The utility of transposon mutagenesis for cancer studies in the era of genome editing
Over the past decade, numerous TMIM studies have identified known and novel cancer
genes that either promote tumor initiation or co-operate with cancer-sensitizing mutations
to drive tumor progression. Recently, novel and elegant ways of employing transposon
mutagenesis to query specific cancer processes have been devised. In this section,
we summarize recent developments in the TMIM field.
Investigating tumor progression and evolution
TMIM screens have been performed in mice harboring various initiating mutations found
in human cancer. Such screens identify drivers of tumor progression and, importantly,
might be influenced by the sensitizing mutation. For example, Alexander and colleagues
performed TMIM in the hematopoietic system, which resulted in multiple leukemias 50]. A Jak2 V617F
-mutant background skewed the disease towards erythroleukemia, and insertions in the
ETS transcription factor genes Erg and Ets1 were identified as the most common events. Conversely, when using an activated ERG allele (TLS-ERG) as the sensitizing mutation, the authors identified frequent activating insertions
in Jak2, thus validating the co-operation between Jak2 and Erg50].
In an elegant study, TMIM was utilized to delineate evolutionary events during the
progression of colorectal cancer (CRC) 51]. Jenkins, Copeland and colleagues crossed the SB system into different sensitizing
backgrounds that carry mutations in genes that act at different stages of CRC: Apc min
, Kras G12D
, Smad4 +/?
or Tp53 R172H
(Fig. 4) 51]. Intriguingly, this approach revealed that functional loss of the wild-type Apc allele was the most crucial event for tumor progression in Apc min
, Kras G12D
and Tp53 R172H
tumors, but not in tumors that were initiated by heterozygous loss of Smad4. Instead, those tumors displayed frequent insertions in the wild-type Smad4 allele along with mutually exclusive insertions in Rspo1 and Rspo2 that promoted overexpression of these R-spondins, which are known enhancers of Wnt
signaling. In addition, 111 candidate cancer genes were identified that were independent
of the initiating mutation.
Fig. 4. Use of transposon-mediated insertional mutagenesis (TMIM) screening to identify mutations
that co-operate with specific genetic lesions associated with different stages of
colorectal cancer development. The top panels illustrate a model of colorectal cancer
initiation and progression 101], along with genetic alterations associated with these stages. TMIM screens using
mouse models carrying mutations in corresponding genes have revealed that Apc was the predominant gene inactivated in tumors from all sensitizing genotypes apart
from Smad4 KO/+
cases, where inactivation of the remaining wild-type Smad4 gene is the most frequent insertional event
These studies illustrate how sensitizing mutations can co-operate with transposon-associated
lesions and how different pre-existing mutations can sometimes influence the trajectory
of subsequent mutation acquisition during tumor development. In the case of human
CRC, loss of APC is thought to be the initiating event, whereas mutations in KRAS, TP53 or SMAD4 occur later during tumor progression. Indeed, transposon-insertion-mediated loss
of Apc appeared to be a prerequisite for colon tumorigenesis in the Apc min
, Kras G12D
and Tp53 R172H
backgrounds, whereas insertions in Kras and Tp53 are rare in Apc-loss-driven CRC 51] (Table 1; Fig. 4). This finding further supports the notion of APC being the gatekeeper of CRC. Conversely, leukemogenesis is initiated by either mutant
Jak2 or Erg and progresses upon transposon insertions in the other gene, suggesting that the
temporal sequence of mutation might be irrelevant 50]. Taken together, TMIM is a valuable tool to delineate tumor progression, and future
studies that unravel the genetic dependencies of co-operating mutations on different
initiating mutations in other cancer types will shed further light on the genetics
of tumor progression and might be useful for devising treatment strategies.
Determining the evolutionary history of mutations within tumors can inform our understanding
of the mutational forces that shape cancer development. To assess tumor clonality
in a more quantitative fashion, new methods to estimate the frequency of transposon
insertions in tumors have been devised. Historical methods to retrieve insertion sites
have been based primarily on PCR amplification of restriction-endonuclease-digested,
adaptor-ligated tumor DNA, followed by high-throughput sequencing. However, sequence
coverage cannot be used to infer tumor clonality accurately owing to PCR biases as
a result of the variable distribution of restriction enzyme sites in the genome. An
alternative approach, called shear-splink, was developed by Jonkers and colleagues
that fragments DNA by acoustic shearing, mitigating this bias 52]. In addition, as DNA is fragmented at random, each fragment harbors a potentially
unique stretch of DNA that can serve as a molecular barcode. Quantification of these
barcodes permits estimation of transposon clonality within a heterogeneous sample.
