites associated with these phenotypes are subsequently recovered. In this study, we describe the application of this system to comprehensively identify and validate genes that promote growth factor independence and transformation in Ba/F3 cells and provide evidence that these newly identified genes play important roles in human leukemogenesis and/or lymphomagenesis.

Cell culture

Ba/F3 cells were cultured in RPMI1640 and supplemented with 10% FBS (Invitrogen), penicillin/streptomycin, and 10 ng/mL mouse IL3 (Cell Signaling Technology).

Mice

The immunocompromised mice NOD.Cg-Prkdcscid TIl2rgtm1Wjl/SzJ (NSG mice) were purchased from The Jackson Laboratories. All procedures involving mice were approved by the Institutional Animal Care and Use Committee of University of Texas Southwestern Medical Center (Dallas, TX).

Sleeping Beauty mutagenesis

Ba/F3 cells cultured in the presence of IL3 were nucleofected (Lonza) with T2/Onc and SB11 or SB100X transposase plasmids, and split immediately into 24- or 48-well plates. As negative controls, cells were transfected with a single T2/Onc or SB plasmid. Seventy-two hours later, IL3 was withdrawn from media. Cells were then cultured for 1 to 4 weeks. Genomic DNA was isolated from cells that grew in the absence of IL3.

IL3 ELISA

Cell cultures were collected three days after changing media, centrifuged, and filtered to remove cells and debris. IL3 concentrations were detected using Mouse IL-3 Quantikine ELISA Kit from R&D Systems following the manufacturer's instructions.

Identification of transposon insertion sites

The junctions of the SB ends and mouse genomic DNA were amplified using ligation-mediated PCR (24). Briefly, samples of genomic DNA were digested with BfaI and ligated to linkers. Next, the ligation products were digested with BamHI and amplified by PCR with primers containing sequences of SB left end and linker; then nested PCR was performed with barcoded primers. The PCR products were purified and sequenced using the Illumina HiSeq2000. The raw sequences were screened for barcodes and the SB end; then the SB sequence and linker sequence were trimmed and the genomic sequences were aligned to the mouse genome (mm10) using Bowtie. The alignment results were filtered to obtain the unique matches. The insertions were further filtered according to their frequency of occurrence. Any insertions with frequencies below 0.01% in their libraries were removed. Finally, the insertions from different libraries were pooled together and the insertions at same chromosome, coordinate, and orientation were collapsed to obtain nonredundant insertions. All primer sequences are available upon request.

Identification of common insertion sites

The common insertion sites (CIS) were determined by two different analyses.

Monte Carlo simulation.

The SB insertions were simulated by randomly random distribution at TA sites throughout the mouse genome and repeated 1,000 times. Then the following “insertions in window size (bp)” were considered as significant (P < 0.01): 2 in 18 bp, 3 in 297 bp, 4 in 958 bp, 5 in 1,895 bp, 6 in 3,029 bp, 7 in 4,309 bp, 8 in 5,698 bp, 9 in 7,175 bp, 10 in 8,723 bp, and 11 in 10,330 bp. The CISs that overlapped with each other were merged and the P values were determined by comparing the insertion density in the CISs and the average insertion density in the entire genome using Pearson χ2 test.

Gene-centric CIS analysis.

For each of the genes in the mouse genome, the entire transcribed units and the region 10 kb upstream of the transcription sites were considered. The insertion densities were calculated by the formula: Density = insertions in the gene/TAs in the gene/2. Then, the densities were compared with the average density of the entire genome to determine the P values using Pearson χ2 test.

Western blotting

Approximately 30 mg of spleen or subcutaneous tumor tissue was homogenized with a Tissue Tearor homogenizer in RIPA buffer (1% v/v NP40, 0.1% w/v SDS, 0.5% w/v sodium deoxycholate in PBS) including phosphatase inhibitor cocktails 2 and 3 (Sigma, P5726 and P0044) diluted 1:100 and protease inhibitor cocktail (Sigma, P8340) diluted 1:50. Protein concentrations were determined by BCA assay (Thermo Scientific, 23228 and 23224). Twenty-five to 30 μg of protein lysate was loaded into each well of a NuPAGE 4%–12% gradient Bis-Tris gel (Life Technologies, NP0335), electrophoresed, and transferred to nitrocellulose. The membranes were incubated overnight at 4°C with rabbit mAbs against either Braf phospho-S729 (Abcam, EPR2207) or GHR [Abcam, EPR5291(2)] diluted 1:1,000 v/v in blocking buffer. HRP-conjugated anti-rabbit IgG secondary antibody (Bio-Rad, 170-6515) diluted 1:4,000 v/v for blots probed with anti-phospho-Braf or 1:10,000 for blots probed with anti-GHR. Membranes were stripped and reprobed with a rabbit mAb against β-actin (Cell Signaling Technology, 8457S) diluted 1:1,000 v/v in blocking buffer. Membranes were developed using SuperSignal West Dura Chemiluminescent Substrate (Thermo Scientific, 37046).

Consensus clustering analysis

Acute myelogenous leukemia (AML) data were downloaded from The Cancer Genome Atlas Research (TCGA, 2013). Acute lymphoblastic leukemia (ALL) data were downloaded from the International Cancer Genome Consortium (ICGC; http://dcc.icgc.org/repository/release_14/ALL-US). To identify patient subgroups, we performed a consensus clustering method with K-means clustering (25). For each k ∈ K = {2, …, 8}, where k represents the number of subgroups in the samples, we ran a K-means algorithm with selected genes on the randomly selected 80% of the samples for 500 times, and then constructed using a patient consensus matrix. On the basis of the consensus matrix CDFs (25) and the visual inspection of the consensus matrices, we chose the optimal subgrouping among K and performed Kaplan–Meier survival analysis. To generate P values, a log-rank test was implemented in R with the survival package. For survival analysis of individual genes, we compared survival of patients with high expression (greater or equal to 75%) versus low expression (greater or equal to 25%) of each gene. A log-rank tested was implemented in R using the survival package.

Expression of candidate genes in Ba/F3 cells using SB system

Full-length or truncated candidate genes were cloned into modified T2 vector (17) downstream of the CAG promoter and a V5 tag using GateWay cloning. The GateWay cassette is followed by a puromycin-IRES-GFP cassette. The entire cassette is flanked by the Sleeping Beauty inverted repeats, IRDR-L and IRDR-R, which is recognized by the SB transposase and then cut and pasted to the host cell DNA. The plasmids were then cotransfected with SB100X plasmid into Ba/F3 cells using the Lonza 4D Nucleofection System (solution SF, program DS-137). Two days later, puromycin was added to the medium at a final concentration of 1 μg/mL to select for stable expression. cDNAs were purchased from Invitrogen Ultimate ORF collection. All cDNAs were human except for Spidr, 4930468A15Riken, and Usp32, which were mouse.

Growth rate assays

Cells were plated 5 × 104 or 1 × 105/mL in the presence or absence of IL3. Cell numbers were counted under the microscope using a hemocytometer (for experiments in Fig. 1B, Supplementary Figs. S8 and S9). For in vitro validation studies shown in Fig. 4B, cell viabilities were determined using CellTiter 96 AQueous nonradioactive cell proliferation assay (MTS) according to the manufacturer's instructions (Promega).

Figure 1.

Transposon mutagenesis promotes IL3 independence and leukemogenesis. A, overview of the ex vivo mutagenesis strategy. B, cell proliferation assay demonstrating that SB mutagenesis results in IL3-independent growth, whereas unmodified Ba/F3 (WT) cells cannot proliferate in the absence of IL3. C, IL3 ELISA assay on mutagenized Ba/F3 pools. D, survival analysis of NSG mice transplanted with WT Ba/F3 cells, or SB mutagenized cells injected subcutaneously (SC) or intravenously(IV). E, spleen weight of NSG mice injected with SB mutagenized cells (intravenously and subcutaneously). P = 2.614E−05 and 0 for intravenous and subcutaneous, respectively, compared with controls (Student t test). F, liver weight of NSG mice injected with SB-mutagenized cells (intravenous and subcutaneous). P = 1.255E−10 and 2.614E−05, respectively, compared with controls. G, representative gross images of liver and spleen from NSG mice injected with SB-mutagenized or control Ba/F3 cells. H, H&E-stained sections of spleen, liver, and kidney of mice injected with control or mutagenized Ba/F3 cells. Scale bar, 50 μm.

