In 2002 the International Mouse Genome Sequencing Consortium released the first draft of the genome from C57BL/6J, an inbred strain of the laboratory mouse 3 , and a finished genome was released in 2009 5 . As one of the most globally used lines, C57BL/6J was a wise choice for the reference mouse strain, but it is by no means the only strain used in research. Therefore, subsequent efforts were initiated to generate genomic sequence of other inbred strains. Firstly, four different strains of the laboratory mouse were included by Celera in a whole-genomic shotgun sequencing project: A/J, DBA/2J, 129X1/SvJ and 129S1/SvImJ 6 . This resulted in 27.4 million sequencing reads, giving a total of 5.3x coverage of the mouse genome. Secondly, more than 150,000 short insert clones were sequenced from the 129S5SvEv^Brd strain covering 4.7% of the reference genome 7 . Thirdly, Perlegen Sciences used hybridization to re-sequence 15 inbred mouse strains 8 ; this set included 11 classical strains and four strains derived from the wild. Unlike the other resources, Perlegen's approach did not generate sequence reads, and their hybridization sequencing technology queried only 1.49 Gigabases of the reference genome (equivalent to about 58% of the C57BL/6J sequence that is non-repetitive). Furthermore, to generate high accuracy calls, high stringency cutoffs were used, resulting in a false negative rate estimated to be as high as 50% 2 . Therefore, available sequence data lacked the coverage and breadth of strains to make it a widely used resource.

The first non-reference mouse chromosomes to be sequenced were A/J and CAST/EiJ chromosome 17, revealing significant variation at the nucleotide level and also considerable structural differences 9 . Building on that work, we commenced the Mouse Genomes Project, which has sequenced the genomes of 17 key mouse strains using next-generation sequencing on the Illumina platform (Box 1). At the last data freeze in December 2009 an average of 25x sequence coverage of each strain had been generated, and a deep catalog of variants 10 . These data provide a comprehensive insight into the genomes of the 17 strains, allowing immediate access to background genetic information for most mouse models of disease in addition to facilitating the analysis of the molecular basis of complex traits with unparalleled resolution.

Technologies for modifying the mouse genome can be split into two broad classes: those for gene-driven analyses and those for random mutagenesis.

The collection and propagation of mice harboring spontaneous mutations with striking phenotypes, such as the obese mouse, served mouse geneticists well for most of the 20th century. When it became clear that the rate of random germline mutation can be significantly increased by exposure to radiation or to chemical mutagens such as N-ethyl-N-nitrosourea (ENU) 11 , large-scale mutagenesis programs followed, resulting in an explosion in the number of mutant lines. Phenotypic screens of these lines led to the identification of many hundreds of new mutations and candidate disease genes 12 13 . One notable example of a successful forward genetic screen, reviewed in 14 , identified 89 ENU-induced mutants that influence the immune system, of which at least 69 have now been characterized at the molecular level. However, mapping random mutations and identifying the affected gene can be an arduous process, often taking years; therefore, causal mutations for only a fraction of mutant lines have been identified thus far. Screening DNA from archived mutagenized lines for mutations in a specific gene of interest is a parallel 'gene driven' strategy that has proven successful 15 16 . With the advent of new sequencing technologies it is now cost effective to sequence mutagenized mouse exomes in their entirety, enabling the rapid identification of candidate disease genes from existing resources and meaning that mutagenesis-driven approaches may return as a powerful tool for studying disease genes. Other methods of random mutagenesis include retroviruses, transposons (reviewed in 17 ), and 'gene traps' 18 . These DNA-based mutagens can be easily mapped using approaches such as splinkerette PCR 19 and are discussed in more detail below.

