Retrotransposons can be used to deliver a gene trap that efficiently splices into genes and disrupts their function. This means that up to 30% of insertions (the percentage of the genome that comprises gene exons and introns) may disrupt a gene. This is a major improvement on chemical mutagenesis, which creates point mutations that often do not affect gene function. Also, unlike DNA transposon insertions, L1 insertions are permanent and stable, eliminating the possibility of losing the mutation over time.

A complete genome-wide mutagenesis strategy requires unbiased gene disruption. Although retrovirus-based mutagenesis strategies result in 'hotspots' of mutation 7, L1 retrotransposons insert into the genome without apparent site specificity. The L1 element contains an apurinic/apyrimidinic endonuclease with an inexact target site preference for 3'-AA/TTTT-5' 8910, which allows the L1 retrotransposon to insert randomly throughout the entire genome. Recent human L1 insertions (within the past few million years) have been found on every chromosome and are present in introns, in intergenic regions, and in DNA regions of high and low GC content 11. The randomness of de novo L1 insertions is well established by the study of recent insertions from the various genome sequencing projects, and we and others have confirmed random L1 insertion in mouse mutagenesis models 1213. Note that the random pattern of de novo insertions is distinct from the pattern that occurs after post-insertion selection bias has taken place, which causes L1 elements to accumulate into AT-rich and gene-poor regions. Post-insertion selection bias occurs over millions of years and does not affect the randomness of mutagenesis systems. Therefore, retrotransposon based mutagenesis offers a significant advantage over retrovirus based mutagenesis. It also offers an advantage over insertional mutagenesis using DNA transposons such as Sleeping Beauty (SB; a member of the Tc1/mariner transposon family) 14 and piggyback (a DNA transposon from the cabbage looper moth) 15, which insert non-randomly when used for in vivo mutagenesis. The majority of de novo transposon insertions occur within 3 megabases of their original genomic location and 80% occur on the same chromosome 16171819, a phenomenon termed 'local hopping'.

L1 retrotransposons have no limit on the number of mutations that they can cause because they replicate by a 'copy and paste' mechanism. This is in contrast to DNA transposons, which move by a 'cut and paste' mechanism and are limited by the number of transposons in the transgenic founder 20. In an ideal germline mutagenesis strategy, each germ cell would contain a single mutation that disrupts a single gene that leads to a phenotype. Fewer mutations introduce inefficiency, although more than one mutation per germ cell is not necessarily ideal because matching mutations to phenotypes becomes problematic. L1 retrotransposon based mutagenesis has a high frequency of mutation, approaching the ideal of one mutation per germ cell.

L1 retrotransposons demonstrate cis preference, which means that the retrotransposition proteins almost exclusively mobilize the RNA that encoded them 21. Therefore, when inserting a tagged L1 retrotransposon into an animal, one can be sure that all de novo insertions will be from tagged L1 retrotransposons and not from retrotransposition of other cellular mRNAs. Any new phenotypes will be due to a tagged insertion that can be easily mapped, as opposed to an untagged insertion that cannot be easily mapped. Unlike N-ethyl-N-nitrosourea (ENU) base substitutions, de novo retrotransposon insertions are large, unique sequences that are relatively easy to map using inverse polymerase chain reaction (PCR) or thermal asymmetric interlaced PCR. Although easy mapping of mutations is a feature shared with mutagenesis using transposons, the SB system has a drawback. Recent experiments demonstrated that de novo SB insertions occasionally occur via multiple sequential 'hops' from one chromosomal region to another 2223, a process that can result in large deletions. After SB excision three terminal base pairs of the transposon remain, and resolution of the double strand break occurs via the nonhomologous end joining repair pathway, resulting in a net 3 base pair insertion. Occasionally, through unknown mechanisms, deletions occur at SB excision sites at lengths of up to 100 base pairs 2425. Deletions created in this manner are not associated with the transposon that caused them because the transposon is excised during the event (cut and paste mechanism), and so these mutations cannot easily be mapped, which confounds attempts to link interesting phenotypes with the causative mutation. L1s can cause genomic deletions at insertion sites 2627, but these deletions remain associated with the L1 element that caused them because the retrotransposon is not subsequently excised (copy and paste mechanism) and so these mutations can be identified more easily.

