PiggyBac Transposon Mutagenesis: A Tool for Cancer Gene Discovery in Mice

See allHide authors and affiliations

Science  19 Nov 2010:
Vol. 330, Issue 6007, pp. 1104-1107
DOI: 10.1126/science.1193004


Transposons are mobile DNA segments that can disrupt gene function by inserting in or near genes. Here, we show that insertional mutagenesis by the PiggyBac transposon can be used for cancer gene discovery in mice. PiggyBac transposition in genetically engineered transposon-transposase mice induced cancers whose type (hematopoietic versus solid) and latency were dependent on the regulatory elements introduced into transposons. Analysis of 63 hematopoietic tumors revealed that PiggyBac is capable of genome-wide mutagenesis. The PiggyBac screen uncovered many cancer genes not identified in previous retroviral or Sleeping Beauty transposon screens, including Spic, which encodes a PU.1-related transcription factor, and Hdac7, a histone deacetylase gene. PiggyBac and Sleeping Beauty have different integration preferences. To maximize the utility of the tool, we engineered 21 mouse lines to be compatible with both transposon systems in constitutive, tissue- or temporal-specific mutagenesis. Mice with different transposon types, copy numbers, and chromosomal locations support wide applicability.

Genetic screening in higher organisms has been hampered for decades by the lack of efficient insertional mutagenesis tools. Retroviruses have been used for cancer gene discovery in mice, but their application has been limited to the study of hematopoietic and mammary tumors because of viral tropism for these tissues (1). DNA transposons, which are the key insertional mutagens in lower organisms, were inactivated in vertebrate genomes millions of years ago. Only recently have new transposons been engineered to be active in mammalian cells, a development that provides opportunities for their use as genetic tools in higher organisms (2). Sleeping Beauty (SB), a TC1/mariner transposon, was reconstructed from dormant elements in fish genomes and optimized to transpose in multiple cell types (3), including mouse embryonic stem cells (4). Further improvements of SB led to its successful application for somatic mutagenesis in mice (5, 6). Another transposon, PiggyBac (PB) from the cabbage looper moth Trichoplusia ni, was recently engineered to be highly active in mammalian cells (7) and has been shown to have biological properties distinct from those of SB (2, 79). PB can move larger DNA fragments (allowing complex transgene designs to be incorporated into the transposon), and it has a weaker tendency for local hopping in vitro (which makes it an attractive candidate for genome-wide mutagenesis). Furthermore, in contrast to SB, PB does not leave undesired footprint mutations after transposition. Lastly, PB and SB have different integration preferences.

To deploy PB for genetic screening in mice, we generated PB transposase knockin mice (RosaPB) and 19 mouse lines carrying different types of transposons (Fig. 1 and fig. S1) (10). All transposons possess PB and SB inverted terminal repeats (ITRs), allowing mobilization with both transposases. Promoter and enhancer elements, a splice donor, bidirectional SV40 polyadenylation signals, and two splice acceptors were introduced in between the ITRs to allow gain- or loss-of-function mutations, depending on the transposon orientation and its spatial relation to genes. Transgenic mice were generated with three variants of these bifunctional activating and inactivating transposons (ATP1, ATP2, and ATP3), which carry different promoter/enhancer elements (CAG, MSCV, and PGK). The chromosomal location and copy number of transposons were determined for each line by fluorescence in situ hybridization (FISH) and quantitative real-ime polymerase chain reaction (QPCR) (figs. S2 and S3). Nineteen transgenic mouse lines were established, with transposon arrays on 14 different chromosomes and 2 to 150 transposon copies (Fig. 1).

