Modeling the Study of DNA Damage Responses in Mice

Summary

Damaged DNA has a profound impact on mammalian health and overall survival. In addition to being the source of mutations that initiate cancer, the accumulation of toxic amounts of DNA damage can cause severe developmental diseases and accelerate ageing. Therefore, understanding how cells respond to DNA damage has become one of the most intense areas of biomedical research in the recent years. However, whereas most mechanistic studies derive from in vitro or in cellulo work, the impact of a given mutation on a living organism is largely unpredictable. For instance, why BRCA1 mutations preferentially lead to breast cancer whereas mutations compromising mismatch repair drive colon cancer is still not understood. In this context, evaluating the specific physiological impact of mutations that compromise genome integrity has become crucial for a better dimensioning of our knowledge. We here describe the various technologies that can be used for modeling mutations in mice, and provide a review of the genes and pathways that have been modeled so far in the context of DNA damage responses.

1. Introduction

Our genetic material is constantly challenged by lesions that distort its integrity. DNA can suffer from a myriad of alterations, which include base modifications, intra- or internucleotide covalent links, and the breakage of one or both strands of the double helix. Accordingly, many specialized DNA repair processes have evolved to be able to maintain genome integrity (Lindahl and Barnes, 2000). Among the different types of DNA lesions, DNA double strand breaks (DSB) are considered the most deleterious, since they may lead to irreversible loss of genetic material or to chromosomal rearrangements, being a serious threat to human health. As such, DSB are highly cytotoxic. The repair of a DSB can occur by a direct ligation of the ends regardless of homology or by a process that involves the processing of the broken ends and the repair of the damage by using the information from a homologous sequence, typically – but not always- this being the sister chromatid. The two canonical pathways are Non-Homologous End Joining (NHEJ) and Homologous Recombination (HR); although multiple variants of each process have been identified in the recent years (i.e. Microhomology Mediated End Joining or Break Induced Replication, among others). NHEJ and HR are exclusive mechanisms, and the decision of whether a DSB is channeled into one or the other is now an area of intense research. NHEJ, a simpler yet error-prone form of DSB repair, operates throughout the cell cycle and involves the KU70/KU80 complex that senses the ends, the Ligase IV/XRCC4 complex for end ligation, and the signaling kinase DNA-PK, among other activities. On the other hand, HR-mediated repair is more precise, yet it is only active during S and G2 phases of the cell cycle, once a sister chromatid is available as a template. In contrast to the limited number of proteins that have been involved in NHEJ, a large number of factors participate in the different activities associated to HR such as end processing (ATM, BRCA1), strand invasion (RAD51 and its paralogues).

Given that DSB constitute one of the most cytotoxic and dangerous genomic lesions, the term DNA damage response (DDR) is commonly used for the signaling machinery that is activated by DSB. In the presence of broken chromosomes, the DDR stimulates the repair of the lesion while it also initiates cytostatic or cytotoxic responses that limit the expansion of the damaged cells (Harper and Elledge, 2007). At the tip of the DDR, DSB activate a phosphorylation-based cascade that starts with the activation of the Ataxia Telangiectasia Mutated (ATM) kinase. If the ends are processed and single stranded DNA (ssDNA) is generated, then the ATM and Rad3-related kinase (ATR) is also activated. Whereas DNA-PK is also activated by DSB, it seems to mostly phosphorylate factors at the lesion to promote its repair, but with a minimal impact on cellular growth. Interestingly, DNAPK can also promote checkpoint responses in the absence of ATM, which might be an interesting concept for the targeting of ATM deficient human tumors (Callen et al., 2009). ATM and ATR can limit the expansion of the damaged cells by the phosphorylation and activation of their target checkpoint kinases CHK2 and CHK1, respectively, which in turn limit CDK activity and therefore cell growth. In addition, the DDR can also phosphorylate and activate apoptotic factors, most prominently p53, to promote the elimination of the damaged cell.

DSB not only arise from exogenous DNA damaging sources, but also as products of physiological recombination reactions during the generation of immune or meiotic diversity. Specifically, the processes of V(D)J and Class-Switch recombination (CRS) generate the repertoire of T-cell receptors (TCR) and Immunoglobulins (Ig) in developing lymphocytes (Dudley et al., 2005); and DSB generated by the SPO11 endonuclease initiate meiotic recombination during spermatogenesis and oogenesis (Zickler and Kleckner, 1998). Consequently, immune deficiencies and infertility are often found in diseases associated to DSB repair deficiencies. These phenotypes, as well as many others, are often faithfully recapitulated in mouse models of the respective human diseases, which consolidated the mouse as a valuable tool for the study of the impact of DNA repair deficiencies in a physiological context.

