Cell. May 16, 2008; 133(4): 727–741.
PMCID: PMC2405818

Large-Scale Mutagenesis in p19ARF- and p53-Deficient Mice Identifies Cancer Genes and Their Collaborative Networks

Summary

p53 and p19ARF are tumor suppressors frequently mutated in human tumors. In a high-throughput screen in mice for mutations collaborating with either p53 or p19ARF deficiency, we identified 10,806 retroviral insertion sites, implicating over 300 loci in tumorigenesis. This dataset reveals 20 genes that are specifically mutated in either p19ARF-deficient, p53-deficient or wild-type mice (including Flt3, mmu-mir-106a-363, Smg6, and Ccnd3), as well as networks of significant collaborative and mutually exclusive interactions between cancer genes. Furthermore, we found candidate tumor suppressor genes, as well as distinct clusters of insertions within genes like Flt3 and Notch1 that induce mutants with different spectra of genetic interactions. Cross species comparative analysis with aCGH data of human cancer cell lines revealed known and candidate oncogenes (Mmp13, Slamf6, and Rreb1) and tumor suppressors (Wwox and Arfrp2). This dataset should prove to be a rich resource for the study of genetic interactions that underlie tumorigenesis.

Keywords: SYSBIO, SIGNALING, HUMDISEASE

Introduction

Retroviral insertional mutagenesis in mice has proven to be an efficient tool for identification of novel cancer genes, providing a valuable complement to the study of human tumors. Slow transforming retroviruses, such as Moloney Murine Leukemia Virus (MuLV), can mutate cellular genes by integration of their proviruses into the host genome. Cells that have acquired a proliferative advantage through proviral mutation of cellular genes, may acquire additional mutations by viral reinfection and eventually develop into tumors (reviewed in Uren et al., 2005). The position of insertions can be determined by amplifying DNA flanking the provirus using linker-mediated PCR and mapping the resulting sequences onto the genome. Many regions that are tagged in multiple independent tumors (termed Common Insertion Sites or Common Integration Sites or CISs) have been identified in previous studies (compiled in the RTCGD, http://rtcgd.abcc.ncifcrf.gov/, (Akagi et al., 2004)). A high proportion of these loci are orthologous with the loci of known human cancer genes, examples being Fli-1, Evi1, and Pim1 (Ben David et al., 1988; Morishita et al., 1988; Cuypers et al., 1984). However, the oncogenic capacity of many of the candidate cancer genes identified by these screens still needs further experimental validation.

Components of the p19ARF-MDM2-p53 signaling pathway are mutated in a large fraction of human cancers (Sharpless, 2005; Levine et al., 2006). p19ARF acts upstream of p53 and can enhance its transcriptional activity by antagonizing MDM2-mediated degradation of p53 (Sherr, 2006). Comutation of p19ARF and p53 rarely occurs (Eischen et al., 1999; Schmitt et al., 1999), and deletion of p19ARF in heterozygous p53 knockout mice reduces the selective pressure for loss of the p53 wild-type allele for tumor formation (Moore et al., 2003). In addition, p19ARF was shown to be pivotal for suppression of chemically- or radiation-induced tumors by p53 in mice (Efeyan et al., 2006; Christophorou et al., 2006). These data suggest the primary function of p19ARF is to activate p53. However, several studies indicate that p19ARF may also suppress tumorigenesis independently of p53. Mice lacking both p19ARF and p53 showed tumors in a wider range of tissue types and more frequently developed multiple primary tumors than mice lacking either of the two genes (Weber et al., 2000). Furthermore, loss of p19ARF increased the number and size of chemically-induced papillomas both in wild-type and p53−/− mice (Kelly-Spratt et al., 2004).

We performed insertional mutagenesis screens in p53−/−, p19ARF−/− and wild-type mice to identify genes that collaborate with loss of either p53 or p19ARF in tumorigenesis and in doing so gained new insight in the functional differences between these tumor suppressors. We also used these data to construct networks of collaborative and mutually exclusive interactions between CIS loci. To date, the primary limitation of identifying genotype-CIS and CIS-CIS collaborations has been not only the number of tumors but also the number of inserts per tumor. If only a fraction of the inserts of each tumor are detected, the power of any statistical test for genotype specificity or for collaboration between loci will be greatly reduced. With this in mind, we optimized our insertion site cloning and analyzed more than 500 tumors yielding over 10,000 independent insertion sites. By comparison, many of the screens published to date identify hundreds of insertions (Li et al., 1999; Hwang et al., 2002; Lund et al., 2002; Mikkers et al., 2002; Suzuki et al., 2002; Johansson et al., 2004; Theodorou et al., 2007; Stewart et al., 2007; Slape et al., 2007), the largest containing 1538 insertions from 245 tumors (Suzuki et al., 2006). As a result, a large number of novel CISs were identified near known and candidate oncogenes and tumor suppressor genes. These data create a resource (http://mutapedia.nki.nl) that will help cancer researchers to identify new cancer genes and further investigate the interactions of established oncogenic lesions.

Results

MuLV Accelerates Lymphomagenesis in p19ARF−/− and p53−/− Mice

We conducted retroviral insertional mutagenesis screens in p19ARF−/−, p53−/− and wild-type mice on a pure FVB genetic background. Mice were infected with MuLV at postnatal day 1 and monitored for tumor growth. Both p53−/− and p19ARF−/− mice are predisposed for development of lymphoma (Donehower et al., 1992; Kamijo et al., 1999), and MuLV infection further accelerated lymphomagenesis in these mice as well as in wild-type mice (p value < 0.0001 for MuLV-infected p19ARF−/− versus noninfected p19ARF−/−, p value < 0.0001 for MuLV-infected p53−/− versus noninfected p53−/−, log-rank test) (Figure 1A). Mice developed tumors almost exclusively in spleen, thymus and lymph nodes. The majority of tumors (n = 349) were analyzed by flow cytometry using T cell and B cell-specific markers (CD3 and B220, respectively, Figure S1A). No large differences were found in the incidence of tumors in these organs between the different genotypes, and the proportion of analyzed tumors from the different organs is roughly the same for each genotype (Figure S1B).

Figure 1
Accelerated Tumor Formation in MuLV-Infected Mice and Identification of Common Insertion Sites Using the GKC Framework

High-Throughput Cloning and Sequencing of MuLV Insertion Sites

To identify genes mutated by proviral insertions, flanking sequences of the retroviral insertions were cloned by linker-mediated PCR (Devon et al., 1995; Mikkers et al., 2002) using tumor DNA digested with either Sau3aI or Tsp509I. For the majority of samples two PCRs were performed (one for each enzyme) and shotgun subcloned. 96 colonies per PCR were picked and sequenced. In total, 510 tumors (220 p19ARF−/−, 123 p53−/−, and 167 wild-type), isolated from 455 mice were analyzed, yielding approximately 100,000 sequence reads. After filtering, 46,197 could be mapped and oriented onto the genome. Sequences cloned more than once from the same mouse likely represent the same insertion. Therefore, these were built into contigs to avoid overcounting the insertions of disseminated tumors. In total, 10,806 independent insertions were retrieved (3194, 4438, and 3174 from wild-type, p19ARF−/− and p53−/− tumors, respectively) (for overview see Figure 1B), yielding an average of 23.7 insertions per mouse.