Figure 1.

Transposon mutagenesis promotes IL3 independence and leukemogenesis. A, overview of the ex vivo mutagenesis strategy. B, cell proliferation assay demonstrating that SB mutagenesis results in IL3-independent growth, whereas unmodified Ba/F3 (WT) cells cannot proliferate in the absence of IL3. C, IL3 ELISA assay on mutagenized Ba/F3 pools. D, survival analysis of NSG mice transplanted with WT Ba/F3 cells, or SB mutagenized cells injected subcutaneously (SC) or intravenously(IV). E, spleen weight of NSG mice injected with SB mutagenized cells (intravenously and subcutaneously). P = 2.614E−05 and 0 for intravenous and subcutaneous, respectively, compared with controls (Student t test). F, liver weight of NSG mice injected with SB-mutagenized cells (intravenous and subcutaneous). P = 1.255E−10 and 2.614E−05, respectively, compared with controls. G, representative gross images of liver and spleen from NSG mice injected with SB-mutagenized or control Ba/F3 cells. H, H&E-stained sections of spleen, liver, and kidney of mice injected with control or mutagenized Ba/F3 cells. Scale bar, 50 μm.

Close modal

JAK inhibitor and MEK inhibitor assays

The JAK inhibitor ruxolitinib was purchased from Selleck Chemicals. MEK inhibitor AZD6244 was purchased from Thermo Fisher Scientific. Cells were plated at 5 × 104/mL in the absence or presence of inhibitors (1 μmol/L ruxolitinib or 10 μmol/L AZD6244) and cultured for 5 days. Cell numbers were counted under the microscope.

In vivo validation assay for leukemogenesis

A total of 2 × 106 IL3-independent cells overexpressing candidate genes or control cells with an empty GFP vector were pelleted, resuspended in PBS solution, and injected into the tail vein of NSG mice. Mice were dissected after or at the time of death. The livers and spleens were harvested for histologic analysis using hematoxylin and eosin (H&E) staining. Bone marrow cells were collected from thigh bones and analyzed for GFP expression using flow cytometry with a BD Accuri C6 flow cytometer.

Computer languages

The scripts for screening the Illumina raw sequences, filtering Bowtie outputs, determining CIS and other bioinformatic analyses were written in Perl language. The vector images of insertion profiles were generated with a Perl script. The raw outputs are in PostScript format, and then converted to .pdf format.

Establishment of an ex vivo SB mutagenesis system

To identify genes that promote IL3 independence and B-cell transformation, we transfected immortalized Ba/F3 cells with the T2/Onc transposon and SB11 or SB100 transposase plasmids, alone or in combination (Fig. 1A). After 72 hours in culture, IL3 was withdrawn from media and the cells were assayed for IL3-independent cell growth. Cotransfection with the transposon and transposase plasmids, but not the individual constructs, resulted in the outgrowth of IL3-independent pools in approximately 20% of transfections after one to several weeks (Fig. 1B). An ELISA assay further confirmed that the majority of mutagenized pools did not produce IL3 and were therefore growth factor independent (Fig. 1C). Although occasional transfectants yielded low levels of IL3, below 45 pg/mL, we confirmed that this level of IL3 production was not sufficient to support logarithmic growth of Ba/F3 cells (Supplementary Fig. S1A). High-throughput sequencing of transposon insertion sites (described in greater detail below) confirmed that these pools harbored SB insertions upstream of the IL3 precursor locus. We speculate that additional mutations in these low-level IL3-producing pools endow cells with the ability to proliferate in the setting of a limiting concentration of growth factor. Importantly, both mutations that confer complete growth factor independence as well as mutations that significantly reduce the threshold of IL3 needed to sustain growth may be relevant to B-cell transformation.

We next transplanted mutagenized IL3-independent Ba/F3 pools subcutaneously or by intravenous tail vein injection into immunocompromised NOD/SCID IL2Rgnull (NSG) mice. Both methods of delivery robustly yielded subcutaneous tumors or leukemias/lymphomas, respectively, within 2–8 weeks (Fig. 1D and Supplementary Fig. S1B), a dramatic increase in tumor penetrance and greatly shortened latency relative to mice injected with nonmutated Ba/F3 cells. Injection with SB-mutagenized cells also significantly reduced the survival of NSG mice, irrespective of whether mutagenized cells were injected intravenously or subcutaneously, as compared with animals transplanted with WT Ba/F3 cells (Fig. 1D). The spleen and liver were consistently enlarged in both cohorts of mutagenized animals relative to controls (Fig. 1E–G) and comparable in size with organs from animals transplanted with Ba/F3 cells overexpressing Bcr-Abl, which served as a positive control (Supplementary Fig. S1C). Histologic analysis of tissues using H&E staining revealed effacement of the normal splenic architecture and invasion of lymphocytes into multiple organs including the spleen, liver, and kidney (Fig. 1H).

To identify the transposon insertions that resulted in oncogenic transformation of mutagenized Ba/F3 cells in these initial pilot studies, we performed ligation-mediated PCR and high-throughput sequencing of 16 IL3-independent pools and 20 tumors derived from the transplanted pools. This revealed insertions in a number of key genes previously associated with hematopoietic malignancy including several genes in the Ras and JAK/STAT signaling pathways. For example, we identified SB insertions in intron 11 of Braf that resulted in a 10-fold upregulation of Braf mRNA (Fig. 2A and B). On the basis of the orientation of the T2/Onc transposon relative to the start of transcription, we predicted that this SB insertion drove expression of a truncated protein with an intact kinase domain at its C-terminus. Western blotting confirmed this prediction, demonstrating overexpression of a truncated 37 kDa phospho-BRAF in pool SB3.1, but not in other mutagenized pools lacking the intronic SB insertion (Fig. 2C, top and bottom).

Figure 2.

Analysis of top candidates identified through SB mutagenesis. A, schematic representation of SB insertions in intron 11 of Braf (top) and intron 1 of Ghr. Black triangles depict unique SB insertions present in the sense orientation relative to the orientation of transcription for each gene. B, quantitative RT-PCR analysis of Braf mRNA expression in growth factor–independent cells and tumors. Error bars, SDs from three independent measurements. C, Western blot analysis confirms overexpression of full-length GHR and truncated 37 kDa phospho-BRAF in cell pools/tumors harboring SB insertions. IV, intravenous (tumor tissue isolated from spleen); SC, subcutaneous tumor. D, Venn diagram depicting the overlap of top candidates identified through MC-CIS analysis or gCIS analysis. Values indicate number of the top 163 CISs identified as statistically significant using one or both methods of analysis. Red asterisks, known oncogenes and tumor suppressors. E, SB insertion profiles in known oncogenes and tumor suppressors. Green triangles, unique SB insertions present in the sense orientation relative to the orientation of gene transcription; red triangles, SB insertions on the antisense strand. Black arrows, transcription start site. 100 kb is shown flanking the 5′ and 3′ end of each gene.

Figure 2.

Analysis of top candidates identified through SB mutagenesis. A, schematic representation of SB insertions in intron 11 of Braf (top) and intron 1 of Ghr. Black triangles depict unique SB insertions present in the sense orientation relative to the orientation of transcription for each gene. B, quantitative RT-PCR analysis of Braf mRNA expression in growth factor–independent cells and tumors. Error bars, SDs from three independent measurements. C, Western blot analysis confirms overexpression of full-length GHR and truncated 37 kDa phospho-BRAF in cell pools/tumors harboring SB insertions. IV, intravenous (tumor tissue isolated from spleen); SC, subcutaneous tumor. D, Venn diagram depicting the overlap of top candidates identified through MC-CIS analysis or gCIS analysis. Values indicate number of the top 163 CISs identified as statistically significant using one or both methods of analysis. Red asterisks, known oncogenes and tumor suppressors. E, SB insertion profiles in known oncogenes and tumor suppressors. Green triangles, unique SB insertions present in the sense orientation relative to the orientation of gene transcription; red triangles, SB insertions on the antisense strand. Black arrows, transcription start site. 100 kb is shown flanking the 5′ and 3′ end of each gene.