2. Modeling genomic instability in mice

During the past 30 years, the generation of mouse models has allowed to study the physiological function of many of DNA repair proteins and their role in cancer and other diseases. Several human diseases caused by inherited mutations that compromise genomic integrity have been successfully modeled in mouse, providing the scientific community with unique platforms to investigate the role of a given mutation in a living mammal (see Table 1). Whereas mouse models recapitulate many of the symptoms found in human diseases associated with deficient DNA repair, this is not the case for all symptoms. A paradigmatic example is Ataxia Telangiectasia, driven by deleterious mutations in the ATM kinase (Savitsky et al., 1995). Whereas ATM deficient mice also suffer from cancer and immune deficiencies like the human patients, they do not present obvert features of Ataxia nor Telangiectasia (Barlow et al., 1996, Elson et al., 1996, Xu et al., 1996). In general terms, mice with compromised DNA repair activities often present one or several of the following alterations: abnormal development, altered pigmentation, sub-Mendelian birth ratios, infertility, immunedeficiencies, premature ageing and/or cancer predisposition.

Table 1. Examples of mouse models related to DNA damage responses.

Not all the strains available for each gene are referenced due to space constrains, but only some of them as examples of the available models.

Mutated gen	Human disease	Mouse models	References
Tp53bp1	Not known	KO	(Manis et al., 2004)
Artemis	RS-SCID	KO	(Rooney et al., 2002)
Atm	Ataxia Telangiectasia (AT)	KO	(Barlow et al., 1996, Elson et al., 1996, Xu et al., 1996)
Atr	Seckel Syndrome	Humanized allele	(Murga et al., 2009)
Bard1	Breast cancer susceptibility	KO, cKO	(McCarthy et al., 2003, Shakya et al., 2008)
Blm	Bloom’s Syndrome	KO, cKO	(Chester et al., 1998, Luo et al., 2000)
Brca1	Familial breast cancer	KO, cKO, KI, cKI	(Hakem et al., 1996, Evers and Jonkers, 2006, Liu et al., 2007, Drost et al., 2011)
Chk1	Not known	KO, cKO, BAC-Tg	(Takai et al., 2000, Zaugg et al., 2007, Lopez-Contreras et al., 2012)
Chk2	Li Fraumeni síndrome	KO	(Hirao et al., 2000)
Csa (Ercc8)	Cockayne syndrome	KO	(van der Horst et al., 2002)
Csb (Ercc6)	Cockayne syndrome	KO	(van der Horst et al., 1997)
CtIP	Jawad síndrome	KO	(Chen et al., 2005)
DNA-PKcs	Not known	KO	(Kirchgessner et al., 1995)
Ercc1	Cerebrooculofacioskeletal syndrome 4	KO	(McWhir et al., 1993)
Ercc3	Xenoderma Pigmentosum	KI	(Andressoo et al., 2009)
FancA	Fanconi anemia	KO	(Wong et al., 2003)
FancC	Fanconi anemia	KO	(Chen et al., 1996, Whitney et al., 1996)
FancD1 (Brca2)	Fanconi anemia; Familial breast cancer	KO, cKO	(Ludwig et al., 1997, Suzuki et al., 1997, Jonkers et al., 2001, Evers and Jonkers, 2006)
FancD2	Fanconi anemia	KO	(Wong et al., 2003)
FancG	Fanconi anemia	KO	(Pulliam-Leath et al., 2010)
H2ax	Not known	KO	(Celeste et al., 2002)
Ku70 (Xrcc6)	Not known	KO	(Li et al., 1998)
Ku80 (Xrcc5)	Not known	KO	(Difilippantonio et al., 2000)
Ligase IV	RS-SCID	KO, Tg	(Frank et al., 1998, Rucci et al., 2010)
Mdc1	Not Known	KO	(Lou et al., 2006)
Mlh1	Hereditary nonpolyposis colorectal cancer	KO, Tg	(Baker et al., 1996)
Mlh3	Colorectal cancer, Endometrial cancer	KO	(Lipkin et al., 2002, Avdievich et al., 2008)
Mre11	AT-like disorder (ATLD)	KI	(Theunissen et al., 2003)
Msh2	Hereditary nonpolyposis colorectal cancer	KO	(de Wind et al., 1995)
Msh6	Hereditary nonpolyposis colorectal cancer; Familial endometrial cancer	KO	(Edelmann et al., 1997)
Mus81	Not known	KO	(McPherson et al., 2004)
c-Myc	Burkitt lymphoma (when overexpressed in B cells)	Tg	(Harris et al., 1988)
Nbs1	Nijmegen breakage síndrome	KO, cKO	(Zhu et al., 2001, Demuth et al., 2004, Stracker et al., 2007)
p21	Not Known	KO, KI	(Wang et al., 1997, Barboza et al., 2006)
p53	Li-Fraumeni	KO, cKO, KI, Tg, BAC-Tg	(Donehower et al., 1992, Godley et al., 1996, Ludwig et al., 1997, Allemand et al., 1999, Garcia-Cao et al., 2002, MacPherson et al., 2004, Sluss et al., 2004, Johnson and Attardi, 2006, Dickins et al., 2007)
Parp1	Not known	KO, KI	(Pieper et al., 1999)
Pms2	Hereditary nonpolyposis colorectal cancer	KO	(Baker et al., 1995)
Ptip	Susceptibility to Alzheimer disease	KO, cKO	(Cho et al., 2003, Daniel et al., 2010)
Rad51c	Fanconi anemia	KO, KI	(Kuznetsov et al., 2007)
Rad54b	Not known	KO	(Couedel et al., 2004, Mills et al., 2004)
Recql2 (Wrn)	Werner síndrome	KO	(Lebel and Leder, 1998)
Rnf168	Riddle síndrome	KO	(Bohgaki et al., 2011)
Rnf8	Not known	KO	(Santos et al., 2010)
Rpa1	Not known	KI	(Wang et al., 2005)
Rtel	Dyskeratosis congénita	KO, cKO	(Ding et al., 2004, Wu et al., 2007)
Slx4	Fanconi anemia	KO	(Crossan et al., 2011)
Topb1	Not known	KO, cKO	(Yamane et al., 2002, Jeon et al., 2011)
Xlf	SCID with microcephaly, growth retardation, and sensitivity to ionizing radiation	KO, cKO	(Li et al., 2008, Zha et al., 2011)
Xpd (Ercc2)	Xeroderma pigmentosum	KI	(de Boer et al., 1998)
Xpf (Ercc4)	Xeroderma pigmentosum; Cockayne syndrome	KO	(Tian et al., 2004)
Xpg (Ercc5)	Xeroderma pigmentosum	KI	(Shiomi et al., 2004)
Xrcc1	Not known	KO	(Tebbs et al., 1999)
Xrcc4	Not known	KO, cKO	(Gao et al., 1998)
Zmpste24 (Face-1)	Hutchinson–Gilford progeria like syndrome	KO	(Pendas et al., 2002)