Direct mutation in the germline produces offspring with gene disruptions by natural breeding. This is important because the use of embryonic stem cells can be costly, time consuming, and technically impossible in most mammalian models, including rat 28.

Table 1 provides a summary of the various methods of mutagenesis in rodents. Many of the advantages of mutagenesis using retrotransposons are shared with DNA transposons, except as noted above, and their use is discussed in other reviews in this supplement.

The first description of an endogenous L1 causing a disease in humans by insertional mutagenesis was reported in 1988. In that instance, an L1 had inserted into the gene encoding coagulation factor VIII, thereby causing hemophilia A 29. Since that time, there have been more than 20 descriptions of disease causing de novo L1 insertions in humans and mice 26. It is not apparent whether these endogenous L1 insertions integrated directly into the germline during spermatogenesis or oogenesis, or whether they integrated early enough in development that they were later incorporated into the germ cells, where they were passed on to future generations. However, it is clear from the many instances of insertional mutagenesis by endogenous L1s and by the hundreds of thousands of L1 elements that have accumulated in mammalian genomes that at least some de novo insertions make their way into the germline and are stably transmitted to offspring.

Proof-of-principle for germline mutagenesis in the mouse using L1 retrotransposons was demonstrated in animals that contained a transgene consisting of a highly active human L1 element tagged with a modified version of an enhanced green fluorescent protein (EGFP) retrotransposition cassette 30. Several lines of transgenic mice were created. In some lines the L1 element used its endogenous promoter for transcription, and in others the L1 was driven by the addition of a heterologous promoter. As expected, the endogenous L1 promoter was active in the germ cells. The addition of the heterologous promoter increased transcript levels in the germ cells and allowed for low levels of transcription in multiple somatic tissues. Mouse lines that contained either type of transgene demonstrated retrotransposition, detected by reverse transcription PCR and PCR of de novo insertions. Two germline insertions, one of which had inserted into a gene intron, were cloned from mice that had the heterologous promoter transgene. However, the transgenes in these initial mice did not contain gene traps and were not expected to disrupt genes when inserting into introns. The frequency of insertion was estimated at approximately one de novo insertion in every 70 sperm for mice with the heterologous promoter, and lower for mice containing the transgene that used only the endogenous promoter. More refined estimates using real-time PCR have demonstrated that some of these lines have germline insertion frequencies as high as one insertion in every 20 sperm.

Subsequent attempts at germline mutagenesis using L1 involved strategies to increase the frequency of retrotransposition, including the use of a more active human L1 element, a different heterologous promoter, and a more efficient retrotransposition cassette 13. The highest reported frequency of germline transmission of de novo L1 insertions was estimated at one insertion in every three sperm 12. These transgenic mice contained a codon optimized mouse L1 31. Although L1-based germline mutagenesis will probably play a role in the creation of a genome-wide collection of mouse knockouts, the existence of other already well established methods for creating germline mutations in mice (knockouts, knockins, and conditional mutations, among others) may limit its usefulness.

L1 elements are also capable of somatic mutagenesis. However, little is known about the regulation, frequency, and relevance of such retrotransposition. Two examples of disease-causing somatic L1 insertions have been described in humans, one associated with a case of colon cancer and the other associated with a breast cancer. An L1 insertion into exon 16 of the APC tumor suppressor, predicted to be an inactivating mutation, was discovered in a human colon cancer and was not present in adjacent normal tissue 3233. An L1 insertion between exons 2 and 3 at the myc locus was also identified in a ductal carcinoma of the breast, although the functional implication of this mutation was not determined 34. These findings indicate that L1 elements are occasionally active in somatic tissue and are capable of driving carcinogenesis. However, the relative contribution to oncogenic transformation is largely unknown. Growing evidence that L1 elements become hypomethylated within cancers suggest that L1 expression may be reactivated in somatic tissue and that retrotransposition may play a significant role as an insertional mutagen, thereby contributing to genomic instability 3536373839. Whether global L1 hypomethylation augments L1 expression and retrotransposition requires more rigorous investigation.