Rad and colleagues used a similar approach, termed quantitative insertion site sequencing
(QIseq), to illustrate the marked genetic complexity of pancreatic tumors 32]. Although these approaches can estimate transposon clonality, they cannot distinguish
between transposon heterogeneity arising during tumor evolution in a monoclonal sample
and multiple distinct insertions in a polyclonal tumor population.
Identifying genes involved in metastasis
In addition to identifying genes involved in tumor initiation and progression, TMIM
has been performed to discover genes that promote tumor dissemination. Largaespada
and colleagues expressed the SB system in p53-deficient mouse osteoblasts and identified
candidate genes involved in metastasis by comparing transposon insertions from osteosarcoma
metastases with those found in primary tumors 53]. Approximately one-third of CIS-associated genes found in metastases were evident
in primary tumors. Furthermore, from this analysis, five candidate oncogenes and 38
tumor suppressors were identified, including nine genes that have been implicated
previously in cancer metastasis. To study further the evolutionary relationships between
metastases and parental ancestors, the authors conducted parsimony analysis of tumors
using transposon integration sites as molecular footprints. Osteosarcoma metastases
were found to be highly clonal but appeared to show different patterns of evolution
from the primary tumor.
Taylor and colleagues performed a TMIM screen aimed at identifying genes affecting
dissemination of medulloblastoma in Ptch1 +/?
heterozygous null or mutant Tp53 mouse backgrounds 54]. Interestingly, the authors found that both transposon-driven mouse and human metastatic
medulloblastoma are clonal but divergent from the primary tumor, suggesting that only
a rare subclone in the primary tumor is able to metastasize. Four of the identified
candidate genes were validated as drivers of medulloblastoma dissemination by retroviral
delivery of these candidates to the cerebellum in combination with overexpression
of the Ptch1 ligand sonic hedgehog (Shh) 55]. These studies demonstrated the utility of TMIM screens to discover drivers of metastatic
spread, and further studies will identify candidate metastasis genes in certain genetic
backgrounds and tumor types. Some mouse cancer models might not be suitable for identification
of metastasis genes by TMIM because the mice have to be sacrificed before the formation
of macroscopic metastases owing to the primary tumor size. However, surgical removal
of the primary tumor to allow more time for metastasis growth or transplantation of
primary tumor cells into syngeneic wild-type mice could circumvent this issue. Nonetheless,
these reports illustrate how TMIM can be employed to query the clonal relationship
of a primary tumor and its metastases, complementing the use of transposons to identify
genes involved in tumor progression.
Identifying alterations in cancer pathways
Apart from identifying genes promoting tumor progression, TMIM screens have been used
to define the most prominent signaling pathways deregulated in tumors. Using the TAPDANCE
tool, Largaespada and colleagues performed a pathway-centric analysis of alterations
in Tp53-mutant, EGFR-driven peripheral nerve sheath tumors to identify roles for the phosphoinositide
3-kinase (PI3K)-AKT-mTOR, mitogen-activated protein kinase (MAPK) and Wnt/?-catenin
pathways in the development of this tumor type 56]. Novel pathways have also been revealed in melanoma driven by oncogenic B-Raf V600E
. Xu and colleagues identified a network involving Magi2 with a PB screen at low transposon
copy number and also found insertions in Map3k1 and Map3k2 that resulted in ERK activation 57]. However, these insertions occurred in melanomas that had not recombined the conditional
oncogenic B-Raf V600E
allele. Although not examined, this suggests that aberrant MAP3K1/MAP3K2 activation
could represent another means to activate the MAPK pathway in human melanoma besides
the common BRAF and NRAS mutations. The melanoma SB screen performed by Jenkins, Copeland and colleagues identified
numerous candidate cancer genes, and pathway analysis found significant enrichment
of CIS-associated genes in many cancer-related signaling pathways, including Wnt/?-catenin,
TGF-?, PI3K and MAPK signaling, as well as in many biological processes 38]. Recently, it was shown that, by integrating SB TMIM in mice and mutation analysis
of human cancer genomes, loss of function of the transcription factor CUX1 drives
myeloid malignancy and other cancer types 20]. It was demonstrated that CUX1 antagonizes the PI3K–AKT signaling pathway by regulating
transcription of the PI3K inhibitor PIK3IP1. Finally, a SB medulloblastoma screen
in Ptch1 +/?
mice identified candidate cancer genes and associated protein networks capable of
distinguishing the molecular subgroups of human medulloblastoma, demonstrating the
power of transposon screens to recapitulate the genetic changes in human cancer 58].