Focusing on cancer, the connection with genomic instability was first noted by Theodor Boveri in the early XX^th century (Boveri, 1914). In addition to the altered chromosome numbers detected by Boveri, tumor cells present important chromosomal rearrangements, which implies a defective DNA repair. Moreover, cancer cells acquire mutations at a faster pace than normal cells, (the so-called “mutator phenotype” (Jackson and Loeb, 1998)), which further supported that –at least some- cancerous mutations might relate to DNA Repair pathways. This concept was finally confirmed by the discovery of mutations affecting DNA Repair genes as drivers of hereditary cancers. Brca1 and Brca2 are the most frequently mutated genes in familiar breast cancer and their gene products are essential for the completion of HR (Apostolou and Fostira, 2013). In fact, given their relevance for human disease, dozens of BRCA1 and BRCA2 mouse models have been developed (Fig. 1) (Evers and Jonkers, 2006). Importantly, some of these models quite remarkably resemble the histopathological features of human BRCA1/2-deficient tumors. Yet, a remarkable difference between organisms exists as well; whereas full Brca1 or Brca2 heterozygosity are sufficient to drive tumorigenesis in humans, both alleles have to be mutated in mouse breast tissue for tumors to arise. In addition to HR, a fraction of hereditary colon cancers (Lynch Syndrome) is linked to deficiencies on Mismatch Repair (MMR), through mutations on genes such as Mlh1, Msh6 and Msh2. Whereas MMR-deficient mice are also tumor prone, tumors principally emerge on the lymphatic compartment and not on the intestinal tract (de Wind et al., 1998). Besides these examples that derived from previous knowledge in human tumors, it is fair to say that a common feature of mice deficient in DNA repair pathways is a higher propensity to develop malignancies (particularly lymphoid). Hence, cancer incidence is a common phenotype that is evaluated on mouse models of DNA Repair factors.

Figure 1. Examples of different gene targeting strategies that have been used to modify Brca1 in mice.

A) Structure of wild type Brca1. B) Example of a constitutive KO allele generated by introducing a neomycin-resistance cassette that replaces exon 5 and interrupts the reading frame. C) Conditional KO allele with loxP sites flanking exons 5 to 11. D) Conditional deletion of exon 11 results in a Δ11 version of BRCA1 protein with hypomorphic activity. This is a frequently used strain for BRCA1 studies given that animals lacking E11 almost complete embryogenesis, and embryonic fibroblasts can be obtained. E) A point mutation introduced in exon 5 (E5) replaces Cys 61 with Gly (C61G) in this constitutive KI allele, this being the mutation with the highest frequency detected in human Brca1 mutation carriers. F) An example of a conditional knockin allele of Brca1 generated by FLEX technology. Upon Cre recombination, the DNA fragment between loxP sites (black arrows) flips around and the E5 containing the C61G mutation is expressed. A second round of recombination between loxP155 sites (white arrows) excises the endogenous E5. The inversion-deletion process can also start with the inversion of the loxP155 sites, followed by the excision of the loxP-flanked sites. In both cases, the end product of the process is an allele that has lost the endogenous E5, and which has gained a C61G mutant version. The introduction of FRT sites flanking the neomycin-resistance cassette (grey arrows) enables the removal of this sequence upon crossing with a FLP expressing mouse strain. Note: Exon 4 of Brca1 gene was originally annotated improperly, which is why E4 is missing on the annotations.