Close modal

We also confirmed that insertions in intron 1 of the growth hormone receptor (Ghr) resulted in the upregulation of full-length GHR protein in multiple mutagenized pools and tumors (Fig. 2A and C). Notably, we and several others have previously identified SB insertions in Ghr in multiple in vivo mutagenesis screens for diverse tumor types, including liver cancer, T-cell ALL, and AML (15, 20, 22). GHR overexpression has also been documented in several human cancers (26, 27). Growth hormone (GH)–GHR signaling activates JAK2 and STAT5, both of which are known to play critical roles in hematopoietic malignancies (5). Taken together, these results demonstrate the potential of our ex vivo mutagenesis approach to identify clinically relevant genes involved in the pathogenesis of leukemias and lymphomas.

We observed that in the vast majority of cases, SB insertions that were present in IL3-independent pools growing in vitro were present in tumors derived from those cells. In addition, all mutagenized IL3-independent pools produced leukemias/lymphomas or subcutaneous tumors when injected into mice. Thus, IL3-independent growth in culture provides a useful surrogate for tumor-forming potential that allows analysis of a significantly larger number of independently mutagenized pools than could feasibly be analyzed following in vivo tumor growth.

Identification of candidate oncogenes and tumor suppressors in a comprehensive mutagenesis screen

On the basis of the promising results from analysis of the initial set of IL3-independent Ba/F3 pools, we generated an additional 1,100 growth factor–independent pools through SB mutagenesis to comprehensively identify genes that contribute to cytokine signaling and B-cell transformation in this system. Transposon insertion sites were mapped using ligation-mediated PCR amplification and high-throughput sequencing, yielding 309,890,027 sequencing reads that mapped to the mouse genome with an intact barcode, a TA dinucleotide (the SB insertion site), and sequence corresponding to the T2/Onc transposon end (Supplementary Fig. S1D). A total of 2,160,208 of these sequences corresponded to nonredundant SB insertions that were analyzed further for identification of CIS, representing potential oncogenes and tumor suppressors.

Two established statistical approaches were used for CIS identification: Monte Carlo simulation analysis (MC-CIS) and gene-centric CIS analysis (gCIS; 24, 28). The MC-CIS method searches for regions of the genome that harbor a significantly higher density of transposon insertions than predicted by random chance while the gCIS method seeks an elevated density of SB insertions within the coding region of RefSeq genes. Whereas MC-CIS can identify CISs anywhere in the genome, gCIS may be more sensitive for detecting CISs that fall within the annotated coding genes. Therefore, the methods are complementary and both were employed to analyze our mutagenesis data. The MC-CIS and gCIS methods resulted in the identification of 4,189 and 3,963 CIS genes, respectively, at the P < 0.01 significance threshold (Supplementary Fig. S1E, Supplementary Dataset S1). For subsequent analyses of genes in clinical datasets, we focused on the top 100 CISs from each list. Thirty seven of the top 100 candidate genes were identified using both analytic methods, whereas 63 genes were unique to the MC-CIS and gCIS methods, respectively (Fig. 2D, Table 1).

Table 1.

Top 37 CISs in transformed Ba/F3 cells

GeneLibraries with gCISP for gCISLibraries with MC-CISP for MC-CISHuman ortholog
Jak1 688 1.0305E−116 734 4.5268E−144 JAK1 
Ghr 603 4.17564E−55 475 1.93785E−50 GHR 
Spidr 315 7.13759E−41 241 2.12995E−36 SPIDR 
Dnahc8 321 2.6757E−40 276 6.14075E−39 DNAH8 
Il3 829 1.9089E−37 1089 1.4909E−114 IL3 
Stat5b 220 4.53283E−17 229 7.42333E−36 STAT5B 
Pik3r5 243 2.44332E−34 344 1.08629E−46 PIK3R5 
Btbd9 317 1.55756E−33 276 6.14075E−39 BTBD9 
Nf1 269 1.22934E−26 113 1.55448E−14 NF1 
Gab2 243 6.14364E−25 110 3.5896E−14 GAB2 
Runx2 226 1.98022E−13 105 1.31608E−13 RUNX2 
Pacs1 180 4.23919E−23 163 1.63059E−20 PACS1 
Ssh2 243 7.49748E−22 124 5.44915E−14 SSH2 
Usp32 299 1.96206E−21 239 4.19956E−24 USP32 
Ptpra 160 9.82393E−21 159 1.04021E−21 PTPRA 
Hira 172 3.28871E−20 95 1.98069E−13 ZNF133 
Gbf1 184 5.10004E−20 80 5.49428E−11 GBF1 
Srgap2 219 2.38571E−19 75 2.01278E−09 SRGAP2 
Foxr2 114 1.19453E−18 147 8.66521E−23 FOXR2 
Supt3h 259 1.57426E−18 105 1.31608E−13 SUPT3H 
Pten 145 2.30594E−18 141 2.09551E−20 PTEN 
Mir1946b 140 4.94008E−18 137 1.39709E−18 — 
Tle4 163 3.2714E−16 99 9.67102E−12 TLE4 
Ppp6r3 142 6.40408E−16 76 3.02041E−11 PPP6R3 
Pcx 139 8.47659E−16 123 3.01141E−13 PC 
Nt5c2 174 2.37144E−13 128 3.04371E−12 NT5C2 
Gpatch8 137 1.08307E−14 140 4.49649E−15 GPATCH8 
Arih1 140 1.65875E−14 108 7.40407E−15 ARIH1 
Rps6ka5 177 1.03289E−13 123 2.71545E−14 RPS6KA5 
Herc1 178 1.12451E−13 116 9.75206E−12 HERC1 
Rapgef6 174 2.5049E−13 88 7.95886E−12 RAPGEF6 
Bicd1 161 6.74809E−13 101 9.65549E−12 BICD1 
Stk38l 95 7.66202E−13 92 9.13045E−13 STK38L 
Gigyf2 143 1.78156E−12 77 1.00184E−10 GIGYF2 
Jak2 120 6.47121E−12 272 4.82099E−41 JAK2 
Zswim5 140 6.81736E−12 84 2.48078E−11 ZSWIM5 
Braf 334 7.14529E−12 284 5.34539E−14 BRAF 
GeneLibraries with gCISP for gCISLibraries with MC-CISP for MC-CISHuman ortholog
Jak1 688 1.0305E−116 734 4.5268E−144 JAK1 
Ghr 603 4.17564E−55 475 1.93785E−50 GHR 
Spidr 315 7.13759E−41 241 2.12995E−36 SPIDR 
Dnahc8 321 2.6757E−40 276 6.14075E−39 DNAH8 
Il3 829 1.9089E−37 1089 1.4909E−114 IL3 
Stat5b 220 4.53283E−17 229 7.42333E−36 STAT5B 
Pik3r5 243 2.44332E−34 344 1.08629E−46 PIK3R5 
Btbd9 317 1.55756E−33 276 6.14075E−39 BTBD9 
Nf1 269 1.22934E−26 113 1.55448E−14 NF1 
Gab2 243 6.14364E−25 110 3.5896E−14 GAB2 
Runx2 226 1.98022E−13 105 1.31608E−13 RUNX2 
Pacs1 180 4.23919E−23 163 1.63059E−20 PACS1 
Ssh2 243 7.49748E−22 124 5.44915E−14 SSH2 
Usp32 299 1.96206E−21 239 4.19956E−24 USP32 
Ptpra 160 9.82393E−21 159 1.04021E−21 PTPRA 
Hira 172 3.28871E−20 95 1.98069E−13 ZNF133 
Gbf1 184 5.10004E−20 80 5.49428E−11 GBF1 
Srgap2 219 2.38571E−19 75 2.01278E−09 SRGAP2 
Foxr2 114 1.19453E−18 147 8.66521E−23 FOXR2 
Supt3h 259 1.57426E−18 105 1.31608E−13 SUPT3H 
Pten 145 2.30594E−18 141 2.09551E−20 PTEN 
Mir1946b 140 4.94008E−18 137 1.39709E−18 — 
Tle4 163 3.2714E−16 99 9.67102E−12 TLE4 
Ppp6r3 142 6.40408E−16 76 3.02041E−11 PPP6R3 
Pcx 139 8.47659E−16 123 3.01141E−13 PC 
Nt5c2 174 2.37144E−13 128 3.04371E−12 NT5C2 
Gpatch8 137 1.08307E−14 140 4.49649E−15 GPATCH8 
Arih1 140 1.65875E−14 108 7.40407E−15 ARIH1 
Rps6ka5 177 1.03289E−13 123 2.71545E−14 RPS6KA5 
Herc1 178 1.12451E−13 116 9.75206E−12 HERC1 
Rapgef6 174 2.5049E−13 88 7.95886E−12 RAPGEF6 
Bicd1 161 6.74809E−13 101 9.65549E−12 BICD1 
Stk38l 95 7.66202E−13 92 9.13045E−13 STK38L 
Gigyf2 143 1.78156E−12 77 1.00184E−10 GIGYF2 
Jak2 120 6.47121E−12 272 4.82099E−41 JAK2 
Zswim5 140 6.81736E−12 84 2.48078E−11 ZSWIM5 
Braf 334 7.14529E−12 284 5.34539E−14 BRAF 

NOTE: Top CIS genes identified in the comprehensive Ba/F3 mutagenesis screen. The top 100 genes from the MC-CIS and gcCIS methods were overlapped. Thirty seven of the top 100 candidate genes were identified using both analytic methods of CIS determination.