The genome of the mouse can also be manipulated by pronuclear injection of DNA into oocytes or by modification of embryonic stem (ES) cells, which can then be injected into blastocysts to make chimeras, allowing modified alleles to be transmitted through the germline. Direct pronuclear injection results in random integration of the injected DNA 20 ; consequently, transgene copy numbers and integration sites differ between lines, potentially resulting in very different phenotypes. Large genomic fragments such as bacterial artificial chromosomes (BACs) may also be injected (reviewed in 21 ); these have proven particularly useful in complementation studies or 'rescue' experiments for identifying genes contributing to a genetically mapped disease trait of interest 22 . By contrast, DNA introduced into ES cells in culture can undergo site-specific, homology-directed recombination 23 , thus enabling the generation of targeted gain- and loss-of-function alleles as well as the engineering of large-scale rearrangements of entire mouse chromosomes (Figure 1) 24 25 . Other recently developed techniques include transgenic small hairpin RNAs (shRNAs), which are often delivered by lentiviral transgenesis 26 27 , single-stranded oligonucleotides (ssODNs; reviewed in 28 ), and zinc-finger nucleases (ZFNs) 29 , which can be used to generate subtle sequence-specific genomic modifications. Here we will address in more detail a few of these technologies, focusing on recent advances uniquely available to mouse geneticists.

The success of the genome sequencing consortia over the past two decades established a model for further large-scale, collaborative projects aimed at functionally characterizing genomes (Table 2). Examples include the International Knockout Mouse Consortium (IKMC), and its constituent regional projects, which collectively aim to generate mutant alleles for every protein-coding gene in the mouse genome and to make the resources available to the scientific community 4 . Researchers can now search the IKMC website and acquire, at minimal cost, mice or ES cells that lack a gene of interest 79 , thereby accelerating the path from a gene of interest to mutant mouse line. By May 2011, IKMC had over 16,000 ES cell lines with mutations in protein coding genes. Many of these alleles are 'knockout first' alleles, which are designed to introduce a LacZ expression marker into a target gene, and the allele can be tailored by using Cre and Flp to generate null and conditional alleles, respectively 80 .

In parallel, a number of past and ongoing standardized phenotyping projects have documented traits in inbred strains and mutant lines for phenotypes relevant to human disease, including the Mouse Phenome Project, the European Mouse Disease Clinic (EUMODIC) and the Mouse Genetics Project (MGP; based at the Wellcome Trust Sanger Institute); see also Table 2 81 82 83 84 . The results from many screens are made available online, enabling researchers to identify potentially interesting phenotypes for detailed analysis (Figure 2). For example, primary MGP analysis of mice lacking the gene Slx4 identified a number of developmental and DNA instability phenotypes. Detailed secondary analysis revealed the mouse to phenocopy a new sub-type of the human genetic illness, Fanconi anemia 85 86 87 .

In an effort to identify quantitative trait loci (QTLs), large stocks of genetically heterogeneous (HS) mice have been generated 88 89 . Individual mice have been phenotyped and genotyped to facilitate high-precision QTL mapping. The Collaborative Cross (CC) is a resource that is using a similar strategy by interbreeding eight strains of mice to generate around 300 new inbred lines 90 , which, unlike HS mice, are being cryopreserved for posterity. It is estimated that the CC will capture approximately 90% of the genetic variability in laboratory mice and will allow the mapping of genetic networks that underlie complex diseases. Moreover, the progenitor strains of the CC were selected for sequencing in the Mouse Genomes Project (Box 1), which should allow the QTLs identified by phenotyping CC mice to be rapidly resolved into a list of candidate variants. When complete, the CC will mark a new era in the discovery of the molecular basis of complex traits in the mouse. Meanwhile, large-scale phenotyping of the strains developed so far is well under way. Finally, EuTRACC is a project to generate ES cells that carry a targeted tandem affinity purification tag (TAP-tag). Initially this will be several hundred transcription factor genes, but this is an effort that is likely to extend genome-wide. This resource will facilitate mass spectrometry of native protein complexes to better understand the mouse 'interactome' 91 .