Human and synthetic mouse L1 sequences have proven to retrotranspose in somatic tissue in transgenic mice 121340. In one transgenic model, a highly active human L1 element (L1_RP) 41, which was tagged with an EGFP retrotransposition cassette 42 and was transcribed using its endogenous promoter, exhibited somatic retrotransposition within neuronal precursors 40. However, it is not known whether retrotransposition from endogenous L1 elements occurs in neuronal precursor cells in mice or humans. In other transgenic mice, a native human L1 element (L1_LRE3), tagged with a markerless retrotransposition cassette and driven by the mouse Hspa2 promoter, demonstrated somatic retrotransposition into sites throughout the genome, including transcriptional units, with 14.6% of insertions found within RefSeq genes 13. A synthetic, codon optimized mouse L1 (ORFeus) tagged with an EGFP retrotransposition cassette and driven by the CAG promoter (a hybrid between the human cytomegalovirus immediate early promoter and chicken β-actin promoter) was also capable of a high level of somatic retrotransposition in transgenic mice 12. Because both high level ubiquitous expression (with CAG) and codon optimization (ORFeus) would both probably enhance retrotransposition, the relative contributions of each of these factors in this transgenic mouse is unknown.

The most practical application of L1-based somatic mutagenesis in mice would be a screen for tumor suppressors and oncogenes. A tissue specific promoter could direct L1 mediated mutagenesis to a particular organ to drive the formation of different types of cancers. The incidence of loss of function mutations could be enhanced by tagging the L1 with a gene trap cassette containing strong splice acceptors. For some trapped tumor suppressors a loss of heterozygosity may be rate limiting for carcinogenesis, but could be expedited by using a Blm mutant background. Germline mutations in the RecQ-like helicase Blm causes a profound increase in sister chromatid exchange, which is mediated by homologous recombination, resulting in widespread cancer 43. Blm inactivation causes accelerated loss of heterozygosity of the wild-type Apc allele in Apc^min/+ mice, causing nearly four times as many polyps 43. Because Blm appears to repress cancer in multiple tissues, Blm mutants may be useful for somatic mutagenesis in many different cell types.

At present, the human genome contains 80 to 100 endogenous full-length active L1s. Although they have similar nucleic acid sequence, they are known to exhibit very different levels of activity. A recent survey of the retrotransposition capability of elements in the human genome demonstrated there are just a few elements with very high activity. These elements were called 'hot' L1s 55. Another study found that more active elements tend to insert larger amounts of DNA during retrotransposition 56. Therefore, hot L1 elements such as L1_RP and L1_LRE3 are a good choice for use in random mutagenesis because they are highly active and better equipped to deliver gene traps. The L1_RP element was used in the initial studies of mouse models of L1 retrotransposition 3057 and in a transgenic mouse model that demonstrated somatic retrotransposition in neuronal precursor cells 40. A more recent study that demonstrated L1 retrotransposition events at relatively high frequency and characterized more than 50 de novo L1 insertions used the L1_LRE3 element 13.

Most mouse L1 elements cloned to date have exhibited poor retrotransposition activity in a cultured cell assay when compared with the hot human elements 5859. However, the poor retrotransposition capability of mouse L1 elements in cultured cells may be due to poor translation 60. To overcome this problem, Han and Boeke 31 designed a synthetic mouse L1 that codon optimized the open reading frames of an endogenous mouse L1 element without changing the amino acid sequence. The codon optimized L1 (ORFeus) was used to create transgenic mice, in which it demonstrated high activity 12. In this model, all transgene positive progeny demonstrated somatic insertions and some insertions were passed through the germline.