In addition to hereditary cancers, mutations on DNA repair genes have also been associated to rare developmental disorders with a broader spectrum of clinical manifestations. As mentioned previously, Ataxia Telangiectasia (AT) is one of such syndromes and is caused by deleterious homozygous mutations in Atm. Again, whereas some of the symptoms of AT patients (i.e. radiosensitivity, immunedeficiency) were recapitulated on Atm deficient mice, the neurological symptoms which are arguably the most dramatic problem for these patients, are not present on the mutant mice. Another complication from modeling DNA repair-related diseases in mice arises from the fact that some of the mutations are essential on the rodents. This is the case for instance of the Bloom Syndrome (BS), which arises from the deficiency of the BLM helicase. Yet, BLM nullyzygosity is not compatible with mouse development (Chester et al., 1998). Perhaps one of the biggest challenges in the field has been the modeling of Fanconi Anemia (FA), a multigenic disease caused by mutations in more than a dozen genes involved on the repair of DNA interstrand crosslinks (ICL) (Moldovan and D'Andrea, 2009). FA is characterized by aplastic anemia, developmental problems and high predisposition to leukemia and other types of cancer. Whereas many of the FA genes have been mutated in mice, the FA phenotypes are very poorly recapitulated on the mouse mutants, if not inexistent (Parmar et al., 2009). On a positive note, however, it was only because of the use of mouse models that recent efforts revealed that endogenous aldehydes might be the natural genotoxic source that generates ICLs on FA patients, and thus responsible for their disease (Garaycoechea et al., 2012).

Intriguingly, rare diseases linked to accelerated ageing (progeria) also seem to have their basis on mutations that drive genomic instability. Two of the most widely known progeroid diseases, Hutchinson-Gilford Progeria (HGP) and the Werner Syndrome (WS) are linked to an accumulation of DNA damage. In the case of the WS, the mutation lies on WRN, a BLM-related helicase. Even if WRN deficient mice are not particularly progeroid, it has been proposed that this might be due to the longer telomere lengths of mice. Accordingly, work with mouse models revealed that WRN deficiency does in fact accelerate ageing in the context of short telomeres (Chang et al., 2004). As for the case of HGP, whereas the syndrome is caused by mutations on LmnA, related to the nuclear lamina and not to DNA repair (Eriksson et al., 2003), work with mouse models revealed that, for reasons yet to be understood, HGP cells accumulate substantial amounts of DNA damage (Liu et al., 2005). Other rare diseases in which patients with features of accelerated ageing are linked to mutations in DNA repair proteins are emerging rapidly with the help of the massive sequencing technologies. In our laboratory, we modeled one of those diseases known as the Seckel Syndrome (SS), which was shown to be driven by hypomorphic mutations on the ATR kinase (O'Driscoll et al., 2003). Based on the human mutation, we generated a humanized mouse model of the SS that faithfully recapitulated many of the diagnostic features of the human patients (Murga et al., 2009). Interestingly, ATR-Seckel mice accumulate high amounts of genomic instability during embryonic development, which then accelerates ageing on born animals. This finding led us to propose the concept of “Intrauterine Programming of Ageing”; namely that ageing rates can be determined by stresses to which we are exposed during embryonic development (Fernandez-Capetillo, 2010). Syndromes aside, the frequent finding of ageing features on DNA Repair mutant mice has been one of the essential facts to support the concept that ageing is driven by a progressive accumulation of DNA damage throughout our lives, particularly on stem cell compartments (Rossi et al., 2008).

We could not end this list of mouse models linked to ageing without mentioning the work done, significantly by the group of Jan Hoeijmakers, on mouse models of nucleotide excision repair (NER). Xeroderma Pigmentosum (XP) and Cockayne Syndrome (CS), among others, are hereditary diseases linked to NER deficiencies. In fact, it was XP and not breast or colon cancers the first link between DNA Repair mutations and cancer (Cleaver and Bootsma, 1975). XP is caused by mutations in NER genes (XPA, XPB, XPC and others), and patients present a high sensitivity to UV-light as well as a high cancer incidence. On the other hand, whereas CS is also caused by mutations in some NER genes, classically in ERCC8 and ERCC6, CS patients are more prone to accelerate ageing than cancer (Natale, 2011). A number of mouse models with mutations in NER genes have been developed presenting a broad spectrum of symptoms resembling XP and CS (Jaarsma et al., 2013). Noteworthy, different mutations on the very same NER gene can be pro-cancer or pro-ageing, further illustrating the intimate connection between these two outcomes. In addition to deepen our understanding of DNA Repair pathways, these mouse models also allowed Hoeijmakers to make the seminal discovery of a transcriptional signature of “aged” tissues (Niedernhofer et al., 2006).

In conclusion, the use of mouse models has allowed to solidify the concept that ageing and cancer can be initiated by DNA damage, and has allowed the generation of research models for human syndromes which can potentially be used for the discovery of new therapeutic strategies. Whereas we have also learned that many of the phenotypes linked to DNA Repair mutations are not shared between mice and humans, it is also true that many of the underlying mutations on the human patients are not simply deleterious, and some extra effort might be needed in order to generate models that faithfully recapitulate the human conditions. We here provide a broad overview, necessarily incomplete, of the most common technologies that have been used and can be used for mutagenizing the mouse genome for these purposes.