Mouse genetics has a bright future. Genome-wide association studies have identified hundreds of alleles statistically associated with human disease, which now demand detailed functional analysis. Early examples suggest the mouse will be the ideal model in moving from genetic association studies to understanding molecular mechanisms leading to complex disease 92 . The ablation of a large proportion of the coding mouse genome within the next 5 years, at least in ES cells, should rapidly accelerate these studies.

The modular design of modern gene targeting cassettes, together with SSRs and RMCE, makes for an incredibly flexible system of genetic engineering in mice. This is establishing the mouse as a leading model in scientific disciplines previously dominated by work in simpler organisms. For example, gene targeting combined with channelrhodopsin, which allows the control of neural activation using light 93 , allows the visualization, and fine manipulation, of precise neural circuits in the mammalian brain that was until recently only possible in Drosophila and C. elegans 94 95 .

However, there are also challenges ahead. A significant number of regions across the mouse genome, typically those containing clusters of highly homologous, tightly arrayed genes, are not amenable to efficient gene targeting. Moreover, the same loci are often difficult to sequence, with some lacking complete coverage even in the high quality reference genome 96 . Thus, as much as 5 to 10% of the functional mouse genome may fall through the cracks of the present large-scale projects unless new technologies, or clever combinations of current technologies, are developed and used to investigate these genes. Nevertheless, the mouse is likely to remain the non-human vertebrate with the most sequenced, and best studied, genome for the foreseeable future. Together, the advances described here will underpin an understanding of mouse genetics within the current decade unthinkable to CC Little when he first began generating inbred lines over a century ago 1 .

The authors declare that they have no competing interests.

The 17 strains being sequenced as part of the Mouse Genomes Project were carefully selected to support other major mouse genetics resources. Three 129 strains were chosen because they serve as the background for thousands of existing gene knock-outs. The C57BL/6N strain is the origin of the highly germline-competent JM8 ES cells that are being used in large-scale gene targeting programs 31 . Nine common lab strains were chosen because of their historical utility, and also because they include the progenitors of the heterogeneous stock and Collaborative Cross mice that are used in dissecting complex traits 88 89 . Finally, four wild-derived strains have been sequenced because they represent some of the founder sub-species of many inbred laboratory lines, and are also important models of cancer and infection resistance 2 .

The first transgenic mice were generated by infecting embryos with viruses 97 , and today viral vectors remain an integral part of the mouse genetics toolkit. Lentiviruses integrate their genome into the host's DNA, making them an effective transgene delivery vector. The lentiviral genome, derived from immunodeficiency viruses, has been deconstructed and distributed across multiple plasmids to minimize the potential formation of replication-competent viruses 98 . A transgene of interest may be included in a plasmid containing a viral packaging signal. This is co-transfected into a cell line (typically human embryonic kidney HEK293T cells) with other plasmids expressing proteins required for viral production, such as envelope proteins. Viruses produced in this way can be introduced into oocytes for transgenesis (reviewed in 99 ). In its simplest form, this method necessitates only a few weeks between target selection and phenotypic analysis, offering a distinct advantage over other approaches. To enable pooled loss-of-function screens to identify complex genetic interactions, lentiviral short hairpin RNA (shRNA) libraries targeting most mouse genes have been generated 27 . Some groups have recently used ultrasound-guided microinjections of lentiviruses to deliver genes to organs and tissues of early mammalian embryos in utero 100 .

Slow transforming retroviruses have been widely used to generate mouse models of cancer 101 . They can re-infect the same cell, randomly inserting their genome into the host DNA multiple times, resulting in an accumulation of mutations. This process of progressive mutagenesis recapitulates the multi-step progression of human tumorigenesis (reviewed in 102 ). The development of next-generation sequencing technologies has dramatically enhanced the process of identifying retroviral insertion sites, and databases, such as the Retroviral Tagged Cancer Gene Database, have been developed to map insertion sites to the reference genome 103 .