Fig. 1

Mouse lines carrying the genetic components of the transposon systems. (A) RosaPB and RosaSB knockin mice express PiggyBac or Sleeping Beauty transposase under control of the constitutively active Rosa26 promoter. (B) Transposon design and transposon mouse lines. (Left) Three transposon constructs were designed, which differ in their promoter and enhancer. All three transposons have PB as well as SB ITRs and can therefore be mobilized with both transposases. (Right) Three types of transposon mouse lines (ATP1, ATP2, and ATP3) were generated by using these constructs. A total of 19 ATP lines were developed and are listed in the tables on the right. For each line, the chromosome harboring the transposon donor locus and the transposon copy number in the donor concatemer are indicated. n.d, not determined; CβASA, Carp β-actin splice acceptor; En2SA, Engrailed-2 exon-2 splice acceptor; SD, Foxf2 exon-1 splice donor; pA, bidirectional SV40 polyadenylation signal; CAG, cytomegalovirus enhancer and chicken beta-actin promoter; MSCV, murine stem cell virus long terminal repeat; PGK, phosphoglycerate-kinase promoter.

To analyze the biology of PB transposition in vivo, we intercrossed RosaPB mice with 14 ATP lines and assessed nearly 900 progeny. When “low copy” ATP lines (<20 copies) were used, double-transgenic offspring were born at Mendelian frequency or close to it, whereas “high copy” ATP mice (e.g., ATP1-H39) did not produce live-born double-positive progeny (Fig. 2A). However, double-transgenic RosaPB;ATP1-H39 embryos were identified, suggesting extensive transposition-induced embryonic lethality (fig. S4). Less pronounced was the effect of the transposon type and donor location on embryonic lethality, although weak positional effects could be observed for some lines (e.g., ATP1-H5). Further characteristics of PB mobilization in vivo are shown in fig. S4.

Fig. 2

Embryonic lethality, tumor spectrum, and latency in different ATP mouse lines upon PiggyBac mobilization. (A) Embryonic lethality in offspring from RosaPB;ATP intercrosses. 14 ATP mouse lines were used. Each line’s transposon copy number is indicated below the graph. The number of mice born in each colony is shown above the graph. Bars represent double transgenic progeny as a percentage of expected. (B, C, and E) Tumor spectrum and latency in different ATP mouse lines upon PB transposon mobilization. The percentage of mice with hematopoietic or solid tumors is indicated for RosaPB;ATP2 (B), RosaPB;ATP1 (C), and RosaPB/ATP3 mice (E). Kaplan-Meier survival curves are shown for indicated genotypes. (D) CAG promoter–driven DsRed expression in mouse tissues. (Left) Red fluorescence detection in organs from wild-type (WT) and hprtCAG-DsRed (R) male mice. H, heart; M, skeletal muscle; K, kidney; B, brain; T, testicle; IN, small intestine; L, lung; S, spleen. (Right) Western blot analysis of the same organs using a DsRed-specific antibody (r) or control antibodies (t, α-tubulin; a, β-actin).

We next investigated whether PB-based mutagenesis induced cancer in the mice. We monitored 393 animals from multiple cohorts and found that beginning at 2 months of age, double-transgenic RosaPB;ATP mice, but not single-transgenic animals, developed cancers. The tumor type and latency were profoundly influenced by the transposon type. Mobilization of ATP2 transposons resulted in highly penetrant hematopoietic cancer formation. More than 90% of RosaPB;ATP2 mice developed aggressive leukemias and lymphomas (Fig. 2B and fig. S5). In contrast, RosaPB;ATP1 animals had almost exclusively solid tumors, including sarcomas and various carcinomas with poor differentiation and metastasis (Fig. 2C, fig. S6, and tables S1 and S2).

To explore whether the lack of hematopoietic carcinogenesis in ATP1 mice is related to tissue specificity of the CAG promoter, we generated reporter mice expressing DsRed constitutively under control of CAG from the Hprt locus. This knockin line has two major advantages over conventional transgenic reporter mice: (i) Promoters inserted into Hprt are not subject to epigenetic silencing and drive gene expression in all tissues (11) and (ii) only one copy of the promoter/reporter is used, which accurately reflects the effect of one ATP1-CAG-transposon. The analysis of Hprt-Cag-DsRed mice revealed robust DsRed expression in most organs but not in the spleen (Fig. 2E), suggesting low CAG activity in the hematopoietic compartment.