3. Technologies for the generation of mouse models

Techniques for the engineering of genetically modified (GM) mice were developed in the 1970ˋs-1980ˋs, fueled by mutual advances in developmental biology and molecular engineering (Fig. 2). In 1974, Rudolf Jaenisch generated the first transgenic mouse by retrovirus mediated infection of an early stage embryo, showing that this leads to integration of the foreign DNA into the host genome (Jaenisch and Mintz, 1974). Yet, it took until 1981 to show that injection of DNA into a pronucleus not only allows to engineer chimeric transgenic mice but also that this genetic material is passed on to subsequent generations (Brinster et al., 1981, Costantini and Lacy, 1981, Gordon and Ruddle, 1981). A key achievement for the field was the observation that murine embryonic stem (ES) cells can be isolated from blastocysts, cultured and re-introduced into a surrogate mother to generate chimeras (Evans and Kaufman, 1981, Martin, 1981, Bradley et al., 1984). Introduction of a selection marker gene (e.g. Neomycin) into ES-cells allowed to screen for successfully transfected clones prior to reintroducing them into the surrogate mother (Gossler et al., 1986) greatly enhancing efficiency of the method.

Figure 2. Timeline of the breakthroughs related to the use of genetically modified mouse models in research.

General breakthroughs are depicted in red, and some of the relevant models mentioned in this work in green.

With those technologies, GM mice were exclusively generated from constructs randomly integrating into the genome, giving rise to transgenic animals. The next breakthrough for gene targeting came, interestingly, with a better understanding of how a foreign DNA element can be introduced into a specific locus by exploiting recombination pathways (HR). The introduction of sequences into the foreign DNA that were identical to an endogenous locus allowed to introduce site-directed modifications and was first reported for the beta-globin locus by Smithies and colleagues (Smithies et al., 1985). Shortly thereafter, two groups reported the first successful generation of chimeric mice from gene-targeted ES-cells (Doetschman et al., 1987, Thomas and Capecchi, 1987). For these and previous findings, the 2007 Nobel prize in medicine was awarded to Capecchi, Sir Martin J. Evans and Oliver Smithies “for their discoveries of principles for introducing specific gene modifications in mice by the use of embryonic stem cells”. These were the seminal discoveries that enabled the explosion of mouse model research. Since then, the field has gradually improved by the creative design of targeting constructs and the introduction of in vivo recombination technologies. We here summarize some of the most frequent types of GM mice, exemplified through mouse models related to genome instability.

3.1. Transgenesis

Transgenic mice contain additional genetic material, ranging from additional copies of an endogenous gene to mutant versions of a gene of interest.

3.1.1. cDNA-based transgenesis

Most frequently, a transgenic mouse line consists of a cDNA of the gene of interest under the control of a short regulatory sequence. Genomic insertion of a transgene, generally delivered into a fertilized egg by pronuclear injection, occurs at random. As a consequence, identification of the insertion site and number of integrations is an important issue. In addition, comparison of the phenotype of various founder mouse lines should be conducted in order to exclude insertion site-specific phenotypes (e.g. embryonic lethality due to interruption of an essential gene). Working in heterozygosity is in these cases preferable to avoid phenotypes caused by transgene-integration.

The use of transgenic mice has dramatically increased our knowledge of (increased) gene function in the context of whole organisms. This was particularly relevant in cancer, due to the increased expression of oncogenes in human tumors. Hence, in the 1980’s, introducing oncogenes into the mouse genome generated the first so-called “oncomice”. Those early experiments led to the verification that ectopic expression of an oncogene in a healthy organism can lead to a heritable predisposition for cancer development (reviewed in (Hanahan et al., 2007). A prominent example of a cDNA-based transgenic mouse model is the Eμ-Myc mouse, a model for lymphomagenesis developed in 1985 by Brinster and colleagues (Adams et al., 1985). Ectopic overexpression of Myc is a frequent oncogenic mutation leading to aberrant regulation of proliferation and resulting in genomic instability (reviewed in (Campaner and Amati). The Eμ-Myc mouse carries a c-myc fusion gene under the control of the Ig heavy chain enhancer, which was based on the translocation found in Burkitt’s B-cell lymphoma (Adams, 1985). Mice carrying the Eμ-Myc transgene develop aggressive lymphoma or leukaemia and die within the first months of life (Adams, 1985, Langdon et al., 1986, Harris et al., 1988), and have been widely used as platforms for the discovery of genetic or chemical conditions that limit Myc-induced carcinogenesis.

Besides constitutive whole-body expression, transgenes can be placed under spatial (with the use of tissue specific promoters) or temporal (promoters that can be activated by orally available inert chemicals) control. Yet, one frequent limitation of cDNA based transgenesis is the lack of proper regulation of the studied gene.

3.1.2. BAC-transgenesis

cDNA-based transgenes lack the natural genomic regulation (introns, promoters, enhancers…), which may limit the interpretations of the results obtained. Moreover, unregulated expression might even limit the generation of the mouse lines. For instance, efforts to generate p53 transgenic mice repeatedly failed due to severe defects in embryogenesis, differentiation and apoptosis (Nakamura et al., 1995, Godley et al., 1996, Allemand et al., 1999). In contrast to classical transgenic vectors, bacterial artificial chromosomes (BAC) can harbor up to 300kb of DNA and can accommodate full genes along with their regulatory elements (Shizuya et al., 1992). The use of a BAC vector containing p53 in its natural environment omits the previously observed side effects of p53 overexpression, and allowed the generation of mice with enhanced p53 function (superP53) (Garcia-Cao et al., 2002). In the context of the DDR, our group showed that mice with just one extra allele of Chk1 showed a more efficient response against replication stress (RS) (Lopez-Contreras et al., 2012). The use of this strain allowed us to demonstrate that, unexpectedly, a modest increase in CHK1 levels (known as a tumor suppressor) could facilitate oncogenic transformation by limiting the genomic instability generated by the oncogenes. It should be noted that the use of these “super” mice, carrying a full extra copy of a gene might actually be physiologically relevant, given that duplications (some of which contain genes) are one of the most frequent source of diversity between human genomes. In fact, increased gene dosage of certain genes involved on the maintenance of genome stability, such as Replication Protein A (RPA), can counterintuitively lead to genomic instability in humans (Outwin et al., 2011).