These studies suggest that pathway and network analyses can provide insight into mechanisms
of human disease and might predict survival and treatment outcomes. Thus, TMIM is
a powerful approach to unravel the functional association of altered signaling pathways
or cell-biological processes with cancer development. Conventional sequencing efforts
can fail to identify such associations because the mutation rate of individual genes
regulating these pathways or processes is not above the background mutation rate.
Moreover, although TMIM cannot recapitulate activating mutations of proto-oncogenes,
pathway analyses of TMIM datasets can reveal the crucial functions downstream of oncogenes
that are commonly mutated in human cancer.
Identification of novel mechanisms of gene deregulation
In cancer cells, loss of mRNA and protein expression can occur without any obvious
genetic alteration in corresponding protein-coding regions. Notably, recent TMIM studies
have identified novel non-coding regulatory regions and other mechanisms of gene deregulation
that promote tumorigenesis. For example, a PB screen identified recurrent transposon
insertions in a 200-kb noncoding region (Ncruc) upstream of the Cdkn2a gene 32], which encodes the tumor suppressors p16Ink4a and p19Arf and is frequently inactivated
by prototypic gene-body insertions in both SB and PB pancreatic cancer screens 24], 32]. Transposon insertions in or genomic loss of the Ncruc region were associated with
reduced expression levels of Cdkn2a in cis, demonstrating the power of PB insertional mutagenesis screens to identify non-coding
DNA regions or genes with crucial roles in tumorigenesis.
Although target-site preferences suggest that PB-based TMIM screens might be more
useful to identify regulatory elements compared with SB transposons (Fig. 3), SB-mediated screens have also been fruitful in identifying atypical mechanisms
of gene deregulation in cancer. For example, Dupuy and colleagues performed a SB-mediated
hepatocellular carcinoma (HCC) screen and found frequent insertions in the complex
imprinted Dlk1-Dio3 locus. A domesticated retrotransposon, Rtl1, located in this locus was shown to be overexpressed in all tumors with Dlk1-Dio3 insertions 59]. Furthermore, ectopic overexpression of Rtl1 in mouse livers induced HCC, validating Rtl1 as a novel cancer driver. Examination of human liver tissue showed that Rtl1 is transcriptionally inactive in normal liver but can be reactivated in human HCC,
supporting a role for Rtl1 in human HCC development.
In a SB-mediated TMIM screen aimed at identifying genes that co-operate with oncogenic
B-Raf in melanoma development, a significant enrichment of genes was discovered among the
CISs that encode mRNAs with the ability to regulate the expression of the tumor suppressor
Pten48]. These so-called competitive endogenous RNAs control Pten levels as microRNA decoys,
in a protein-coding-independent fashion. While these CIS-associated genes are classical
protein-coding genes, our analysis highlighted a non-coding function of their mRNAs.
Only 2Â % of the mammalian genome encodes protein-coding genes; however, the non-coding
portion of the genome, both transcribed (e.g., microRNAs, long non-coding RNAs) and
non-transcribed (e.g., enhancers), plays crucial roles in physiology and pathology.
TMIM screens have barely scratched the surface of the non-coding space, and re-analyzing
existing SB and PB mutagenesis data might reveal additional non-coding insertion hotspots.
Identifying mechanisms of resistance to therapy
TMIM has been useful in identifying genes that mediate therapeutic drug resistance
both in vitro and in vivo. Schmidt and colleagues conducted a PB screen in four different
human cell lines derived from neuroblastoma, breast and cervical cancer to identify
genes whose overexpression mediates resistance to paclitaxel 60]. Interestingly, while the authors identified multiple CISs in the four cell lines,
the only CIS that was common to all four cell lines was the ABCB1 gene 60], which encodes an ABC-transporter associated with multi-drug resistance 61]. This suggests the existence of both cancer-type-specific and common mechanisms of
drug resistance. In addition, Xu and colleagues performed a PB screen in melanoma
cells and identified BRAF and CRAF as mediators of resistance to the BRAF inhibitor
vemurafenib 62], recapitulating previous observations in human melanoma patients and cell lines treated
with vemurafenib 63]–65].