Identification of CISs Using a Kernel Convolution-Based Statistical Framework

To identify common insertion sites (CISs), i.e., regions in the genome that are significantly more frequently mutated by insertions than would be expected by chance, we used a statistical framework based on Gaussian kernel convolution (GKC), which estimates a smoothed density distribution of inserts over the entire genome (Figure S2A) (de Ridder et al., 2006). Depending on kernel size and p value, the total number of statistically significant CISs varies (Figure 1C, Table S1). Increasing kernel size may result in merging independent CISs that may influence the same gene (as illustrated for the Myc locus in Figure 1D). Smaller kernels sizes may, conversely, reveal separate CISs that affect the same gene, which may be preferentially mutated in specific genetic backgrounds or result in expression of different mutant proteins (see below). Unless stated otherwise a kernel size of 30 kb was used in this paper.

Large-Scale Identification of CISs Near Known and Unknown Cancer Genes

Applying the GKC framework to the 10,806 insertions from the p53−/−, p19ARF−/− and wild-type tumors identified 346 CISs (p value < 0.05) (Figure 1E). By comparison, similar analysis of the MuLV insertion data present in the RTCGD database identified 160 CISs from 5435 insertions, 54 of these being shared between these datasets (Figure S2B). Combined analysis of both datasets yielded 473 CISs. To select genes that are the most likely candidate cancer genes, we used the mechanisms of retroviral mutation described by Jonkers and Berns, 1996 (Table S2). Established proto-oncogenes like Myc, Nmyc1, Ccnd3, Pim1 and Notch1, rank among the most significant CISs of this screen (Figure 1E, Table S2). In addition, established cancer genes like Lmo2, RhoH, Trim33, Mll, and Hspca (Futreal et al., 2004) are candidate target genes of less frequently mutated CISs, indicating that lower ranked CISs (carrying 4–5 insertions) may represent bona fide cancer genes. Importantly, we found a large number of highly significant CISs near genes that have not previously been linked to tumorigenesis, including Pik3r5 and Pik3cd, both regulatory subunits of PI3K that upon retroviral activation may enhance PI3K signaling (Brock et al., 2003), miRNA genes like e. g. mmu-mir-24-2/27a/23a and mmu-mir-142, Lunatic Fringe (Lfng), a modulator of Notch receptor activity (Haines and Irvine, 2003) and Smg6 (Est1a) which is involved in nonsense-mediated RNA decay (Fukuhara et al., 2005).

CISs Enriched for Established Cancer Genes

To determine what proportion of our CISs corresponds to verified human cancer genes, we compared our CISs with lists of known cancer genes. The Cancer Gene Census is an actively curated list of genes the mutation of which contributes to tumor formation (Futreal et al., 2004). Another list of mutated genes has been derived from sequencing the coding regions of a panel of breast and colon tumors (Sjoblom et al., 2006). From these, we composed a list of 516 unambiguous murine orthologs of human cancer genes. Depending on window size around the CIS midpoint (+/− 50 kb to 300 kb), we found between 30–79 CISs in the vicinity of murine orthologs of human cancer genes, a highly significant enrichment compared to an equal number of randomly selected genes (p value ≈0 for all window sizes). This indicates that CISs are found in the vicinity of cancer genes more frequently than expected by chance. We then used this overlap as a measure of the saturation of our screen i.e., to see how many of these cancer genes might have been identified using fewer tumors. Subsets of the total dataset were selected by stepwise addition of randomly selected groups of 50 mice. CISs were then identified for these random subsets, and we determined the number of known cancer gene orthologs within 200 kb either side of each CIS midpoint. The number of cancer genes identified does not reach its maximum until all the tumors are included (Figure S3), indicating that significant numbers of known (and by extension unknown) cancer genes might still be found by performing more extensive studies with larger amounts of tumors.

Insertions Mutate Pathways Involved in Human Cancer

To identify signaling pathways activated by retroviral insertions, we took the 346 candidate target genes and used Ingenuity Pathway Analysis software to investigate whether these genes play a role in canonical signaling pathways. Our 346 candidate genes were most significantly enriched for genes implicated in T cell and B cell receptor signaling, GM-CSF signaling and IL-2 signaling (Table S3), which may have been expected since all tumors were of lymphoid origin. In addition, we find significant enrichment for mutation of genes involved in ERK/MAPK, PI3K/AKT and G1/S checkpoint regulation. Genes in these pathways are also commonly mutated in human cancers, indicating that the signaling pathways mutated in our screen overlap with signaling pathways deregulated in human cancer (Table S3, Cancer Gene Census genes). Selection of the 346 candidate genes may be biased by manual curation. To rule out possible biases, we also compiled a list of the nearest gene to the midpoint of each of the 346 CISs. This yielded similar results, demonstrating that these results are not due to manual curation of CIS candidate target genes.

Identification of Genes Collaborating with p19ARF or p53 Deficiency

To identify genes that specifically collaborate with deficiency for either p19ARF or p53, the dataset was analyzed in two ways: (a) CISs were determined using all the insertions from the three panels together and CIS-genotype interactions were identified by comparing the number of insertions in a CIS in one genotype versus another, or (b) CISs were determined per panel, and candidate genes of the CISs were compared between panels (see Figure 2A).

Figure 2
Genotype-Specific Distribution of Insertions

Using the first approach (a), with all panels combined, we found 21 CISs (20 genes) with a significant bias (p < 0.05) toward one of the genotypes (Figure 2B, Table 1). For example, Runx1, Ccnd3, the miRNA cluster mmu-mir-106a-363 (encoding mmu-mir-106a/20b/19b-2/92-2 and 363), Flt3 and Smg6 are preferentially mutated in p19ARF−/− tumors compared to wild-type tumors. Runx1, mmu-mir-106a-363, Ccnd3, and Flt3, but not Smg6, are also specific for p53−/− compared to wild-type, indicating that mutation of these genes is selected for in cells that lack the p19ARF-p53 tumor suppression pathway. In contrast, mutation of Smg6 is highly specific for p19ARF−/− (p19ARF−/− versus wild-type p value = 0.0047, p19ARF−/− versus p53−/− p value = 0.0003) and may therefore only contribute to tumorigenesis in a p19ARF-deficient background. P53 can inhibit tumor development in the absence of p19ARF. We find that Notch1, which is able to suppress p53 activity through a p19ARF-independent, MDM2-dependent pathway (Beverly et al., 2005), is more frequently mutated in wild-type and p19ARF−/− compared to p53−/− (Figure 2B), suggesting that activation of Notch1 in p19ARF−/− and wild-type tumors might be instrumental for suppression of p53 activity. Similar numbers of genes are genotype-specific for p53−/− compared to p19ARF−/− and p53−/− compared to wild-type. However, genes preferentially mutated in p53−/− versus wild-type collaborate more strongly with p53 deficiency (yield lower p values) than genes preferentially mutated in p53−/− versus p19ARF−/−.