3.1.3. shRNA-transgenesis

Inferring the function of a gene can also be done by turning it off, and looking at the consequences. This can also be done by transgenesis, taking advantage of the phenomenon of RNA interference (Hannon, 2002). Upon expression a transgenic shRNA triggers RNA interference of a specific messenger RNA in vivo, thus inducing a gene knockdown, which mimics genetic deletion to variable extents. shRNA transgenes can be made conditional by using the Tet system. The integrity of the endogenous target gene combined with the use of the Tet system allows for reversibility of the gene silencing. A well-known example of RNA interference-based mouse modeling is a strain carrying a transgene coding for a p53-targeting shRNA regulated by a tet-responsive promoter (rtTA) (Dickins et al., 2007). This strain develops lymphomas upon p53 downregulation, and lymphoma regression upon inactivation of the system. To follow-up with previous examples, this technology has also been used for proteins such as RPA, allowing the study of the consequences of sudden and broad genomic instability in adult mice (McJunkin et al., 2011).

3.2. Gene Targeting

3.2.1. Constitutive knockout

Knockout (KO) models are characterized by the inactivation of a specific gene, resulting in truncated, non-functional or absent protein expression. This method allows the study of gene function in vivo and is of notable interest since many human diseases, including some types of cancer, are caused by gene loss of function (for review see (Mak, 1998)). The vast majority of KO mice are generated from embryonic stem (ES) cell clones in which a particular gene was mutagenized through HR by a replacement vector lacking one or more essential coding exons. The replacement vector usually contains a drug resistances markers (e.g. Neomycin) to allow both negative and positive selection of the targeted cells (Hall et al., 2009). This being the most widely spread technology used, to date hundreds of different knockout mice have been generated. In fact, almost every discovery of a new DNA repair protein is followed up within the subsequent years with the generation of a knockout strain. There are thus too many knockout mice related to genomic instability to summarize. As examples of some of the most widely used we could single out ATM, Chk2, 53BP1 or H2AX deficient mice, perhaps ATM^-/- being the most widely used as a DDR model by many investigators around the world.

3.2.2. Conditional knockout

Although traditional knockout mouse technology is a convenient way of studying protein function, the method has a number of important restrictions. Roughly 15 percent of constitutive KO models are unable to complete embryogenesis, this percentage being higher for proteins involved on essential repair pathways such as HR. This is the case for the ATM- and Rad3-related kinase ATR, which is the essential kinase that coordinates the response against replication stress (RS) (Brown and Baltimore, 2000, de Klein et al., 2000). Hence, in order to study ATR function at the organismal level, a conditional KO strain was generated (Brown and Baltimore, 2003). The use of these mice allowed to reveal that ATR deletion in adult mice accelerates ageing (Ruzankina et al., 2007). Of note, the ageing phenotype was not directly related to ATR in this case, but rather to the loss of adult stem cells that lost ATR, followed by a compensatory proliferation of the remaining ATR wild type cells. This phenotype of accelerated ageing is frequently observed upon the deletion of an essential gene in adult mice (which is usually not absolute, so that the wild type cells need to compensate for the loss of cells).

The conditional knockout technology relies on DNA recombinases such as Cre or FLP. These enzymes catalyze a recombination reaction between two target sequences, loxP in the case of Cre and FRT sites in the case of FLP (Sternberg et al., 1981, Sauer and Henderson, 1988). Depending on the relative orientation of the target sites, the result is the excision of the DNA contained in between the target sites (when the sequences are on the same orientation) or its inversion (for inverted sequences). For the generation of conditional KO mice, loxP sites are usually placed at both sides of an essential exon of the target gene, preferably by targeting an exon (or exons) that would lead to an altered reading frame upon deletion. The “floxed” mice are then bred with Cre expressing transgenic animals that can direct the expression to the recombinase, and thus the deletion, to a particular tissue. The use of conditional knockout models demands a similar development of Cre recombinases. We now have hundredths of tissue specific Cre lines (an updated database is mantained by the Nagy laboratory at http://nagy.mshri.on.ca/cre_new/) which can be used in combination of floxed alleles. The number of Flp carrying lines is also rapidly evolving. Importantly, a tamoxifen-inducible version of Cre (Cre-ERT2 in its most evolved version), allows for not only a tissue-specific expression of the Cre (given by the promoter), but also for a temporal control of its activation (in response to an inert derivative of tamoxifen, 4-hydroxy-tamoxifen).