Each IL3-independent pool contained an average of 232 mutated CIS genes, ranging between tens to hundreds of CISs identified per individual pool (Supplementary Fig. S1F). In keeping with results from our initial pilot studies, IL3 was identified as a CIS gene (Supplementary Dataset 1). Although SB insertions in the IL3 were detectable in 75% of pools, in over 93% of such pools, these insertions represented <10% of the total number of detected insertions (Supplementary Fig. S1G). As such SB insertions near the IL3 gene result in low-level IL3 production that is insufficient to support bulk Ba/F3 cell growth (Fig. 1C and Supplementary Fig. S1A), we speculate that these insertions are synergistic with other SB-induced mutations that reduce the concentration of IL3 necessary to support population growth. Alternatively, IL3 insertions may support autocrine stimulation of the IL3 pathway in a minor subpopulation of cells within these pools. Notably, the functional consequence of SB insertions near the IL3 gene in Ba/F3 cells is distinct from a previously reported case wherein a retrovirus-like IAP insertion close to the 5′ end of the IL3 gene in WEHI-3B leukemia cells leads to constitutive high-level synthesis of IL3 (29). It is possible that this difference is due to stronger promoter or enhancer elements within the IAP element compared with the SB gene trap.

A significant number of well-characterized oncogenes and tumor suppressors were identified as CIS genes in this screen, including Jak1, Gab2, Stat5b, Pten, Nf1, and Foxr2. We also identified 90 genes with human orthologs that are mutated and/or dysregulated in human ALL and/or AML, only some of which have been functionally linked to oncogenic transformation, including Ezh2, Smc3, Nras, Flt3, Runx1, Idh2, and Idh1 for AML and Jak2, Kras, Ikzf1, and Abl1 for ALL (Supplementary Fig. S2A; Supplementary Tables S1 and S2; refs. 10, 30–32). Moreover, many genes mutated by SB in IL3-independent pools encode proteins that function at multiple levels in signaling pathways that are known to be important for diverse malignancies including leukemia, such as the MAPK, JAK/STAT, PI3K-AKT, and WNT pathways (Supplementary Fig. S2B and S2C). These results demonstrate that SB mutagenesis of Ba/F3 cells identifies genes that contribute to the pathogenesis of human cancers and suggest that the CIS genes identified in this screen may represent functionally relevant drivers in hematopoietic malignancies.

The T2/Onc transposon harbors bidirectional stop cassettes to inactivate tumor suppressors and a strong promoter in one orientation that drives oncogene expression. Thus, the transposon insertion site patterns are predictive of gain- or loss-of-function mutational mechanisms. In general, an enrichment of SB insertions in the same transcriptional orientation as a CIS gene signifies gain-of-function of a potential oncogene, whereas SB insertions distributed throughout a gene in either orientation denote loss-of-function of a putative tumor suppressor gene. For example, known oncogenes or genes whose hyperactivation is predicted to drive growth factor–independent proliferation (Jak1, Ghr, Stat5b, Gab2, and Foxr2) exhibited a strong enrichment for sense-strand SB insertions within the first intron or within 10 kb upstream of the transcription start site (Fig. 2E). These insertions are predicted to drive overexpression of full-length proteins. We also identified many sense-strand insertions in intron 11 of Braf, similar to those observed in our initial sequencing studies (Fig. 2A and B) that are predicted to drive overexpression of a truncated protein. In contrast, insertions at the well-characterized tumor suppressor Pten locus showed no directional bias (Fig. 2E).

On the basis of these findings, we generated a predictive metric, termed the sense-strand insertional bias (SIB) score, which reflects the relative enrichment of sense-strand insertions relative to the total insertions within a genomic region consisting of a gene body and an upstream 10-kb window. SIB scores significantly greater than 0.5 suggest that the gene of interest is preferentially mutated through a gain-of-function mechanism and is therefore likely functioning as an oncogene in this system. SIB scores for all identified CIS genes are provided in Supplementary Dataset S1. Previous studies have predicted the effect on gene function based on the position and orientation of T2/Onc insertions relative to gene transcription (33, 34). Although similar in principle, the SIB score provides a quantitative metric for determining the effect of SB insertions on gene function.

Analysis of CIS genes in human leukemia datasets

Given that we recovered recurrent SB insertions in a number of genes with human orthologs that are mutated and/or dysregulated in leukemias, we next sought to determine whether other CIS genes identified in our screen are relevant to human leukemogenesis. We first selected the top 100 CIS genes identified with the MC-CIS and gCIS methods (163 genes total) for analysis in existing TCGA and ICGC leukemia datasets with available survival information. Expression of 135 human orthologs of the 163 CIS genes were analyzed in a recent TCGA study of 161 AML patients (30). K-means clustering suggested an optimum of seven patient subgroups based on the expression of these genes and Kaplan–Meier survival analysis demonstrated statistically significant differences in survival between the subgroups (Fig. 3A and B). This pattern of association was not observed with a random set of genes not identified in our screen. Moreover, the expression of 66 CIS genes was sufficient to define the highest and lowest risk patient subgroups (Fig. 3C), perhaps highlighting genes with a particularly important role in AML pathogenesis. Indeed, known genes in the IL3 signaling pathway that promote leukemia development such as JAK1, JAK2, STAT5A, and STAT5B, as well as many CIS genes without a previously known role in leukemia, are highly expressed in rapidly progressing patients. A similar association with survival was observed when all CIS genes were used as input in K-means clustering (Supplementary Fig. S3A). Finally, an independent supervised clustering method was used to analyze CIS gene expression and AML patient survival, further demonstrating that expression of CIS genes is sufficient to stratify patients into high and low-risk subgroups (Supplementary Fig. S3B).

Figure 3.

Expression of CIS genes predicts survival in AML. A, consensus clustering of human AML data (TCGA, 2013; ref. 30) based on the expression of 135 human orthologs of CIS genes with SB insertions. B, Kaplan–Meier analysis depicting overall survival of AML patients clustered into 7 subgroups based on expression of CIS genes; n = 161 patients, P < 0.000217, log-rank test. C, heatmap depicting gene expression data of human orthologs of CIS genes in the low-risk and high-risk AML patient subgroups. Genes that are marked with an asterisk were tested in validation studies (shown in Figs. 4 and 5).

Figure 3.

Expression of CIS genes predicts survival in AML. A, consensus clustering of human AML data (TCGA, 2013; ref. 30) based on the expression of 135 human orthologs of CIS genes with SB insertions. B, Kaplan–Meier analysis depicting overall survival of AML patients clustered into 7 subgroups based on expression of CIS genes; n = 161 patients, P < 0.000217, log-rank test. C, heatmap depicting gene expression data of human orthologs of CIS genes in the low-risk and high-risk AML patient subgroups. Genes that are marked with an asterisk were tested in validation studies (shown in Figs. 4 and 5).

Close modal

Similar analyses were performed with the ICGC ALL dataset, which included gene expression data for 111 human orthologs of the 163 top CIS genes in 207 patients. Unsupervised clustering suggested an optimum of four patient subgroups that showed statistically significant survival differences (Supplementary Fig. S4A and S4B). Similar to our findings in AML, a specific set of 77 CIS genes was sufficient to stratify the highest and lowest risk ALL patients (Supplementary Fig. S4C), highlighting potential genes with particularly important roles in this malignancy.

Notably, we observed significant associations between the expression of 38 individual CIS gene orthologs and overall survival of AML (Supplementary Fig. S5) or ALL (Supplementary Fig. S6) patients, further highlighting genes that may contribute to the pathogenesis of these malignancies and providing compelling candidates for detailed mechanistic studies.