Table 1
Genotype-Specific 30 kb CISs

In our second approach (b), we determined CISs per panel, thus analyzing insertions that were retrieved from tumors isolated from mice with the same germ-line genotype. We find 113 CISs in p19ARF−/− tumors, 85 in p53−/− and 87 in wild-type tumors. The Venn diagram (Figure 2C) shows that 25 CISs near frequently mutated genes like Myc, Gfi1, Rasgrp1 and Rras2 are found in all panels (Figure 2C, Table S4). However, all panels also have a relatively large number of unique CISs, including 17 new CISs that were not found when insertions from the different panels were analyzed together. Fourteen of these newly identified loci are exclusively mutated in one of the three genotypes (Table S4). To determine the activated canonical signaling pathways, we analyzed candidate target genes of CISs unique for a single genotype (e.g., p19ARF−/− but not wild-type and p53−/−) with Ingenuity Pathway Analysis. We also analyzed all CISs that are found in one or two genotypes, but not the third (e.g., p19ARF−/− and/or p53−/− but not wild-type). We found that genes implicated in p53 signaling such as Akt1, Bcl2l1, Gadd45b and Ccnd1 are mutated in wild-type or p19ARF−/− tumors but not in p53−/− tumors (Tables S4 and S5). Activation of Akt1 can induce MDM2-mediated degradation of p53 and may thereby suppress p53-dependent effects on tumor formation (Mayo and Donner, 2001). Bcl2l1 can inhibit p53-induced apoptosis (Eischen et al., 2001), whereas Gadd45 has been demonstrated to have anti-apoptotic activity in hematopoietic cells in response to genotoxic stress (Gupta et al., 2005). Thus, activation of these genes in wild-type and p19ARF−/− tumors may abrogate p53-mediated apoptosis and contribute to tumor formation.

Creating an Interaction Map for All CISs

Previous analyses of insertions in the RTCGD have noted that some CISs appear to cooperate in oncogenesis (Dave et al., 2004; de Ridder et al., 2007). We looked for CIS-CIS interactions by determining whether particular CISs are found mutated together in the same tumor at higher or lower rates than expected by chance. We performed analysis of all pairs of CISs, using contingency tables that assume one CIS is the predisposing event, whereas inserts in the second CIS are assumed to be subsequent events. The reciprocal assumption was also tested.

The set of interactions between all 300 kb CISs is depicted in Figure S4, with Table S6 ranking the most significant interactions by p value. Interactions between the top 25 300 kb CISs are depicted as a heat map in Figures 3A and 3B and as a network in Figure 3C. The symmetry observed over the diagonal of these plots indicates that the contingency table tests yield similar results regardless of which insert is assumed to have occurred earlier. Nonetheless, some CIS pairs do break from this trend; for instance, insertions mutating Notch1 in the presence of an existing Rai17 insertion (p value = 0.001) appear to be more strongly selected for than insertions mutating Rai17 in the presence of an existing Notch1 insertion (0.0356).

Figure 3
Mapping Interaction Networks between Common Insertion Sites

Several of these interactions have some precedent within the literature. For example, Ikaros (Zfpn1a1) and Notch1 insertions frequently co-occur within the same tumor. Ikaros, a transcriptional regulator of hematopoietic differentiation, is deleted in acute lymphoblastic leukemia (ALL) (Sun et al., 1999) and has previously been identified as a CIS in a screen conducted in transgenic mice expressing the Notch1 intracellular domain (NotchIC) (Beverly and Capobianco, 2003). Myc and Nmyc1 insertions are mutually exclusive with each other and both loci are mutually exclusive with Notch1. These three loci may be functionally redundant because Myc and Nmyc1 are related in sequence and function, and Notch1 insertions can activate Myc expression (Sharma et al., 2007).

We also find a rationale within the literature for some of our novel interactions such as the comutation of Notch1 and Lfng. Lfng modulates the activity of the Notch1 receptor by glycosylation and frucosylation of the N-terminal extracellular EGF domains (Stirewalt and Radich, 2003). This suggests that in tumors where Notch1 is activated Lfng insertions can cooperate to enhance its activity.

Complete lists of interacting loci for 5 kb, 30 kb and 300 kb CISs with p values < 0.05 are included (Tables S7, S8, and S6, respectively). As expected for lower kernel widths, we find some adjacent CISs that may affect the same gene are mutually exclusive. For instance, some of the CISs near Myc using a 5 kb or 30 kb scale (Figures S5A and S5B) are mutually exclusive events. This may suggest that once an insert is obtained within a CIS there is apparently no selection for insertion within an adjacent CIS affecting the same gene, since this would be a redundant duplication of the same oncogenic function.

Separate Clusters of Insertions Create Distinct Flt3 and Notch1 Mutants

Using a kernel width of 5 kb, we found that insertions in frequently mutated genes like Flt3, Notch1, Jundm2 and Ikaros are unevenly distributed into clusters that may mutate the same genes by different mechanisms (Figure 4A). Two CISs were found in or near Flt3, a gene frequently mutated in human hematopoietic malignancies (Stirewalt and Radich, 2003). One CIS (CIS25 kb) upstream of the gene has 4 insertions where the retroviral genome is inserted in the antisense direction relative to the Flt3 transcript, suggesting that they act by an enhancer effect on the Flt3 promoter. The other CIS (CIS15 kb) resides in Flt3 intron 9–10. In tumors with CIS15 kb insertions, RT-PCR identified chimeric transcripts that fuse the MuLV transcript to Flt3 coding sequences and encode an N-terminally truncated protein of approximately 65 kDa (Figure 4B and data not shown). Western blot analysis showed high levels of a mutant Flt3 protein of 65kd in tumors having an insertion in Flt3 intron 9-10 but not in tumors with insertions upstream of Flt3 or normal thymus tissue. We also found that tumors with insertions upstream of Flt3 (CIS25 kb) are significantly enriched for insertions near Evi1 and Ets1, but this is not the case for tumors with Flt3 truncating insertions (CIS15 kb) (Figure 4C). Rather, truncated Flt3 is mutually exclusive with mutation of Myc, Gfi1, and Rasgrp1, whereas there is no selection against mutation of these genes in tumors with upstream Flt3 insertions.