Lastly, we analyzed RosaPB;ATP3 mice, which developed solid and/or hematopoietic tumors, with a high percentage of mice having both (Fig. 2D and fig. S7). Further characterization of PB-induced tumors is shown in the supporting online material (SOM), including immunohistochemical subtype analysis of hematologic cancers (fig. S8), evidence for lack of generalized transposon-induced genomic instability (fig. S9), proof of clonality of insertions in tumors (fig. S10), analysis of transposase expression and transposon copy numbers in tumors (fig. S11), and the distribution pattern of different transposon types in genes (table S3).

SB has a strong tendency for local hopping, with 32 to 45% of all integrations mapping to the donor chromosome in SB-induced tumors (12, 13). To examine the potential of PB for genome-wide mutagenesis, we cloned more than 50,000 transposon integration sites in 63 hematopoietic tumors originating from three transposon mouse lines. A total of 5590 nonredundant insertions were identified. Figure 3 and table S4 show that PB integrations were uniformly distributed across the genome and correlated with the size and the TTAA frequency (the PB integration site) of individual chromosomes. There was a slightly increased insertion density on chromosomes 2, 9, 11, and 19 in all three groups of mice, which might be related to the high gene density (table S5) or the high number of oncogenic insertions (fig. S12) on those chromosomes. On donor chromosomes, the number of integrations was as expected or only slightly higher, and local hopping was confined to a small region (1 to 3 Mb) around the donor locus (Fig. 3). Together these results demonstrate the unique qualities of PB for genome-wide mutagenesis.

Fig. 3

PB integration pattern in tumors from different ATP2 mouse lines. (A) Analysis of tumors from RosaPB;ATP2-S1 mice (50 tumors, 4379 insertions). The bottom graph compares the percentage of transposon insertions to the size and TTAA frequency on individual chromosomes. The donor chromosome and locus were identified by FISH and are indicated by blue arrows. The top graph shows the distribution of insertions on Chr17 at a 1-Mb resolution. The numbers of expected (exp) and observed insertions outside (odl) and inside (idl) a 3-Mb interval surrounding the donor locus are indicated. Results from (B) RosaPB;ATP2-S2 mice (seven tumors; 594 insertions; donor Chr12) and (C) RosaPB;ATP2-H32 mice (six tumors; 618 insertions; donor Chr2) are presented accordingly.

To discover candidate cancer genes, we used a statistical framework based on Gaussian kernel convolution (14), which identified 72 unique common integration sites (CISs) at 67 independent loci (Fig. 4 and table S6). Figure S12 visualizes the global transposon integration pattern for each chromosome at a 1-Mb resolution and indicates all significant CISs. A search in the Retrovirus Tagged Cancer Gene Database (15) and in recent large-scale insertional mutagenesis studies (13, 16) revealed, as expected, that some of these CISs (or genes within CISs) have also been identified in earlier retroviral or SB screens (Fig. 4A). However, 42% of the CISs have not been reported before our study. Given that thousands of murine hematologic cancers have been generated and analyzed earlier by using retroviral or SB mutagenesis, this high number of novel CISs was unexpected and underscores the potential of PB for genetic screening.

Fig. 4

Identification of CISs in PiggyBac-induced hematopoietic tumors. (A) Pie chart representing all CISs identified: Some were reported in earlier screens that used retroviruses (RV, 36%), Sleeping Beauty (SB, 4%), or both (18%). Forty-two percent have not been identified in earlier screens. (B) Insertion pattern in known oncogenes and TSGs. Arrows indicate individual insertions and their orientation (red or blue, sense or antisense orientation of the transposon’s promoter to genes, respectively). (C) Promoter insertions in Spic in nine tumors. MSCV-Spic fusion transcripts were detected by reverse transcription PCR and sequencing. Spic expression was analyzed by QPCR and normalized to Gapdh expression. Tumors with Spic integrations are myeloid leukemias, as shown by immunohistochemical myeloperoxidase staining (left, spleen; right, lymph node from one mouse). Scale bars, 25 μm; error bars, SEM. ***P < 0.001.