The use of conditional knockout alleles is of particular relevance for the study of essential genes or for cases in which the phenotype of a KO strain is so aggressive that precludes the study of additional phenotypes. For instance, while p53 KO mice provided an important model for studying p53ˋs role in tumor development (Donehower et al., 1992), these mice almost invariably succumb to aggressive lymphomas. The incidence of lymphoma (and in some cases sarcoma) does not allow for other tumor types to develop before animals succumb to disease. However, the most frequently observed human cancer type carrying p53-mutations are carcinomas (for review see (Johnson and Attardi, 2006). Hence, in order to induce different types of tumors and study the consequences of p53 loss in other tissues, a number of tissue-specific p53 KO models have been employed. In regards to genomic instability, the use of conditional knockouts has been particularly useful for the study of BRCA1-associated breast tumors (reviewed in (Evers and Jonkers, 2006)). In contrast to the high incidence of tumors on human BRCA1 mutation carriers, mammary-specific deletion of BRCA1, together with p53, is needed in mice for tumors to develop.

3.2.3. Knockin

Besides deleting a protein, specific mutations can also be targeted or knocked in at a particular genomic location (knockin; KI). KI alleles can introduce point mutations, incorporate exogenous DNA fragments, or replace large endogenous DNA fragments with exogenous ones. These strategies provide a variety of possibilities for studying specific functions of proteins in vivo, modeling disease and generating tools for mouse engineering. Knockin strategies have been widely used to interrogate the role of specific protein residues, such as phosphorylation sites or key residues at catalytic sites. Relevant examples are the p53 phosphorylation defective mutant mice generated at the laboratories of Stephen N. Jones and Tyler Jacks (MacPherson et al., 2004, Sluss et al., 2004). Absence of p53 protein is known to be a predisposing condition to tumor development as shown in various KO models (Donehower et al., 1992). Serine residues 18 and 23 were proposed to be critical phosphorylation sites for p53-mediated tumor suppression (Oliner et al., 1993). To test this in a physiological context, several studies used KI mice carrying point mutations at the endogenous p53 locus that replaced serines 18 and 23 with alanines (MacPherson et al., 2004, Sluss et al., 2004, Chao et al., 2006, Armata et al., 2007). Unexpectedly, these studies showed that such phosphorylations only had a minor contribution to p53 tumor suppressive function. In contrast, in some other cases the use of knockin alleles of phosphorylation sites previously shown to be relevant in vitro revealed a lack of a phenotype in vivo. This was for instance the case of the ATM kinase, where previous studies had shown that autophosphorylation at S1981 was essential for its activation (Bakkenist and Kastan, 2003). However, mice with the corresponding residue mutated to Alanine do not show any obvious phenotype (Pellegrini et al., 2006).

3.2.3. a Humanized alleles

In some cases, the study of a mouse model might have the limitations that the orthologue sequence is not identical to the human sequence. Yet, strategies have also evolved that enable the study of the human sequence within the mouse genome. One of those strategies implies the modification of a mouse locus with the corresponding human sequences, so that the allele becomes “humanized”. This strategy was used to develop a mouse model for ATR-Seckel syndrome (Murga et al., 2009). The Seckel syndrome is a genetic disease characterized by a severe growth retardation, microcephaly. A subset of Seckel patients carry a synonymous point mutation in the ATR gene that leads to reduced protein levels due to abnormal splicing (O’Driscoll et al., 2003). On ATR-Seckel mice, the human region containing the mutated exon, together with the surrounding 2 exons and introns was swapped with the corresponding mouse region. This humanized allele rendered mice expressing very low levels of ATR, which faithfully recapitulated the symptoms observed in the patients (Murga et al., 2009). In addition, ATR-Seckel mice were used to reveal that low ATR levels are particularly toxic for cells expressing oncogenes (Murga et al., 2011, Toledo et al., 2011). Of note, other strategies have also been used within the genomic instability for the generation of humanized mice. For instance, transgene-introduction of mutated BACs containing the entire human NBS1 locus, in the context of the deletion of the murine NBS1 gene, generated humanized alleles of disease-relevant NBS1 mutations (Difilippantonio et al., 2005).

3.2.4. Conditional knockin

Some targeted mutations are early embryonic lethal in mice when expressed constitutively. One possibility to overcome this limitation is having one KI allele harboring the mutation and the other a cKO. Upon Cre-mediated recombination the conditional allele is excised and only the KI mutant allele is expressed. One example of this strategy was done with BRCA1 C61G mutant mice (Drost et al., 2011). The C61G mutation at the RING domain of BRCA is the most frequent missense mutation in familial breast cancer. Yet, homozygosity for this mutation is embryonic lethal in mice. Using this strategy the authors were able to express the C61G mutant protein specifically in mammary gland, which led to mammary tumors (Drost et al., 2011). However, it must be noted that using this approach only half of the mutant protein is expressed. In the recent years, genome-engineering technologies have led to the development of technologies that enable a controlled expression of the mutant allele. One of such approaches is the FLEX switch (Schnutgen et al., 2003) (Fig. 1e). This strategy relies on the use of heterotypic target sites for the Cre recombinase placed on opposite orientations, which promote the inversion of the flanked sequence. FLEX conditional mutant alleles are built around an exon that harbours the mutation of interest. The allele is composed of a forward-oriented endogenous version of the exon, followed by its mutated version in the reverse orientation. Both are flanked by two pairs of different heterotypic recombinase target sites with a specific orientation. While the endogenous version is expressed constitutively, an allele subjected to Cre recombination will “flip around” and stabilize the reverse orientation. In this manner the mutant version is switched on, while the endogenous one is switched off. Whereas no examples of the FLEX technology have been published on the field of genomic instability yet, we have developed a conditional mouse model of the BRCA1 C61G mutation described above which works as expected (Lopez-Contreras et al., unpublished).