Figure 4
Clusters of Insertions within Flt3 and Notch1 Induce Oncogenic Proteins with Distinct Properties

Similarly, CISs found within and upstream of Notch1 suggest that Notch1 may also give rise to functionally distinct oncogenic mutants (Figure 4A). CIS45 kb, upstream of Notch1, most likely enhances Notch1 expression, whereas CIS35 kb in intron 2 may induce overexpression of full length Notch1 or mutant Notch1 proteins that lack approximately 50 N-terminal amino acids. The other CISs in Notch1 give rise to a constitutively active mutant protein consisting of the intracellular domain of Notch1 (NotchIC) (CIS25 kb), or remove the destabilizing COOH-terminal PEST-domain and thereby increase Notch1 activity (CIS15 kb) (Hoemann et al., 2000; Weng et al., 2004). Both NotchIC and Notch1 lacking the PEST-domain closely resemble NOTCH1 mutants found in human cancers (Weng et al., 2004). Interestingly, NotchIC (CIS25 kb) mutations are almost exclusively found in wild-type tumors (wild-type versus p19ARF−/−, 11/3, p value = 0.006, wild-type versus p53−/−, 11/2, p value = 0.011) indicating that expression of the Notch1IC mutant may be particularly oncogenic in the wild-type background. Moreover, mutation of Ikaros is strongly selected for in tumors that have a NotchIC mutation (p = 7.6 × 10−5) but no significant co-occurrence is found with other Notch1 mutations. Lastly, in contrast to CIS15 kb and CIS35 kb, mutations in CIS25 kb (NotchIC) do not co-occur with Lfng mutations, most likely because NotchIC does not contain the N-terminal extracellular EGF domains and therefore will not be activated by increased Lfng levels (Figure 4C). Together, these data illustrate that increasing the coverage of insertional mutagenesis screens and analysis with a range of kernel sizes is informative even for CIS genes that have previously been established as cancer genes.

Using Intragenic Insertions to Identify Tumor Suppressor Genes

In some cases, insertions are selected for because they disrupt and inactivate tumor suppressor genes (Suzuki et al., 2006; reviewed in Uren et al., 2005). 4700 of our 10,806 inserts land within the transcribed regions of genes (Table S9). Some of these mutations are within known oncogenes where truncating or enhancer mutations within the gene are oncogenic (as previously discussed for Flt3 and Notch1). However, we also find known tumor suppressor loci.

The most prevalent known tumor suppressor on this list is Ikaros (Zfpn1a1), harboring 50 insertions. Another family member, Zfpn1a3 (Aiolos), harbors 11 insertions and like Ikaros is also implicated in ALL (Mullighan et al., 2007). Other known tumor suppressor loci also carry disrupting insertions including NfI (20 insertions), Ovca2 (6) and Wwox (7). Candidates of particular note are: E2f2 (which can act as a haploinsufficient tumor suppressor of Myc-induced lymphoma in mice (Opavsky et al., 2007)), Raptor (a binding partner and inhibitor of mTOR (target of rapamycin) (Kim et al., 2002; Hara et al., 2002)), Nfatc3 (found in a previous retroviral screen to suppress SL3-3 induced lymphoma (Glud et al., 2005) and also found to suppress mammary adenocarcinoma (Lee et al., 2005)), Xrcc6 (Ku70) (knockouts of which develop thymic and disseminated T cell lymphomas [Li et al., 1998]) and Ablim1, which is located in a chromosomal region frequently lost in human tumors (Kim et al., 1997)(Figure S6). Notably, we observe few inserts within commonly mutated tumor suppressors such as Cdkn2a, Pten and Rb1. These genes may be poor targets for integration due to insertion site preferences of the virus or inability of the virus to inactivate these genes by insertion.

We further postulated that finding multiple intragenic insertions within both alleles of a gene is more likely for tumor suppressor genes than oncogenes. 55 genes carry more than one intragenic insertion within the same tumor (Table 2). To estimate the significance of such events, we compared how often genes were hit twice in real data versus 100,000 permutations of randomized data (shuffling inserts between tumors) and ranked them by this p value.

Table 2
Genes with Multiple Intragenic Insertions

Ikaros (Zfpn1a1) is disrupted in 33 tumors, 11 of which carry more than one intragenic insertion. After shuffling, we find on average only 3 tumors carry more than one insertion within Ikaros (p value = 1 x 10−5). By comparison, the known oncogene CyclinD3 (with 148 inserts) is hit more than once in 25 tumors, however this number is similar to that expected by chance (in randomized data 22 tumors have more than one hit, yielding a p value of 0.2). Other known or candidate tumor suppressors with more hits per tumor than expected are indicated in Table 2, including Mobkl2a (hMOB1), Ablim1, Adrbk1 and Nf1 (p values 0.02, 0.03, 0.04, 0.08, respectively). We cannot rule out that there may be selection for multiple insertions within the same gene in a single tumor clone even when these insertions are activating mutations. Nonetheless the most established oncogenes of this list (AhiI, Evi1, Pim1, Flt3, Gfi1, Evi5, Notch1, and Ccnd3) are those with higher p values. Thus, it appears that genes with lower p values in this analysis are better tumor suppressor candidates than those with higher p values.

Identifying Candidate Cancer Genes by Comparison of Human Tumor Amplicons and Deletions with Murine Retroviral Insertions

As part of the Wellcome Trust Sanger Institute's Cancer Genome Project, 713 human cell lines were hybridized to 10,000 probe SNP arrays and copy number information was extracted from these data. Some amplicons and deletions span megabases and contain many genes, thus making cancer gene identification difficult. To this end, we mapped our CIS loci to their orthologous loci in the human genome and looked for overlap with amplicons and deletions in the tumor cell lines. The human orthologs of our list of CIS candidate target genes were found to be amplified at a significantly higher frequency than a random list of genes (p value = 0.03). Significant results were also obtained using the list of genes nearest to the midpoint of our CISs (p value = 0.006). Known oncogenes like Fgfr2, Kit and Evi1 and other notable examples identified in this manner are illustrated in Figure S7 and Figure 5.

Figure 5
Insertions Identify Tumor Suppressors and Oncogenes in Human Cancers

A recurrent amplicon on chromosome 6 is orthologous to a CIS in the vicinity of Rreb1 (Figure 5A). Rreb1 binds and represses expression of the p16(Ink4a) promoter, and the development of pristane-induced plasma cell tumors in Balb/C mice is attributable to a polymorphism in this Rreb1 binding site (Zhang et al., 2003).

Another amplified region on chromosome 1 contains at least 21 genes, of which Slamf6 appears to be the most likely target gene of our insertions (Figure 5B). Polymorphisms within the region of Slamf6 (Ly108) have been implicated in systemic lupus erythematosus (SLE) (Wandstrat et al., 2004).

A recurrent amplicon on Chromosome 11 contains at least 24 genes, including a cluster of genes encoding matrix metalloproteinases (MMPs). Several MMPs within this amplicon have been previously implicated in cancer, however the only one of these genes implicated by our insertions is Mmp13, which to date has not been ascribed a role in cancer (Figure 5C).