Functional validation of CIS genes in vitro

To validate the ability of individual genes identified in our mutagenesis screen to confer growth factor independence and leukemogenesis, we selected CIS genes to evaluate using in vitro and in vivo growth and tumorigenesis assays. These genes were chosen based on the following criteria: (i) they were expressed at high levels in high-risk leukemia patients (Fig. 3C and Supplementary Fig. S4C), (ii) they were associated with survival in AML or ALL patients (Supplementary Figs. S5 and S6), or (iii) they had not been previously linked to leukemia- or lymphomagenesis in the literature. We selected 34 genes for validation studies, of which 21 were among the top 163 candidates identified by MC-CIS and g-CIS as well as 13 genes that were ranked lower on the CIS list.

An overview of our validation scheme is shown in Fig. 4A. Candidate oncogenes were overexpressed using the SB system in Ba/F3 cells. In this assay, cells were cotransfected with the SB transposase and a modified T2 vector harboring a candidate CIS cDNA and a puromycin-IRES-GFP cassette. GFP+ cells were then assayed for IL3-independent growth, transplanted into NSG mice, and monitored for leukemia/lymphoma development. Importantly, transfection of Ba/F3 cells with transposase plus T2 vector expressing GFP only does not confer growth factor–independent proliferation or tumorigenesis (Fig. 4B and Fig. 5A).

Figure 4.

Functional validation of CIS genes in Ba/F3 cells. A, overview of in vitro and in vivo validation studies. B, quantification of cell proliferation of Ba/F3 cells overexpressing candidate CIS genes or transfected with empty vector grown in the absence of IL3. Growth curves were generated using a colorimetric assay based on absorbance. For this and all subsequent panels, error bars represent SDs from three independent measurements. C, growth of Ba/F3 cells overexpressing CIS genes after treatment with the JAK inhibitor, ruxolitinib. For each gene, the percentage of cells remaining after 5 days of ruxolitinib treatment was measured relative to untreated cells. Vector control cells were grown in the presence of IL3. n = 3 replicates per condition; *, P < 0.05; **, P < 0.01 (Student t test). D, growth of Ba/F3 cells overexpressing CIS genes after treatment with the MEK inhibitor, AZD6244. For each gene, the percentage of cells remaining after 5 days of AZD6244 treatment was measured relative to untreated cells. n = 3 replicates per condition; *, P < 0.05; **, P < 0.01 (Student t test).

Figure 4.

Functional validation of CIS genes in Ba/F3 cells. A, overview of in vitro and in vivo validation studies. B, quantification of cell proliferation of Ba/F3 cells overexpressing candidate CIS genes or transfected with empty vector grown in the absence of IL3. Growth curves were generated using a colorimetric assay based on absorbance. For this and all subsequent panels, error bars represent SDs from three independent measurements. C, growth of Ba/F3 cells overexpressing CIS genes after treatment with the JAK inhibitor, ruxolitinib. For each gene, the percentage of cells remaining after 5 days of ruxolitinib treatment was measured relative to untreated cells. Vector control cells were grown in the presence of IL3. n = 3 replicates per condition; *, P < 0.05; **, P < 0.01 (Student t test). D, growth of Ba/F3 cells overexpressing CIS genes after treatment with the MEK inhibitor, AZD6244. For each gene, the percentage of cells remaining after 5 days of AZD6244 treatment was measured relative to untreated cells. n = 3 replicates per condition; *, P < 0.05; **, P < 0.01 (Student t test).

Close modal
Figure 5.

Overexpression of CIS genes promotes leukemogenesis in vivo. A, Kaplan–Meier survival analysis of NSG mice transplanted with Ba/F3 cells overexpressing candidate CIS genes. B, spleen weights of animals analyzed in A. N = 6 animals for each group. Dotted line, the spleen weight of control animals. C, liver weights of animals analyzed in A. N = 6 animals for each group. Dotted line, the liver weight of control animals. D, flow cytometry data demonstrating that IL3-independent cells are GFP positive in culture (middle) and populate the bone marrow in vivo (right). E, quantification of percentage of GFP-positive cells in bone marrow. F, representative H&E-stained spleen and liver sections of NSG mice transplanted with Ba/F3 cells overexpressing CIS genes and/or an empty vector control. Scale bar, 50 μm.

Figure 5.

Overexpression of CIS genes promotes leukemogenesis in vivo. A, Kaplan–Meier survival analysis of NSG mice transplanted with Ba/F3 cells overexpressing candidate CIS genes. B, spleen weights of animals analyzed in A. N = 6 animals for each group. Dotted line, the spleen weight of control animals. C, liver weights of animals analyzed in A. N = 6 animals for each group. Dotted line, the liver weight of control animals. D, flow cytometry data demonstrating that IL3-independent cells are GFP positive in culture (middle) and populate the bone marrow in vivo (right). E, quantification of percentage of GFP-positive cells in bone marrow. F, representative H&E-stained spleen and liver sections of NSG mice transplanted with Ba/F3 cells overexpressing CIS genes and/or an empty vector control. Scale bar, 50 μm.

Close modal

For most CIS genes, we overexpressed full-length cDNAs. However, genes such as Braf and Usp32 exhibited a distinct pattern of SB insertions that suggested strong selection for overexpression of truncated proteins (Supplementary Fig. S7A). Analysis of Usp32 mRNA and protein levels was consistent with SB insertions driving expression of a truncated transcript (Supplementary Fig. S7B and S7C). We therefore expressed only the C-terminus of this gene, encompassing all exons downstream of the dominant transposon integrations.

Twenty-one of 34 (∼62%) individual genes tested promoted IL3-independent growth in Ba/F3 cells (Fig. 4B; Supplementary Table S3). Notably, although 13 of 34 genes tested were not sufficient to promote IL3 independence when tested individually, we found that transfecting different combinations of CIS genes as pools into Ba/F3 cells could promote IL3-independent proliferation, suggesting that multiple CIS genes may cooperate to promote growth factor independence and leukemogenesis (Supplementary Fig. S8). Determination of the copy number of the SB integrations in transfected cells using a quantitative real-time PCR assay demonstrated between 5 and 45 copies for 5 representative genes tested, with 3 of 5 genes present below 10 copies per cell (Supplementary Fig. S9A). In addition, we expressed a subset of these genes using lentiviral or retroviral vectors, further confirming that enforced expression of candidate CIS genes conferred IL3-independent growth (Supplementary Fig. S9B). Together with our observation that the control SB vector (empty or expressing GFP) does not confer IL3-independent proliferation or tumorigenesis, these findings suggest that insertional mutagenesis is not a major mechanism in the functional validation studies.

To determine whether the novel genes that conferred IL3-independent growth required ongoing JAK-STAT or MAPK signaling, Ba/F3 cells overexpressing putative oncogenes were treated with ruxolitinib, a JAK inhibitor, or AZD6244, a MEK inhibitor (35, 36). As positive controls, we confirmed that BRAF-C overexpression resulted in sensitivity to AZD6244, whereas CSF2RA, a receptor known to signal through the JAK–STAT pathway, conferred sensitivity to ruxolitinib (Fig. 4C and D). Proliferation driven by several of the tested genes required continued JAK–STAT signaling, as indicated by sensitivity to ruxolitinib but not AZD6244 (4930468A15Rik, ARCN1, HIRA, MSI2, SPAG9, and Usp32-C), suggesting that these proteins may represent novel activators of the JAK–STAT pathway. Consistent with this, overexpression of several genes resulted in highly elevated phospho-STAT5 levels (Supplementary Fig. S10A and S10B). An interesting exception was CSGALNACT1, expression of which resulted in sensitivity to AZD6244 but complete resistance to ruxolitinib, providing evidence that CSGALNACT1 functions upstream of MEK in the MAPK signaling pathway. The remaining 13 genes displayed modest to strong sensitivity to both inhibitors, suggesting that they activate multiple signaling pathways.