The integration pattern in oncogenes suggests a high selective pressure for insertions in specific regions where transposons cause gene activation. All Nras or Evi1 integrations, for example, were upstream of the translation start site, and transposons were sense-oriented, supporting gene expression from the unidirectional murine stem cell virus (MSCV) promoter (Fig. 4B and fig. S13). Some oncogenes had more complex integration patterns. Notch1, for example, had one insertion cluster in intron 27, leading to overexpression of an intracellular Notch1 domain like in human T-ALL with t(7;9) translocations (17), and a second cluster of truncating integrations in intron 2, leading to a “decoupled” intracellular domain that can be expressed from a cryptic Notch1 promoter (18).

In contrast to oncogene insertions, hits in tumor suppressor genes (TSGs) were randomly distributed throughout genes (e.g., Ikzf1 or Pten; Fig. 4B) and were either sense- or antisense-oriented, suggesting that gene disruption is the disease-causing mechanism. In some tumors multiple integrations in TSGs occurred, leading to trapping of both alleles (fig. S14). Pten was frequently hit in this screen but not in previous retroviral screens, highlighting the different integration preferences of these insertional mutagens.

Examples of novel candidate cancer genes with promoter insertions identified in this screen are the PU.1-related transcription factor Spic, the histone deacetylase Hdac7, the Wnt pathway component Bcl9, and the cell cycle regulatory phosphatase Cdc14 (Fig. 4 and fig. S13). Spic was hit in nine tumors, which were all myeloperoxidase (MPO)-positive myeloid leukemias. In all tumors, transposons were sense-oriented and located within a narrow region [760 base pairs (bp)] upstream of the translation start site. MSCV-Spic fusion transcripts were confirmed by sequencing, and Spic overexpression was shown by QPCR (Fig. 4C). Little is known about the biological function of Spic, but a recent study has suggested its role in myelomonocytic development (19), which further supports our findings. HDAC7 is specifically expressed in CD4/CD8 double-positive human thymocytes (20), but its role in hematopoietic tumorigenesis has not been studied so far. The β-catenin cofactor Bcl9 is a rare translocation partner in human hematopoietic malignancies (21), and a recent study revealed frequent amplifications of a chromosomal region that includes BCL9 in various human malignancies (22).

The mouse is a major model organism for the study of human physiology and disease. Our study adds an efficient and versatile tool kit for genetic screening in mice that can be exploited for a wide range of applications, including constitutive, tissue-specific, or spatiotemporally controlled somatic mutagenesis, using conditional and tamoxifen-inducible PB (23). In tissue-specific screens, we were able to induce aggressive metastatic solid tumors, which in principle will facilitate future studies of tumor initiation, progression, and metastasis, as well as the genetic basis of various histological subtypes of individual cancers. PB can be used alone for genome-wide mutagenesis, but the ATP lines described here with donor loci on 14 different chromosomes can also be used in combination with SB [e.g., to exploit its local hopping tendency (24) for chromosome-specific screens in known cancer susceptibility regions]. These transposon tool kits in mice have the potential to increase the speed and efficiency with which cancer genes and oncogenic signaling networks are discovered and should likewise facilitate functional analysis of these genes.

Supporting Online Material

Materials and Methods

Figs. S1 to S14

Tables S1 to S7


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank the research support facility team, B.L. Ng, B. Fu, and Y. Hooks for technical assistance; N. Carter and C. Lopez-Otin for experimental resources; L. Wessels, J. ten Hoeve, and J. de Ridder for software used in data analysis; D. Saur for technical advice; and H. Prosser for reagents. The work was supported by the Wellcome Trust. R.R. and J.R. are recipients of a fellowship from the Deutsche Forschungsgemeinschaft and the Fundación Maria Cristina Masaveu Peterson, respectively.
View Abstract

Stay Connected to Science

Navigate This Article