We did not see significant global overlap with our lists of tumor suppressor candidate genes and the deletions of the human CGH data. This is perhaps not surprising since, while these lists may be enriched for tumor suppressor genes, they still contain many oncogenes that have insertions within their transcribed regions. Nonetheless, we also find candidate tumor suppressor genes that overlap recurrent deletions. Wwox is disrupted by seven intragenic insertions and deleted within 10 of the human cell lines (Figure 5D). A novel candidate tumor suppressor emerging from this comparison is Arfrp2 (Arl15)(Figure 5E). Arfrp2 is a member of the ADP-ribosylation factor-like family. Notably, another member of this family, ARL11 (ARLTS1), is a tumor suppressor gene where truncating germline mutations or promoter methylation contribute to leukemia, breast cancer, ovarian cancer, and melanoma (Sych et al., 1978; Frank et al., 2005; Petrocca et al., 2006).

Discussion

Here, we report on a large-scale retroviral insertional mutagenesis screen using more than 500 tumors from p19ARF−/−, p53−/− and wild-type mice. This scale of analysis allows identification of a high number of new candidate oncogenes and tumor suppressors and detects highly significant combinations of co-occurring or mutually exclusive genes. Notably, a significant proportion of CISs that map to orthologs of known human cancer loci are not only identified in hematopoietic tumors, but also in other tumor types such as lung, colon, breast, and prostate tumors.

We have identified 25 CISs that were significantly more mutated in one of the germline genotypes compared to either one or both of the two other genotypes. Together with p53 and RAS, FLT3 is the most commonly mutated gene in human AML, occurring in approximately 25% of cases (Stirewalt and Radich, 2003). Flt3 is almost exclusively mutated in p53−/− and p19ARF−/− tumors and not in wild-type tumors (11, 13 and 2 insertions, respectively), indicating that Flt3 mutations may be particularly oncogenic in the absence of a functional p19ARF-MDM2-p53 pathway. Flt3 induces Ras signaling (Stirewalt and Radich, 2003), and mutation of Flt3 is mutually exclusive with Ras mutations in human cancers suggesting that mutation of Flt3 has similar effects as Ras mutation (Stirewalt et al., 2001). We find that mutations in the Ras-activating Rasgrp1 (Ebinu et al., 2000) are mutually exclusive with mutation of Flt3, also suggesting that activation of Ras signaling may be an important effect of Flt3 mutation in tumors.

The utility of retroviral insertions is further illustrated by comparison to array CGH data from cancer cell lines. We identify several novel candidate cancer genes including Rreb1, Mmp13 and Arfrp2 (Arl15). Similarly useful comparisons can also be envisaged for tumor resequencing data. Human populations carry many polymorphisms, only some of which contribute to tumor susceptibility and many tumors have a mutator phenotype that creates a background of irrelevant mutations. In such studies polymorphisms and background mutations can only be distinguished from oncogenic mutations by stringent statistics that may inadvertently exclude rare but genuine oncogenic events. CIS loci should be a useful tool to help focus future resequencing studies, as illustrated by a recent study which finds overlap between mutations in coding regions of human breast cancer genes and CIS loci from an MMTV insertional mutagenesis study (Theodorou et al., 2007; Wood et al., 2007). As such, CIS associations could be used to prioritize which rarely mutated genes should be sequenced within a tumor set i.e., identifying the commonly mutated genes in a set of tumors and then sequencing their known collaborators from insertional mutagenesis screens.

Due to the tropism of MuLV, our screen is limited in its ability to identify cancer genes from nonhematopoietic tumors. As the technology of transposon mutagenesis matures a greater range of tumor types will become amenable to analysis by insertional mutagenesis. New models notwithstanding, there is still much to be gained from traditional MuLV screens. Larger numbers of inserts and better estimates of tumor clonality will improve the power of association studies, giving rise to interaction maps that are denser and more informative. Even if complete saturating coverage of each tumor is possible, it remains unclear how many more loci might be identified by increasing the number of tumors analyzed. Certainly it appears that screening a greater variety of genotypes and predisposing mutations will expand the range of mutations found. Also, given the differences in susceptibility for tumor development between different inbred mouse strains, it would be interesting to compare the spectrum of oncogenic mutations between these strains. Analysis of MuLV tumors by expression arrays and by linker-mediated PCR on cDNA of all tumors will increase the accuracy with which target genes of insertions can be identified. CGH copy number arrays and ORF resequencing might also identify cooperating mutations that are less amenable to mutation by insertion.

Ultimately there may also be clinical applications of mutually exclusive interactions of CIS loci. Recent studies have indicated that EGFR mutations are mutually exclusive with KRAS mutations in lung adenocarcinomas (Pao et al., 2005). Mutant EGFR is a target for gefitinib and erlotinib, and patients with mutations in KRAS instead of EGFR do not see any benefit from the use of these drugs. In cases where the target of a drug is unknown or unclear, knowing the associations of nontarget mutations that correlate with treatment outcomes can inform the search for the actual target and/or suggest novel indications for established therapies.

Experimental Procedures

MuLV Infection of p19ARF−/−, p53−/−, and Wild-Type Mice

Crosses between F0 parental FVB mice +/− for the p19ARF knockout allele (Kamijo et al., 1997) were performed to generate F1 p19ARF−/−, p19ARF+/− and p19ARF+/+ (wild-type) offspring. F0 parental FVB mice+/− for the p53 knockout allele (Donehower et al., 1992) were crossed to generate p53−/−, p53+/− and p53+/+ F1 offspring. Newborn F1 pups were injected i.p. with 1.105 infectious units of MuLV. Animals were monitored in time for the development of tumors, moribund mice were sacrificed and tumors were isolated. All animal experiments were done conform national regulatory standards approved by the DEC (Animal Experiments Committee).

Identification of Retroviral Insertion Sites

Genomic DNA was isolated using PureGene from Gentra Systems, Inc. Insertions sites were identified using an linker-mediated PCR protocol adapted from Mikkers et al. (Mikkers et al., 2002). PCR products were shotgun subcloned using a protocol developed at the Wellcome Trust Sanger Institute (both protocols available on http://mutapedia.nki.nl/).

Mapping of Insertions

Detailed instructions and parameters used for informatics are available on request. Briefly, we used cross_match (Dr. Philip Green, unpublished data) to identify vector, primer, linker and viral U5 LTR sequences in the reads. We used SSAHA2 (Ning et al., 2001) to map individual reads onto the mouse genome (NIH Build 34). Chimeric and concatameric subcloning products were separated where possible or else discarded. Sequences containing a splinkerette but lacking an LTR were discarded, unless they were within 2000 bp of contigs that containing an LTR-genome junction, in which case they were added to these contigs where orientation was consistent. A final round of filtering to remove PCR artifacts was then applied.

Software

Pathway analysis was performed using Ingenuity Pathway Analysis software (Ingenuity® Systems, www.ingenuity.com). Cytoscape® software (version 2.4.0, www.cytoscape.org) was used to create the CIS interaction network.