Functional validation of CIS genes in vivo

To directly test whether the candidates identified in our screen promote tumor development in vivo, we transplanted growth factor–independent Ba/F3 cells overexpressing 11 individual CIS genes or Ba/F3 cells transfected with control vector into NSG mice. These experiments confirmed that overexpression of all 11 genes resulted in leukemia/lymphoma development in vivo. All experimental animals, but not controls, were moribund with tumors prior to 40 days after transplantation (Fig. 5A). For all genes except HIRA, lymphoma development was associated with hepatosplenomegaly (Fig. 5B and C) and a near complete repopulation of the bone marrow with GFP+ cells (Fig. 5D and E, Supplementary Fig. S11A). Histologic analysis verified the presence of hematopoietic malignancy with the disruption of normal splenic architecture and varying degrees of lymphocyte infiltration in the liver and kidney for all tested genes (Fig. 5F and Supplementary Fig. S11B). Consistent with prior reports, we observed that tumors derived from Ba/F3 cells express both lymphoid (CD79a) and myeloid (CD11b) markers (Supplementary Fig. S12; refs. 37, 38). Therefore, genes identified through our transposon mutagenesis screen function as oncogenes in mice.

To further support and extend these findings, we determined whether USP32, a poorly characterized deubiquitinating enzyme that drives growth factor–independent proliferation and tumorigenesis in Ba/F3 cells (Figs. 4B and 5A), is also essential for growth in an independent human cancer cell line. Three independent sgRNAs were designed and delivered along with Cas9 to SU-DHL-1 cells, a lymphoma cell line with high USP32 expression based on analysis of the Cell Line Encyclopedia (39). Consistent with its growth-promoting properties in Ba/F3 cells, USP32 knockout using the CRISPR/Cas9 system reduced SU-DHL-1 cell proliferation in vitro and prolonged survival of NSG mice in xenograft assays (Supplementary Fig. S13A–S13C). Future studies will focus on elucidating the mechanisms through which USP32 and other novel genes identified in this screen promote leukemia and lymphoma development.

Unbiased genetic screens illuminate the functional relevance of genes mutated in human tumors. Many groups have successfully employed reverse genetic approaches such as RNA interference and cDNA overexpression screens for deciphering cancer genes (40, 41). Unlike these approaches, transposon-mediated mutagenesis generates both gain- and loss-of-function mutations in the same pool and/or tumor, which may more accurately model the process of tumor initiation, progression, and metastasis. The ex vivo transposon mutagenesis system described here is complementary to the traditional insertional mutagenesis approaches using transgenic mice and offers a rapid and cost-effective system of identifying genes that drive tumor development. The utilization of SB in ex vivo screening avoids local hopping, a phenomenon inherent to in vivo studies whereby the SB transposon favors transposition to sites adjacent to the donor concatemer in transgenic mice (42). Another advantage of ex vivo mutagenesis is the adaptability of this approach for identifying genes that promote or suppress tumorigenesis directly in human cancer cells. Supporting the feasibility of this application, somatic mutagenesis of human bone explant mesenchymal cells was recently performed with a hybrid lentiviral and SB mutagenesis system (termed Lentihop), generating genetically tractable myxofibrosarcomas in mice (43). In contrast to the Lentihop system where lentiviral vectors were used to introduce a mutagenic SB transposon into the genome, our ex vivo mutagenesis approach relied solely on transient SB mobilization. This system may be modified for conditional and/or dose-dependent expression of the transposon or transposase in different settings. Finally, there are many possible selection strategies that may be adapted for ex vivo mutagenesis screens, including the acquisition of anchorage-independent growth or metastasis of tumor cells implanted subcutaneously in mice.

In the current study, adaptation of the ex vivo SB mutagenesis system described here allowed the comprehensive identification of genes that confer growth factor independence of Ba/F3 cells. Several lines of evidence suggest that many of these genes play a role in human leukemia and lymphoma pathogenesis. For example, expression of CIS genes identified in growth factor–independent Ba/F3 cells was sufficient to stratify AML and ALL patients into subgroups with distinct survival probabilities. Moreover, a large fraction of tested CIS genes (62%), when overexpressed individually in Ba/F3 cells, were able to confer growth factor independence. Eleven genes were further selected and were shown to promote leukemogenesis in vivo. Importantly, as each IL3-independent Ba/F3 pool harbored SB insertions at numerous CISs, it is likely that an even higher percentage of genes that were identified in this screen are functionally relevant as they may need to act cooperatively to promote B-cell transformation. Indeed, we found that transfection of pools of genes that do not independently confer growth factor independence leads to IL3-independent growth.

The fact that well-characterized oncogenes and tumor suppressors are enriched at the top of our CIS lists indicates that ranking by P values using our statistical methods is effective. However, we found that some highly ranked CIS genes did not promote cytokine independence, whereas other lower ranked CIS genes did confer growth in the absence of IL3. In future studies, it may be possible to improve CIS calling methods by taking into account SIB scores in addition to P values, by ranking CIS genes according to the number of independent pools/tumors with SB insertions, or by applying a more stringent cutoff by removing the insertions with frequencies below 0.05% in each mutagenized pool or tumor.

Our study identified many known genes previously implicated in hematopoietic cancers such as Braf, Jak1, Gab2, Nf1, Pten, and Stat5b. More significantly, we uncovered numerous novel genes not previously functionally linked to leukemogenesis but recently shown to be dysregulated or altered in these malignancies. For example, we identified 62 of the approximately 300 genes that found to be mutated in human AML patients (30, 44). Similarly, we identified 28 of 74 genes that are mutated and/or dysregulated in ALL (Supplementary Fig. S2A). Although these genes are mutated at low frequencies in AML and ALL patients, we believe that they merit in depth investigation based on their identification as CISs in our functional screen. Furthermore, it is likely that identified genes are functionally relevant to nonhematopoietic malignancies. For example, BRAF, recurrently mutated in our study, is a key oncogene in melanoma. PTEN is a critical tumor suppressor in diverse human malignancies. The dataset provided here therefore represents a resource that may be useful for prioritizing the functional interrogation of genes that are recurrently mutated in human tumors.

Recent studies have led to the hypothesis that haploinsufficiency of candidate genes in regions commonly deleted in cancers may have detectable tumor suppressing activity only in the context of other cooperating genetic events (45). In support of this, MLL was recently identified as a haploinsufficient tumor suppressor on chromosome 7q that, in combination with TP53 deficiency and NF1 suppression, mediated AML and myelodysplastic syndrome (MDS) phenotypes in mouse models and human cells (46). Given that multiple gene mutations occur in individual cell pools and tumors, the mutagenesis system described here may be useful for uncovering additional context-dependent haploinsufficient tumor suppressors, which have recently been reported as important alterations in AML and myelodysplastic syndrome.

Among the genes that we validated in our screen, several represent potential therapeutic targets and therefore warrant future investigation. For example, G-protein coupled receptor kinase 5, GRK5, is a member of the family of serine/threonine protein kinases that phosphorylate G protein-coupled receptors as well as nonreceptor substrates. GRK5 was recently shown to phosphorylate Nucleophosmin (NPM1), which is mutated in approximately 28% of AML patients (30, 47). We observed that GRK5 is expressed at higher levels in high-risk AML patients compared with low-risk patients (Fig. 3C) and we demonstrated that GRK5 overexpression resulted in IL3 independence in Ba/F3 cells and leukemogenesis in mice (Fig. 5A–F).

Overexpression of PI3K regulatory subunit 5 (PIK3R5) resulted in growth factor independence and leukemogenesis while high expression of this gene is associated with poor prognosis of AML patients. PIK3R5 functions as a regulator of the PI3K pathway, a major oncogenic signaling node in leukemias and many other tumor types. Corroborating the potential importance of this gene to leukemogenesis, Pik3r5 was recently identified in two in vivo insertional mutagenesis screens for genes that cooperate with Mll-AF9 and NPM1 in AML (14, 22).

The distinct pattern of SB insertions observed in Ubiquitin specific peptidase 32 (Usp32) suggested a strong selection for overexpression of the C-terminus of this gene product in IL3-independent Ba/F3 cells (Supplementary Fig. S7A). Indeed, Ba/F3 cells harboring insertions in Usp32 exhibited highly elevated expression of mRNA downstream of the SB insertion site (Supplementary Fig. S7B). Accordingly, overexpression of truncated Usp32 was sufficient to transform Ba/F3 cells in vitro and in vivo (Figs. 4 and 5) while USP32 knockout reduced proliferation and tumorigenesis in a human lymphoma cell line (Supplementary Fig. S13). USP32 is frequently amplified in multiple tumor types (Supplementary Fig. S13D) and high expression is associated with poor survival in ALL (Supplementary Fig. S6). USP32 protein interaction partners were recently identified in a global proteomic analysis of deubiquitinating enzymes and their associated protein complexes (48). Interestingly, several of the genes that encode USP32-interacting partners are known to be mutated and/or exhibit altered expression in human leukemia cells, including SMC1 and FARSA (49, 50). Future studies are therefore warranted to further dissect the role of USP32 in human leukemogenesis and other tumor types.