The L1 element may be tagged with a retrotransposition cassette to facilitate detection of de novo L1 insertions. Retrotransposition cassettes are disrupted by an intron sequence and are cloned into the 3' untranslated region of an L1 element. The intron is removed by splicing during a retrotransposition event, allowing the differentiation between a de novo L1 insertion and the L1 element(s) present in the transgene. The cassette can contain a gene that is disrupted by the intron and activated upon retrotransposition, thereby allowing positive or negative selection. For example, the EGFP cassette allows detection of cells that contain a retrotransposition event by fluorescence 3042. A negative selection strategy could rescue cells that would otherwise die by the expression of a rescue gene upon retrotransposition. There are also cassettes that contain no selectable marker. In these 'markerless' cassettes, retrotransposition can be detected by PCR detection of loss of the intron 13. The main advantage of using a markerless cassette is that it facilitates detection of smaller insertions. The majority of L1 insertions are 5' truncated during the integration process. Therefore, a smaller cassette permits detection of more insertions. Recent L1 transgenic mouse studies demonstrated that more than 90% of de novo insertions were 5' truncated, and some of them were less than 500 nucleotides 1213.

Gene trapping technology can be used to maximize the rate at which retrotransposition insertions interrupt coding exons. Only about 1% of the genome is composed of exons. Therefore, only one in every 100 random insertions is expected to disrupt the coding sequence of a gene. A gene trap that disrupts the normal splicing sequence of genes during transcription can be included in the L1 retrotransposon, such that retrotransposons that land within gene introns in the correct orientation will disrupt gene expression. Because approximately 30% of the mouse genome consists of exons and introns 61, and new insertions are likely to be in the correct orientation 50% of the time, this strategy should increase the incidence of gene disruptions resulting from retrotransposition events from 1% to approximately 15%. To increase this rate of disruption even further, we designed a unique bi-directional gene trap that disrupts gene expression when inserted into an intron in either orientation (and that can therefore be expected to increase the rate of gene disruptions to approximately 30% of all retrotransposition events) 13. The gene trap that we chose utilizes the very strong splice acceptor from the human BCL2 gene, which is able to splice into exons located more than 100 kilobases away without alternative splicing 62. This particular gene trap demonstrated 100% efficiency (complete disruption of a gene with no alternative splicing) when tested in a cultured cell system 63 and has successfully been used to disrupt genes in the mouse 64. We have demonstrated an L1 element tagged with the bi-directional gene trap retrotransposition cassette can disrupt genes in cultured cells and have used the cassette in transgenic animals 13.

L1 elements contain an endogenous promoter 65, which is sufficient to direct expression of the L1 transgene during spermatogenesis and oogenesis in transgenic animals 30. Transcripts produced by the endogenous promoter are integrated either in the germline or early during development, so the endogenous promoter can be used for germline mutagenesis. The endogenous promoter has also been used in transgenic mice that demonstrated retrotransposition in neuronal precursor cells 40. Because L1 retrotransposition appears to be limited by the steady-state levels of the processed L1 mRNA 66, a reasonable method to increase retrotransposition frequency is by incorporating a strong heterologous promoter. We and others have demonstrated that a heterologous promoter can boost retrotransposition for both germline 30 and somatic mutagenesis 1213.

A feasible somatic mutagenesis screen in mice would hinge on robust and reliable expression in the desired tissues. This can be achieved by removing the endogenous L1 promoter and adding a strong tissue specific promoter. Unless the promoter possesses a locus control region, expression of the tagged L1 transgene may be affected significantly by the genomic integration site. Ideally, L1 transgenes designed for mutagenesis should have a promoter/enhancer that is resistant to position effect silencing, such as the albumin, a-fetoprotein, elastase, or β-globin promoters 676869707172. Alternatively, L1 transgenes can be targeted to specific loci in mouse ES cells to recapitulate the tissue-specific expression of that gene entirely. Recombinase mediated cassette exchange is a useful method that is increasingly being used to target particular loci, such as the β-actin locus 73. Another alternative would be the incorporation of insulators into the transgene, such as a chicken β-globin insulator 747576.