Identification of CISs Near Orthologs of Human Cancer Genes

Each of the 30 kb CISs was examined for the presence of a murine ortholog of a human cancer gene within 200 kb up and downstream of the CIS peak. The number of unique orthologs was counted. Next, the number of insertions in the dataset was decreased step-wise to zero by removing insertions from 50 randomly selected mice. CIS positions are determined each time 50 mice are taken out of the dataset, as well as the number of unique orthologs. The experiment was repeated 20 times, average and standard deviation of the number of murine orthologs was determined. To reduce computation time required for the saturation analysis, CISs were determined using an alternative method that applies a more stringent α-level then GKC to determine CISs and therefore detects less CISs.

Statistical Methods

Significance of genotype-CIS interactions and CIS-CIS interactions were estimated by permutation. Briefly, inserts were placed within a 2 × 2 contingency table based on whether the insert is found within a given CIS and either the genotype of the mouse or the presence or absence of insertions from another CIS within the same mouse. Similar tables were constructed for 100,000 random permutations of data where the entire set of insertions was shuffled between all tumors. A Chi squared test statistic was calculated for each of the real and permuted tables. p values were calculated as the proportion of permuted test statistics that are greater than the real test statistic.

CGH Comparison

We obtained 10K SNP array CGH data for 713 human cancer cell lines from the Wellcome Trust Sanger Institute (ftp://ftp.sanger.ac.uk/pub/cgp/10kdata). We identified regions of copy number change in each cell line and identified human orthologs of mouse CIS candidate genes overlapping with amplicons and deletions in the human cancer cell lines. For further details see Supplemental Experimental Procedures.

Transfection and Plasmids

Phoenix cells were transfected using CaPO4 precipitation. The LXSN-Flt3 vector used as a positive control for Flt3 expression was obtained from Dr. Olivier Rosnet, Inserm, Marseille, France.

Protein Extraction and Western Blotting

Protein was isolated from frozen tumor material or Phoenix cells and loaded on SDS-PAGE gels. Antibodies used are Flt3 (8F2) (Cell Signaling Technology, Inc) and alpha tubulin (Sigma-Aldrich, Inc). A protein lysate from a thymus isolated from a 40-day-old wild-type mouse was used as control.

Acknowledgments

Work in A. Berns's and M. van Lohuizen's lab was funded by the NWO Genomics program and the Netherlands Genomics Initiative/Netherlands Organization for Scientific Research (NWO). J.d.R. was supported by the BioRange program of the Netherlands Bioinformatics Centre (NBIC), which is supported by a BSIK grant through the Netherlands Genomics Initiative (NGI). M.v.U. was supported by a Horizon Breakthrough Project grant from the NGI. All sequencing was carried out by the Wellcome Trust Sanger Institute sequencing facility. Work in the D.J. Adams laboratory is funded by Cancer Research-UK and the Wellcome Trust (76943). We would like to thank Richard Wooster and members of the Sanger Institute's Cancer Genome Project for generating the CGH data and for assistance in the analysis of this dataset. We would like to thank Keiko Akagi for providing us with RTCGD data. We thank the staff of the animal facility at the Netherlands Cancer Institute for animal husbandry and Bas van Steensel and Colin Pritchard for critically reading the manuscript.

Supplemental Data

Document S1. Seven Figures and Six Tables:
Document S2. Table S6, Table S7, and Table S8:

References

Akagi K., Suzuki T., Stephens R.M., Jenkins N.A., Copeland N.G. RTCGD: retroviral tagged cancer gene database. Nucleic Acids Res. 2004;32:D523–D527. [PMC free article] [PubMed]
Ben David Y., Prideaux V.R., Chow V., Benchimol S., Bernstein A. Inactivation of the p53 oncogene by internal deletion or retroviral integration in erythroleukemic cell lines induced by Friend leukemia virus. Oncogene. 1988;3:179–185. [PubMed]
Beverly L.J., Capobianco A.J. Perturbation of Ikaros isoform selection by MLV integration is a cooperative event in Notch(IC)-induced T cell leukemogenesis. Cancer Cell. 2003;3:551–564. [PubMed]
Beverly L.J., Felsher D.W., Capobianco A.J. Suppression of p53 by Notch in lymphomagenesis: implications for initiation and regression. Cancer Res. 2005;65:7159–7168. [PubMed]
Brock C., Schaefer M., Reusch H.P., Czupalla C., Michalke M., Spicher K., Schultz G., Nurnberg B. Roles of G beta gamma in membrane recruitment and activation of p110 gamma/p101 phosphoinositide 3-kinase gamma. J. Cell Biol. 2003;160:89–99. [PMC free article] [PubMed]
Christophorou M.A., Ringshausen I., Finch A.J., Swigart L.B., Evan G.I. The pathological response to DNA damage does not contribute to p53-mediated tumour suppression. Nature. 2006;443:214–217. [PubMed]
Cuypers H.T., Selten G., Quint W., Zijlstra M., Maandag E.R., Boelens W., van Wezenbeek P., Melief C., Berns A. Murine leukemia virus-induced T-cell lymphomagenesis: integration of proviruses in a distinct chromosomal region. Cell. 1984;37:141–150. [PubMed]
Dave U.P., Jenkins N.A., Copeland N.G. Gene therapy insertional mutagenesis insights. Science. 2004;303:333. [PubMed]
de Ridder J., Kool J., Uren A., Bot J., Wessels L., Reinders M. Co-occurrence analysis of insertional mutagenesis data reveals cooperating oncogenes. Bioinformatics. 2007;23:i133–i141. [PubMed]
de Ridder J., Uren A., Kool J., Reinders M., Wessels L. Detecting Statistically Significant Common Insertion Sites in Retroviral Insertional Mutagenesis Screens. PLoS Comput Biol. 2006;2:e166. [PMC free article] [PubMed]
Devon R.S., Porteous D.J., Brookes A.J. Splinkerettes–improved vectorettes for greater efficiency in PCR walking. Nucleic Acids Res. 1995;23:1644–1645. [PMC free article] [PubMed]
Donehower L.A., Harvey M., Slagle B.L., McArthur M.J., Montgomery C.A., Jr., Butel J.S., Bradley A. Mice deficient for p53 are developmentally normal but susceptible to spontaneous tumours. Nature. 1992;356:215–221. [PubMed]
Ebinu J.O., Stang S.L., Teixeira C., Bottorff D.A., Hooton J., Blumberg P.M., Barry M., Bleakley R.C., Ostergaard H.L., Stone J.C. RasGRP links T-cell receptor signaling to Ras. Blood. 2000;95:3199–3203. [PubMed]
Efeyan A., Garcia-Cao I., Herranz D., Velasco-Miguel S., Serrano M. Tumour biology: Policing of oncogene activity by p53. Nature. 2006;443:159. [PubMed]
Eischen C.M., Weber J.D., Roussel M.F., Sherr C.J., Cleveland J.L. Disruption of the ARF-Mdm2-p53 tumor suppressor pathway in Myc-induced lymphomagenesis. Genes Dev. 1999;13:2658–2669. [PMC free article] [PubMed]
Eischen C.M., Woo D., Roussel M.F., Cleveland J.L. Apoptosis triggered by Myc-induced suppression of Bcl-X(L) or Bcl-2 is bypassed during lymphomagenesis. Mol. Cell. Biol. 2001;21:5063–5070. [PMC free article] [PubMed]
Frank B., Klaes R., Burwinkel B. Familial cancer and ARLTS1. N. Engl. J. Med. 2005;353:313–314. [PubMed]
Fukuhara N., Ebert J., Unterholzner L., Lindner D., Izaurralde E., Conti E. SMG7 is a 14–3-3-like adaptor in the nonsense-mediated mRNA decay pathway. Mol. Cell. 2005;17:537–547. [PubMed]
Futreal P.A., Coin L., Marshall M., Down T., Hubbard T., Wooster R., Rahman N., Stratton M.R. A census of human cancer genes. Nat. Rev. Cancer. 2004;4:177–183. [PMC free article] [PubMed]
Glud S.Z., Sorensen A.B., Andrulis M., Wang B., Kondo E., Jessen R., Krenacs L., Stelkovics E., Wabl M., Serfling E. A tumor-suppressor function for NFATc3 in T-cell lymphomagenesis by murine leukemia virus. Blood. 2005;106:3546–3552. [PMC free article] [PubMed]
Gupta M., Gupta S.K., Balliet A.G., Hollander M.C., Fornace A.J., Hoffman B., Liebermann D.A. Hematopoietic cells from Gadd45a- and Gadd45b-deficient mice are sensitized to genotoxic-stress-induced apoptosis. Oncogene. 2005;24:7170–7179. [PubMed]
Haines N., Irvine K.D. Glycosylation regulates Notch signalling. Nat. Rev. Mol. Cell Biol. 2003;4:786–797. [PubMed]
Hara K., Maruki Y., Long X., Yoshino K., Oshiro N., Hidayat S., Tokunaga C., Avruch J., Yonezawa K. Raptor, a binding partner of target of rapamycin (TOR), mediates TOR action. Cell. 2002;110:177–189. [PubMed]
Hoemann C.D., Beaulieu N., Girard L., Rebai N., Jolicoeur P. Two distinct Notch1 mutant alleles are involved in the induction of T-cell leukemia in c-myc transgenic mice. Mol. Cell. Biol. 2000;20:3831–3842. [PMC free article] [PubMed]
Hwang H.C., Martins C.P., Bronkhorst Y., Randel E., Berns A., Fero M., Clurman B.E. Identification of oncogenes collaborating with p27Kip1 loss by insertional mutagenesis and high-throughput insertion site analysis. Proc. Natl. Acad. Sci. USA. 2002;99:11293–11298. [PMC free article] [PubMed]
Johansson F.K., Brodd J., Eklof C., Ferletta M., Hesselager G., Tiger C.F., Uhrbom L., Westermark B. Identification of candidate cancer-causing genes in mouse brain tumors by retroviral tagging. Proc. Natl. Acad. Sci. USA. 2004;101:11334–11337. [PMC free article] [PubMed]
Jonkers J., Berns A. Retroviral insertional mutagenesis as a strategy to identify cancer genes. Biochim. Biophys. Acta. 1996;1287:29–57. [PubMed]
Kamijo T., Bodner S., van de Kamp E., Randle D.H., Sherr C.J. Tumor spectrum in ARF-deficient mice. Cancer Res. 1999;59:2217–2222. [PubMed]
Kamijo T., Zindy F., Roussel M.F., Quelle D.E., Downing J.R., Ashmun R.A., Grosveld G., Sherr C.J. Tumor suppression at the mouse INK4a locus mediated by the alternative reading frame product p19ARF. Cell. 1997;91:649–659. [PubMed]
Kelly-Spratt K.S., Gurley K.E., Yasui Y., Kemp C.J. p19Arf suppresses growth, progression, and metastasis of Hras-driven carcinomas through p53-dependent and -independent pathways. PLoS Biol. 2004;2:E242. [PMC free article] [PubMed]
Kim A.C., Peters L.L., Knoll J.H., van H.C., Ciciotte S.L., Kleyn P.W., Chishti A.H. Limatin (LIMAB1), an actin-binding LIM protein, maps to mouse chromosome 19 and human chromosome 10q25, a region frequently deleted in human cancers. Genomics. 1997;46:291–293. [PubMed]
Kim D.H., Sarbassov D.D., Ali S.M., King J.E., Latek R.R., Erdjument-Bromage H., Tempst P., Sabatini D.M. mTOR interacts with raptor to form a nutrient-sensitive complex that signals to the cell growth machinery. Cell. 2002;110:163–175. [PubMed]
Lee H., Chouinard L., Bonin M., Michel R.N. NFATc3 deficiency may contribute to the development of mammary gland adenocarcinoma in aging female mice. Mol. Carcinog. 2005;44:219–222. [PubMed]