3.2.5. Gene trapping strategies

While directed gene targeting provides an invaluable tool to study specific gene function, it remains a relatively labor intensive and expensive method. One alternative to directed mutagenesis is to randomly mutagenize murine ES cells with gene trapping constructs, which can then be used for the generation of mutant mice. A typical gene trap cassette contains a strong 5´splice acceptor sequence followed by a promotorless antibiotic resistance gene and/or a reporter gene (e.g. beta-gal). When inserted within an intron of an active gene, upstream exons are spliced to the resistance-reporter cassette, leading to a truncated protein. In this manner, it is preferable to obtain alleles that have been trapped at the beginning of the gene, to maximize the probability of getting severe protein dysfunction. Since the endogenous promoter of the trapped gene drives the expression of the reporter, these alleles can also be used to study the dynamics of gene expression in vivo. Mutagenesis of the ES cells is achieved by the integration of the genetrap construct, which has been mostly done by retroviral transduction, but also by other methods (i.e. use of transposons). Once mutant cells are selected with the use of the selectable markers, the identification of the trapped gene can be done either by identifying the exon previous to the genetrap (5’RACE), or by directly mapping the precise position of the insertion on the genome through inverse PCR and related protocols. Genetrap-mediated gene silencing in ES cells was already described in the early 90’s (von Melchner et al., 1992), and subsequently used for screenings in vitro (Forrester et al., 1996). Genetrap based screenings have the limitation that only one allele is trapped, so that the phenotypes have to be evident in heterozygosis. This limitation was recently overcome by the development of haploid murine ES cells (see below). Yet, the easiness of this process launched the creation of large libraries of mutant ES cells, which investigators could then obtain to generate their mouse strains. For instance, one of the earliest collections was that of the German Genetrap Consortium (http://www.genetrap.de), which already in 2003 reported the generation of more than five thousand characterized ES clones. Other consortia are available such as the International Gene Trap Consortium (http://www.genetrap.org), BayGenomics (https://www.mmrrc.org/catalog/overview_BG.php), the Center for Modelling Human Disease (http://www.cmhd.ca/genetrap/index.html) or the collections at the Japanese RIKEN Institute (http://www2.brc.riken.jp/lab/mouse_es/index.html.en). These collections have been of great use to investigators on many fields, including the DDR. For instance, 53BP1 (Morales et al., 2003), and its upstream regulators RNF8 (Minter-Dykhouse et al., 2008) and RNF168 (Bohgaki et al., 2011) were all modeled by genetrapped strains. Besides genetrap collections, it should be mentioned that the International Knockout Mouse Consortium has more recently took the challenge of developing constitutive and conditional gene targeting constructs for all mouse genes (Skarnes et al., 2011). The Consortium distributes modified ES cells, targeting vectors, or even living mice in the case that they have already been generated. In fact, the ultimate aim of the project is to phenotype mouse mutants for every murine gene, and the effort is now run under a collaborative project named the International Mouse Phenotyping Consortium (http://www.mousephenotype.org/martsearch_ikmc_project/). Many investigators, including us, have benefited from this effort that greatly speeds up the creation of KO or cKO strains.

4. Conclusions and future

The accumulation of DNA damage has its roots on pathologies such as cancer, immunedeficiencies or accelerated ageing, which can only be investigated in the context of a living animal. We have here summarized the technologies that can be used to mutagenize the mouse genome in order to generate mutant mice, with a focus on the efforts that have been placed on mice related to DNA damage repair. Given the large number of mice and technologies available, this work is not fully comprehensive, and we apologize for the many mice and technologies that have not been mentioned for space reasons. As for the future, among the many technologies that are emerging, two deserve particular attention. On one hand, the rapid development of sequence specific nucleases of the CRISPR-Cas9 system allows for a very effective site-specific manipulation of the mammalian genome, even in cell lines that are not particularly recombinogenic such as most of the human cell lines used in the laboratory (Ran et al., 2013). Not surprisingly, this method has been rapidly adapted to mutagenize murine ES cells. Its high targeting efficiency and ease of use allows for the simultaneous targeting of several genes, which will greatly reduce the time to generate compound mutants (Wang et al., 2013, Yang et al., 2013). In addition to CRISPR-Cas9, the generation of mammalian haploid cells, including murine ES cells, constitutes another recent breakthrough that will further facilitate the fast generation of mice homozygous for one or more mutations (Leeb and Wutz, 2011). With the incoming technologies, there has never been a time in which the generation of mutant mice has been easier than it is now. All fields of biomedicine should benefit from these developments that will help us dimension the relevance of our in vitro findings in the context of a living mammal.