In summary, our findings support the feasibility of ex vivo transposon mutagenesis screening and illustrate how combining cancer genomics with unbiased forward genetic approaches enable functional annotation of genes in hematopoietic malignancies. Although we identified a large number of CISs, intersecting these genes with those that are commonly mutated in cancer may aid future efforts to prioritize and design functional experiments. Future studies will focus on elucidating the mechanisms through which novel genes identified in this screen promote leukemogenesis. It is likely that this integrative approach will accelerate the development of novel diagnostic and therapeutic strategies for these malignancies.

No potential conflicts of interest were disclosed.

Conception and design: Y. Guo, K.A. O'Donnell

Development of methodology: Y. Guo, K.A. O'Donnell

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Guo, B.L. Updegraff, D. Durakoglugil, V.H. Cruz, S. Maddux, K.A. O'Donnell

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Guo, B.L. Updegraff, S. Park, S. Maddux, T.H. Hwang, K.A. O'Donnell

Writing, review, and/or revision of the manuscript: Y. Guo, B.L. Updegraff, S. Park, T.H. Hwang, K.A. O'Donnell

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Guo, S. Maddux

Study supervision: K.A. O'Donnell

Other (performing experiments): Y. Guo

The authors thank Z. Ivics and Z. Izsvak for sharing the SB100x plasmid, J. Shelton and J. Richardson for assistance and advice with histopathology, B. Druker and P. Scaglioni for sharing Ba/F3 and Ba/F3-Bcr-Abl cells, respectively, and J. Mendell, E. Olson, A. Zhang, M. White, J. Boeke, and members of the O'Donnell laboratory for critical reading of the manuscript. We also thank the McDermott Center Next Generation Sequencing Core and Jose Cabrera for assistance with the figures.