Levine A.J., Hu W., Feng Z. The P53 pathway: what questions remain to be explored? Cell Death Differ. 2006;13:1027–1036. [PubMed]
Li G.C., Ouyang H., Li X., Nagasawa H., Little J.B., Chen D.J., Ling C.C., Fuks Z., Cordon-Cardo C. Ku70: a candidate tumor suppressor gene for murine T cell lymphoma. Mol. Cell. 1998;2:1–8. [PubMed]
Li J., Shen H., Himmel K.L., Dupuy A.J., Largaespada D.A., Nakamura T., Shaughnessy J.D., Jr., Jenkins N.A., Copeland N.G. Leukaemia disease genes: large-scale cloning and pathway predictions. Nat. Genet. 1999;23:348–353. [PubMed]
Lund A.H., Turner G., Trubetskoy A., Verhoeven E., Wientjens E., Hulsman D., Russell R., DePinho R.A., Lenz J., van Lohuizen M. Genome-wide retroviral insertional tagging of genes involved in cancer in Cdkn2a-deficient mice. Nat. Genet. 2002;32:160–165. [PubMed]
Mayo L.D., Donner D.B. A phosphatidylinositol 3-kinase/Akt pathway promotes translocation of Mdm2 from the cytoplasm to the nucleus. Proc. Natl. Acad. Sci. USA. 2001;98:11598–11603. [PMC free article] [PubMed]
Mikkers H., Allen J., Knipscheer P., Romeijn L., Hart A., Vink E., Berns A. High-throughput retroviral tagging to identify components of specific signaling pathways in cancer. Nat. Genet. 2002;32:153–159. [PubMed]
Moore L., Venkatachalam S., Vogel H., Watt J.C., Wu C.L., Steinman H., Jones S.N., Donehower L.A. Cooperativity of p19ARF, Mdm2, and p53 in murine tumorigenesis. Oncogene. 2003;22:7831–7837. [PubMed]
Morishita K., Parker D.S., Mucenski M.L., Jenkins N.A., Copeland N.G., Ihle J.N. Retroviral activation of a novel gene encoding a zinc finger protein in IL-3-dependent myeloid leukemia cell lines. Cell. 1988;54:831–840. [PubMed]
Mullighan C.G., Goorha S., Radtke I., Miller C.B., Coustan-Smith E., Dalton J.D., Girtman K., Mathew S., Ma J., Pounds S.B. Genome-wide analysis of genetic alterations in acute lymphoblastic leukaemia. Nature. 2007;446:758–764. [PubMed]
Ning Z., Cox A.J., Mullikin J.C. SSAHA: a fast search method for large DNA databases. Genome Res. 2001;11:1725–1729. [PMC free article] [PubMed]
Opavsky R., Tsai S.Y., Guimond M., Arora A., Opavska J., Becknell B., Kaufmann M., Walton N.A., Stephens J.A., Fernandez S.A. Specific tumor suppressor function for E2F2 in Myc-induced T cell lymphomagenesis. Proc. Natl. Acad. Sci. USA. 2007;104:15400–15405. [PMC free article] [PubMed]
Pao W., Wang T.Y., Riely G.J., Miller V.A., Pan Q., Ladanyi M., Zakowski M.F., Heelan R.T., Kris M.G., Varmus H.E. KRAS mutations and primary resistance of lung adenocarcinomas to gefitinib or erlotinib. PLoS Med. 2005;2:e17. [PMC free article] [PubMed]
Petrocca F., Iliopoulos D., Qin H.R., Nicoloso M.S., Yendamuri S., Wojcik S.E., Shimizu M., Di Leva G., Vecchione A., Trapasso F. Alterations of the tumor suppressor gene ARLTS1 in ovarian cancer. Cancer Res. 2006;66:10287–10291. [PubMed]
Schmitt C.A., McCurrach M.E., de Stanchina E., Wallace-Brodeur R.R., Lowe S.W. INK4a/ARF mutations accelerate lymphomagenesis and promote chemoresistance by disabling p53. Genes Dev. 1999;13:2670–2677. [PMC free article] [PubMed]
Sharma V.M., Draheim K.M., Kelliher M.A. The Notch1/c-Myc pathway in T cell leukemia. Cell Cycle. 2007;6:927–930. [PubMed]
Sharpless N.E. INK4a/ARF: a multifunctional tumor suppressor locus. Mutat. Res. 2005;576:22–38. [PubMed]
Sherr C.J. Divorcing ARF and p53: an unsettled case. Nat. Rev. Cancer. 2006;6:663–673. [PubMed]
Sjoblom T., Jones S., Wood L.D., Parsons D.W., Lin J., Barber T.D., Mandelker D., Leary R.J., Ptak J., Silliman N. The consensus coding sequences of human breast and colorectal cancers. Science. 2006;314:268–274. [PubMed]
Slape C., Hartung H., Lin Y.W., Bies J., Wolff L., Aplan P.D. Retroviral insertional mutagenesis identifies genes that collaborate with NUP98-HOXD13 during leukemic transformation. Cancer Res. 2007;67:5148–5155. [PMC free article] [PubMed]
Stewart M., Mackay N., Hanlon L., Blyth K., Scobie L., Cameron E., Neil J.C. Insertional mutagenesis reveals progression genes and checkpoints in MYC/Runx2 lymphomas. Cancer Res. 2007;67:5126–5133. [PMC free article] [PubMed]
Stirewalt D.L., Kopecky K.J., Meshinchi S., Appelbaum F.R., Slovak M.L., Willman C.L., Radich J.P. FLT3, RAS, and TP53 mutations in elderly patients with acute myeloid leukemia. Blood. 2001;97:3589–3595. [PubMed]
Stirewalt D.L., Radich J.P. The role of FLT3 in haematopoietic malignancies. Nat. Rev. Cancer. 2003;3:650–665. [PubMed]
Sun L., Heerema N., Crotty L., Wu X., Navara C., Vassilev A., Sensel M., Reaman G.H., Uckun F.M. Expression of dominant-negative and mutant isoforms of the antileukemic transcription factor Ikaros in infant acute lymphoblastic leukemia. Proc. Natl. Acad. Sci. USA. 1999;96:680–685. [PMC free article] [PubMed]
Suzuki T., Minehata K., Akagi K., Jenkins N.A., Copeland N.G. Tumor suppressor gene identification using retroviral insertional mutagenesis in Blm-deficient mice. EMBO J. 2006;25:3422–3431. [PMC free article] [PubMed]
Suzuki T., Shen H., Akagi K., Morse H.C., Malley J.D., Naiman D.Q., Jenkins N.A., Copeland N.G. New genes involved in cancer identified by retroviral tagging. Nat. Genet. 2002;32:166–174. [PubMed]
Sych F., Habs M., Schmahl D. Chemotherapy studies in autochthonous rat tumors intestinal cancer. Z. Krebsforsch. Klin. Onkol. Cancer Res. Clin. Oncol. 1978;92:105–117. [PubMed]
Theodorou V., Kimm M.A., Boer M., Wessels L., Theelen W., Jonkers J., Hilkens J. MMTV insertional mutagenesis identifies genes, gene families and pathways involved in mammary cancer. Nat. Genet. 2007;39:759–769. [PubMed]
Uren A.G., Kool J., Berns A., van Lohuizen M. Retroviral insertional mutagenesis: past, present and future. Oncogene. 2005;24:7656–7672. [PubMed]
Wandstrat A.E., Nguyen C., Limaye N., Chan A.Y., Subramanian S., Tian X.H., Yim Y.S., Pertsemlidis A., Garner H.R., Jr., Morel L. Association of extensive polymorphisms in the SLAM/CD2 gene cluster with murine lupus. Immunity. 2004;21:769–780. [PubMed]
Weber J.D., Jeffers J.R., Rehg J.E., Randle D.H., Lozano G., Roussel M.F., Sherr C.J., Zambetti G.P. p53-independent functions of the p19(ARF) tumor suppressor. Genes Dev. 2000;14:2358–2365. [PMC free article] [PubMed]
Weng A.P., Ferrando A.A., Lee W., Morris J.P., Silverman L.B., Sanchez-Irizarry C., Blacklow S.C., Look A.T., Aster J.C. Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia. Science. 2004;306:269–271. [PubMed]
Wood L.D., Parsons D.W., Jones S., Lin J., Sjoblom T., Leary R.J., Shen D., Boca S.M., Barber T., Ptak J. The Genomic Landscapes of Human Breast and Colorectal Cancers. Science. 2007;318:1108–1113. [PubMed]
Zhang S., Qian X., Redman C., Bliskovski V., Ramsay E.S., Lowy D.R., Mock B.A. p16 INK4a gene promoter variation and differential binding of a repressor, the ras-responsive zinc-finger transcription factor, RREB. Oncogene. 2003;22:2285–2295. [PubMed]

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...