In diploid cells, biallelic inactivating transposon insertions that completely abrogate
gene expression are rare compared with monoallelic events, thus hampering the identification
of genes that promote drug resistance only upon complete loss of expression. To tackle
this issue, Ashworth and colleagues 66] took advantage of a haploid mouse embryonic stem (ES) cell system to screen for mediators
of olaparib toxicity, in which inactivating transposon insertion can result in complete
loss of gene expression. The authors identified the poly [ADP-ribose] polymerase 1
gene Parp1 as a mediator of olaparib toxicity, and their results suggested that loss of Parp1
could result in olaparib resistance in patients 66]. In another mouse ES cell screen, Jonkers and colleagues identified loss of the gene
53bp1 as a mediator of survival and DNA-damage responses in Brca1-null cells 67]. Reduced 53BP1 expression was associated with basal-like, triple-negative, and BRCA1/2-mutant breast cancer in humans, suggesting that downregulation of 53BP1 might be
an important survival factor in such tumors, particularly during chemotherapy-induced
DNA damage. These studies demonstrate the utility of TMIM to identify mediators of
resistance in human cancer cell lines as well as ES cells.
Drug resistance in patients develops in the context of a supporting microenvironment
and, thus, in vitro approaches might be limited in their ability to identify resistance
genes. To avoid this shortcoming of in vitro drug-resistance screens, a SB screen
in a B-Raf V600E
-driven mouse model of melanoma was performed. This identified transposon insertion
sites in treatment-naïve tumors as well as melanomas treated with the vemurafenib
progenitor compound PLX4720 49]. Insertions in several known mediators of resistance were enriched in the PLX4720-treated
tumors, validating this approach for resistance gene discovery. An ERAS-AKT-BAD signaling
axis was validated as a mediator of drug resistance, which mimics the paracrine mechanism
of stromal hepatocyte growth factor-mediated resistance 68], 69]. Curiously, many of the genes that have been previously identified in cell lines
as promoters of resistance through reactivation of MAPK signaling were not identified
in this in vivo study. A possible explanation is that such mutations are preexisting
in patients only in a minor tumor subclone that no longer relies on oncogenic BRAF
signaling. Conversely, transposon mobilization was induced concomitantly with the
initiating B-Raf mutation in the resistance TMIM. In these tumor cells, transposon
insertions that would otherwise result in MAPK activation might be negatively selected
owing to functional redundancy with oncogenic B-Raf. Thus, additional insight might
be gained from studies in which transposon mobilization is induced at the time of
drug treatment.
Novel approaches of employing transposon mutagenesis
In vivo transposon mutagenesis requires up to four transgenic alleles to accelerate
tumorigenesis in a tissue-specific manner in a sensitizing background. Generating
and maintaining compound mutant mouse strains is time consuming and costly, prompting
alternative ways of utilizing the transposon systems. Molyneux and colleagues transduced
immortalized primary human bone mesenchymal cells with SB and a lentivirus harboring
the elements of a SB transposon, and, when injected into mice, the transplanted cells
produced myxofibrosarcomas 70]. For human candidate cancer gene discovery, both the insertions of the parental lentivirus
as well as the remobilized transposons were mapped. In another study, neural stem
cells were derived from transgenic mice harboring the SB system and a Nestin-Cre allele
71]. Following in vitro differentiation, the neural stem cells were immortalized through
SB mutagenesis and the resulting immortalized astroglial-like cells were injected
into SCID mice to identify genes that drive glioblastoma formation. CIS mapping of
immortalized cell lines and tumors identified partially overlapping CISs, suggesting
differential roles of the identified genes during immortalization and tumorigenesis.
In vitro delivery of the transposon system components followed by orthotopic or subcutaneous
transplantation thus represents another means for in vivo selection and identification
of candidate cancer genes.
The SB transposon system has also been used as a reverse-genetics tool to validate
candidate cancer genes. Futreal and colleagues created transposons with both SB and
PB terminal repeats that also harbored IRES-cDNA cassettes 72], such that the cDNA cargo was expressed only when transposon insertion occurred in
transcribed genes. Using these transposons, the authors tested kinases with point
mutations encoding putative gain-of-function oncogenic alleles. Mice were generated
carrying multiple transposons with different cDNA cargos and crossed to SB transgenic
mice, leading to tumorigenesis by in vivo selection of the kinase mutants with the
highest oncogenic potential in somatic cells. This report elegantly displays how the
transposon system can be utilized to discern the relative oncogenic properties of
several candidate genes simultaneously in all or selected organs.
To extend the utility of TMIM to another model system, transgenic rats carrying the
components of the SB or PB system have been created 73]. The transposons carried both SB and PB terminal repeats as well as a tyrosinase
expression cassette, permitting coat-color-based phenotyping for transposon zygosity
and genomic position effects on tyrosinase expression in albino rat backgrounds. In
the future, it will be interesting to determine the overlap in cancer genes identified
by TMIM screens in mouse and rat and their relevance to human cancer.