This work was supported by The Cancer Prevention Research Institute of Texas (R1101; RP140110), The American Cancer Society (ACS-IRG-02-196), The Welch Foundation (I-1881), and The Sidney Kimmel Foundation (SKF-15-067). K.A. O′Donnell is a CPRIT Scholar in Cancer Research and a Kimmel Scholar.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Alexander
WS
,
Nicola
NA
. 
Hemopoietic growth factor receptor abnormalities in leukemia
.
Leukemia Res
1998
;
22
:
1097
111
.
2.
Broughton
SE
,
Dhagat
U
,
Hercus
TR
,
Nero
TL
,
Grimbaldeston
MA
,
Bonder
CS
, et al
The GM-CSF/IL-3/IL-5 cytokine receptor family: from ligand recognition to initiation of signaling
.
Immunol Rev
2012
;
250
:
277
302
.
3.
Testa
U
,
Riccioni
R
,
Diverio
D
,
Rossini
A
,
Lo Coco
F
,
Peschle
C
. 
Interleukin-3 receptor in acute leukemia
.
Leukemia
2004
;
18
:
219
26
.
4.
Baker
SJ
,
Rane
SG
,
Reddy
EP
. 
Hematopoietic cytokine receptor signaling
.
Oncogene
2007
;
26
:
6724
37
.
5.
Chen
E
,
Staudt
LM
,
Green
AR
. 
Janus kinase deregulation in leukemia and lymphoma
.
Immunity
2012
;
36
:
529
41
.
6.
Palacios
R
,
Henson
G
,
Steinmetz
M
,
McKearn
JP
. 
Interleukin-3 supports growth of mouse pre-B-cell clones in vitro
.
Nature
1984
;
309
:
126
31
.
7.
Palacios
R
,
Steinmetz
M
. 
Il-3-dependent mouse clones that express B-220 surface antigen, contain Ig genes in germ-line configuration, and generate B lymphocytes in vivo
.
Cell
1985
;
41
:
727
34
.
8.
Lacronique
V
,
Boureux
A
,
Monni
R
,
Dumon
S
,
Mauchauffe
M
,
Mayeux
P
, et al
Transforming properties of chimeric TEL-JAK proteins in Ba/F3 cells
.
Blood
2000
;
95
:
2076
83
.
9.
Maxson
JE
,
Davare
MA
,
Luty
SB
,
Eide
CA
,
Chang
BH
,
Loriaux
MM
, et al
Therapeutically targetable ALK mutations in leukemia
.
Cancer Res
2015
;
75
:
2146
50
.
10.
Roberts
KG
,
Morin
RD
,
Zhang
J
,
Hirst
M
,
Zhao
Y
,
Su
X
, et al
Genetic alterations activating kinase and cytokine receptor signaling in high-risk acute lymphoblastic leukemia
.
Cancer Cell
2012
;
22
:
153
66
.
11.
Walters
DK
,
Mercher
T
,
Gu
TL
,
O'Hare
T
,
Tyner
JW
,
Loriaux
M
, et al
Activating alleles of JAK3 in acute megakaryoblastic leukemia
.
Cancer Cell
2006
;
10
:
65
75
.
12.
Daley
GQ
,
Baltimore
D
. 
Transformation of an interleukin 3-dependent hematopoietic cell line by the chronic myelogenous leukemia-specific P210bcr/abl protein
.
Proc Natl Acad Sci U S A
1988
;
85
:
9312
6
.
13.
Warmuth
M
,
Kim
S
,
Gu
XJ
,
Xia
G
,
Adrian
F
. 
Ba/F3 cells and their use in kinase drug discovery
.
Curr Opin Oncol
2007
;
19
:
55
60
.
14.
Bergerson
RJ
,
Collier
LS
,
Sarver
AL
,
Been
RA
,
Lugthart
S
,
Diers
MD
, et al
An insertional mutagenesis screen identifies genes that cooperate with Mll-AF9 in a murine leukemogenesis model
.
Blood
2012
;
119
:
4512
23
.
15.
Berquam-Vrieze
KE
,
Nannapaneni
K
,
Brett
BT
,
Holmfeldt
L
,
Ma
J
,
Zagorodna
O
, et al
Cell of origin strongly influences genetic selection in a mouse model of T-ALL
.
Blood
2011
;
118
:
4646
56
.
16.
Collier
LS
,
Adams
DJ
,
Hackett
CS
,
Bendzick
LE
,
Akagi
K
,
Davies
MN
, et al
Whole-body sleeping beauty mutagenesis can cause penetrant leukemia/lymphoma and rare high-grade glioma without associated embryonic lethality
.
Cancer Res
2009
;
69
:
8429
37
.
17.
Collier
LS
,
Carlson
CM
,
Ravimohan
S
,
Dupuy
AJ
,
Largaespada
DA
. 
Cancer gene discovery in solid tumours using transposon-based somatic mutagenesis in the mouse
.
Nature
2005
;
436
:
272
6
.
18.
Dupuy
AJ
,
Akagi
K
,
Largaespada
DA
,
Copeland
NG
,
Jenkins
NA
. 
Mammalian mutagenesis using a highly mobile somatic Sleeping Beauty transposon system
.
Nature
2005
;
436
:
221
6
.
19.
Mann
MB
,
Jenkins
NA
,
Copeland
NG
,
Mann
KM
. 
Sleeping Beauty mutagenesis: exploiting forward genetic screens for cancer gene discovery
.
Curr Opin Genet Dev
2014
;
24
:
16
22
.
20.
O'Donnell
KA
,
Keng
VW
,
York
B
,
Reineke
EL
,
Seo
D
,
Fan
D
, et al
A Sleeping Beauty mutagenesis screen reveals a tumor suppressor role for Ncoa2/Src-2 in liver cancer
.
Proc Natl Acad Sci U S A
2012
;
109
:
E1377
86
.
21.
Starr
TK
,
Scott
PM
,
Marsh
BM
,
Zhao
L
,
Than
BL
,
O'Sullivan
MG
, et al
A Sleeping Beauty transposon-mediated screen identifies murine susceptibility genes for adenomatous polyposis coli (Apc)-dependent intestinal tumorigenesis
.
Proc Natl Acad Sci U S A
2011
;
108
:
5765
70
.
22.
Vassiliou
GS
,
Cooper
JL
,
Rad
R
,
Li
J
,
Rice
S
,
Uren
A
, et al
Mutant nucleophosmin and cooperating pathways drive leukemia initiation and progression in mice
.
Nat Genet
2011
;
43
:
470
5
.
23.
Wu
X
,
Northcott
PA
,
Dubuc
A
,
Dupuy
AJ
,
Shih
DJ
,
Witt
H
, et al
Clonal selection drives genetic divergence of metastatic medulloblastoma
.
Nature
2012
;
482
:
529
33
.
24.
Brett
BT
,
Berquam-Vrieze
KE
,
Nannapaneni
K
,
Huang
J
,
Scheetz
TE
,
Dupuy
AJ
. 
Novel molecular and computational methods improve the accuracy of insertion site analysis in Sleeping Beauty-induced tumors
.
PLoS One
2011
;
6
:
e24668
.
25.
Monti
S
,
Tamayo
P
,
Mesirov
J
,
Golub
T
. 
Consensus clustering: a resampling-based method for class discovery and visualization of gene expression microarray data
.
Mach Learn
2003
;
52
:
91
118
.
26.
Garcia-Caballero
T
,
Mertani
HM
,
Lambert
A
,
Gallego
R
,
Fraga
M
,
Pintos
E
, et al
Increased expression of growth hormone and prolactin receptors in hepatocellular carcinomas
.
Endocrine
2000
;
12
:
265
71
.
27.
Gebre-Medhin
M
,
Kindblom
LG
,
Wennbo
H
,
Tornell
J
,
Meis-Kindblom
JM
. 
Growth hormone receptor is expressed in human breast cancer
.
Am J Pathol
2001
;
158
:
1217
22
.
28.
Starr
TK
,
Allaei
R
,
Silverstein
KA
,
Staggs
RA
,
Sarver
AL
,
Bergemann
TL
, et al
A transposon-based genetic screen in mice identifies genes altered in colorectal cancer
.
Science
2009
;
323
:
1747
50
.
29.
Ymer
S
,
Tucker
WQ
,
Sanderson
CJ
,
Hapel
AJ
,
Campbell
HD
,
Young
IG
. 
Constitutive synthesis of interleukin-3 by leukaemia cell line WEHI-3B is due to retroviral insertion near the gene
.
Nature
1985
;
317
:
255
8
.
30.
Cancer Genome Atlas Research N
. 
Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia
.
N Engl J Med
2013
;
368
:
2059
74
.
31.
Holmfeldt
L
,
Wei
L
,
Diaz-Flores
E
,
Walsh
M
,
Zhang
J
,
Ding
L
, et al
The genomic landscape of hypodiploid acute lymphoblastic leukemia
.
Nat Genet
2013
;
45
:
242
52
.
32.
Lilljebjorn
H
,
Rissler
M
,
Lassen
C
,
Heldrup
J
,
Behrendtz
M
,
Mitelman
F
, et al
Whole-exome sequencing of pediatric acute lymphoblastic leukemia
.
Leukemia
2012
;
26
:
1602
7
.
33.
Rahrmann
EP
,
Watson
AL
,
Keng
VW
,
Choi
K
,
Moriarity
BS
,
Beckmann
DA
, et al
Forward genetic screen for malignant peripheral nerve sheath tumor formation identifies new genes and pathways driving tumorigenesis
.
Nat Genet
2013
;
45
:
756
66
.
34.
Takeda
H
,
Wei
Z
,
Koso
H
,
Rust
AG
,
Yew
CC
,
Mann
MB
, et al
Transposon mutagenesis identifies genes and evolutionary forces driving gastrointestinal tract tumor progression
.
Nat Genet
2015
;
47
:
142
50
.
35.
Davies
BR
,
Logie
A
,
Mckay
JS
,
Martin
P
,
Steele
S
,
Jenkins
R
, et al
AZD6244 (ARRY-142886), a potent inhibitor of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase 1/2 kinases: mechanism of action in vivo, pharmacokinetic/pharmacodynamic relationship, and potential for combination in preclinical models
.
Mol Cancer Ther
2007
;
6
:
2209
19
.
36.
Eghtedar
A
,
Verstovsek
S
,
Estrov
Z
,
Burger
J
,
Cortes
J
,
Bivins
C
, et al
Phase 2 study of the JAK kinase inhibitor ruxolitinib in patients with refractory leukemias, including postmyeloproliferative neoplasm acute myeloid leukemia
.
Blood
2012
;
119
:
4614
8
.
37.
Sigvardsson
M
,
O'Riordan
M
,
Grosschedl
R
. 
EBF and E47 collaborate to induce expression of the endogenous immunoglobulin surrogate light chain genes
.
Immunity
1997
;
7
:
25
36
.
38.
Wilson
RB
,
Kiledjian
M
,
Shen
CP
,
Benezra
R
,
Zwollo
P
,
Dymecki
SM
, et al
Repression of immunoglobulin enhancers by the helix-loop-helix protein Id: implications for B-lymphoid-cell development
.
Mol Cell Biol
1991
;
11
:
6185
91
.
39.
Barretina
J
,
Caponigro
G
,
Stransky
N
,
Venkatesan
K
,
Margolin
AA
,
Kim
S
, et al
The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity
.
Nature
2012
;
483
:
603
7
.
40.
Oricchio
E
,
Nanjangud
G
,
Wolfe
AL
,
Schatz
JH
,
Mavrakis
KJ
,
Jiang
M
, et al
The Eph-receptor A7 is a soluble tumor suppressor for follicular lymphoma
.
Cell
2011
;
147
:
554
64
.
41.
Zender
L
,
Xue
W
,
Zuber
J
,
Semighini
CP
,
Krasnitz
A
,
Ma
B
, et al
An oncogenomics-based in vivo RNAi screen identifies tumor suppressors in liver cancer
.
Cell
2008
;
135
:
852
64
.
42.
Keng
VW
,
Yae
K
,
Hayakawa
T
,
Mizuno
S
,
Uno
Y
,
Yusa
K
, et al
Region-specific saturation germline mutagenesis in mice using the Sleeping Beauty transposon system
.
Nat Methods
2005
;
2
:
763
9
.
43.
Molyneux
SD
,
Waterhouse
PD
,
Shelton
D
,
Shao
YW
,
Watling
CM
,
Tang
QL
, et al
Human somatic cell mutagenesis creates genetically tractable sarcomas
.
Nat Genet
2014
;
46
:
964
72
.
44.
Riva
L
,
Luzi
L
,
Pelicci
PG
. 
Genomics of acute myeloid leukemia: the next generation
.
Front Oncol
2012
;
2
:
40
.
45.
Will
B
,
Steidl
U
. 
Combinatorial haplo-deficient tumor suppression in 7q-deficient myelodysplastic syndrome and acute myeloid leukemia
.
Cancer Cell
2014
;
25
:
555
7
.
46.
Chen
C
,
Liu
Y
,
Rappaport
AR
,
Kitzing
T
,
Schultz
N
,
Zhao
Z
, et al
MLL3 is a haploinsufficient 7q tumor suppressor in acute myeloid leukemia
.
Cancer Cell
2014
;
25
:
652
65
.
47.
So
CH
,
Michal
AM
,
Mashayekhi
R
,
Benovic
JL
. 
G protein-coupled receptor kinase 5 phosphorylates nucleophosmin and regulates cell sensitivity to polo-like kinase 1 inhibition
.
J Biol Chem
2012
;
287
:
17088
99
.
48.
Sowa
ME
,
Bennett
EJ
,
Gygi
SP
,
Harper
JW
. 
Defining the human deubiquitinating enzyme interaction landscape
.
Cell
2009
;
138
:
389
403
.
49.
Sen
S
,
Zhou
H
,
Ripmaster
T
,
Hittelman
WN
,
Schimmel
P
,
White
RA
. 
Expression of a gene encoding a tRNA synthetase-like protein is enhanced in tumorigenic human myeloid leukemia cells and is cell cycle stage- and differentiation-dependent
.
Proc Natl Acad Sci U S A
1997
;
94
:
6164
9
.
50.
Thol
F
,
Bollin
R
,
Gehlhaar
M
,
Walter
C
,
Dugas
M
,
Suchanek
KJ
, et al
Mutations in the cohesin complex in acute myeloid leukemia: clinical and prognostic implications
.
Blood
2014
;
123
:
914
20
.

Supplementary data