COMMENTARY The Knockout Mouse Project Mouse knockout technology provides a powerful means of elucidating gene function in vivo, and a publicly available genome-wide collection of mouse knockouts would be significantly enabling for biomedical discovery. To date, published knockouts exist for only about 10% of mouse genes. Furthermore, many of these are limited in utility because they have not been made or phenotyped in standardized ways, and many are not freely available to researchers. It is time to harness new technologies and efficiencies of production to mount a high-throughput international effort to produce and phenotype knockouts for all mouse genes, and place these resources into the public domain. Now that the human and mouse genome expression pattern of the knocked-out gene in a dedicated project to produce knockout alleles http://www.nature.com/naturegenetics sequences are known1–3, attention has turned mouse tissues. Such marking of cells by a for all mouse genes and place them into the to elucidating gene function and identifying reporter gene facilitates the identification of public domain. The meeting took place from gene products that might have therapeutic new cell types according to their gene expres- 30 September to 1 October 2003 at the value. The laboratory mouse (Mus musculus) sion patterns and allows further characteriza- Banbury Conference Center at Cold Spring has had a prominent role in the study of tion of marked tissues and single cells. Harbor Laboratory. The attendees of the meet- human disease mechanisms throughout the Appreciation of the power of mouse genet- ing are the authors of this paper. rich, 100-year history of classical mouse genet- ics to inform the study of mammalian physi- ics, exemplified by the lessons learned from ology and disease, coupled with the advent of Is a systematic project warranted? naturally occurring mutants such as agouti4, the mouse genome sequence and the ease of A coordinated project to systematically knock reeler5 and obese6. The large-scale production producing mutated alleles, has catalyzed pub- out all mouse genes is likely to be of enormous and analysis of induced genetic mutations in lic and private sector initiatives to produce benefit to the research community, given the worms, flies, zebrafish and mice have greatly mouse mutants on a large scale, with the goal demonstrated power of knockout mice to eluci- accelerated the understanding of gene function of eventually knocking out a substantial por- date gene function, the frequency of unpre- © 2004 Nature Publishing Group in these organisms. Among the model organ- tion of the mouse genome12,13. Large-scale, dicted phenotypes in knockout mice, the isms, the mouse offers particular advantages publicly funded gene-trap programs have potential economies of scale in an organized for the study of human biology and disease: (i) been initiated in several countries, with the and carefully planned project, and the high cost the mouse is a mammal, and its development, International Gene Trap Consortium coordi- and lack of availability of knockout mice being body plan, physiology, behavior and diseases nating certain efforts and resources14–17. made in current efforts. Moreover, implement- have much in common with those of humans; Despite these efforts, the total number of ing such a systematic and comprehensive plan (ii) almost all (99%) mouse genes have knockout mice described in the literature is will greatly accelerate the translation of genome homologs in humans; and (iii) the mouse relatively modest, corresponding to only ∼10% sequences into biological insights. Knockout ES genome supports targeted mutagenesis in spe- of the ∼25,000 mouse genes. The curated cells and mice currently available from the pub- cific genes by homologous recombination in Mouse Knockout & Mutation Database lists lic and private sectors should be incorporated embryonic stem (ES) cells, allowing genes to be 2,669 unique genes (C. Rathbone, personal into the genome-wide initiative as much as altered efficiently and precisely. communication), the curated Mouse Genome possible, although some may be need to be pro- The ability to disrupt, or knock out, a spe- Database lists 2,847 unique genes, and an duced again if they were made with suboptimal cific gene in ES cells and mice was developed analysis at Lexicon Genetics identified 2,492 methods (e.g., not including a marker) or if in the late 1980s (ref. 7), and the use of knock- unique genes (B.Z., unpublished data). Most their use is restricted by intellectual property or out mice has led to many insights into human of these knockouts are not readily available to other constraints. The advantages of such a sys- biology and disease8–11. Current technology scientists who may want to use them in their tematic and coordinated effort include efficient also permits insertion of ‘reporter’ genes into research; for example, only 415 unique genes production with reduced costs; uniform use of the knocked-out gene, which can then be are represented as targeted mutations in the knockout methods, allowing for more compa- used to determine the temporal and spatial Jackson Laboratory’s Induced Mutant rability between knockout mice; and ready Resource database (S. Rockwood, personal access to mice, their derivatives and data to all The Comprehensive Knockout Mouse communication). researchers without encumbrance. Solutions to Project Consortium* The converging interests of multiple mem- the logistical, organizational and informatics *Authors and their affiliations are listed at the bers of the genomics community led to a meet- issues associated with producing, characteriz- end of the paper. ing to discuss the advisability and feasibility of ing and distributing such a large number of NATURE GENETICS VOLUME 36 | NUMBER 9 | SEPTEMBER 2004 921 COMMENTARY of the genome that can ultimately be covered by gene-trap mutations. Trapping is not entirely random but shows preference for larger transcription units and genes more highly expressed in ES cells. In recent studies, Specialized gene trapping was estimated to potentially phenotyping produce null alleles for 50–60% of all genes, perhaps more if a variety of gene-trap vectors with different insertion characteristics is Tier 2 phenotyping used17,21. RNA interference offers enormous promise for analysis of gene function in Transcriptome analysis mice22 but is not yet sufficiently developed for large-scale production of gene modifications Tier 1 phenotyping capable of reliably producing true null alleles. Tissue expression analysis Both gene-targeting and gene-trapping meth- ods are suitable for producing large numbers Knockout mice of knockout alleles, and, given their comple- mentary advantages, a combination of these methods should be used to produce the ES cells genome-wide collection of null-reporter alle- les most efficiently. Mouse genome sequence What should the deliverables be? http://www.nature.com/naturegenetics A genome-wide knockout mouse project could deliver to the research community a Figure 1 Structure of resource production in the proposed KOMP. Using the mouse genome sequence as a foundation, knockout alleles in ES cells will be produced for all genes. A subset of ES cell trove of valuable reagents and data, including knockouts will be used each year to produce knockout mice, determine the expression pattern of the targeting and trapping constructs and vec- targeted gene in a variety of tissues and carry out screening-level (Tier 1) phenotyping. In a subset of tors, mutant ES cell lines, live mice, frozen mouse lines, transcriptome analysis and more detailed system-specific (Tier 2) phenotyping will be sperm, frozen embryos, phenotypic data at a done. Finally, specialized phenotyping will be done on a smaller number of mouse lines with variety of levels and detail, and a database particularly interesting phenotypes. All stages will occur within the purview of the KOMP except for the with data visualization and mining tools. At a specialized phenotyping, which will occur in individual laboratories with particular expertise. minimum, we believe that a comprehensive genome-wide resource of mutant ES cell lines from an inbred strain, each with a different mice will draw from the experience of related be desirable, but we do not advocate this as gene knocked out, should be produced and projects in the private sector and in academia, part of the mutagenesis strategy unless the made available to the community. Choosing © 2004 Nature Publishing Group which have made or phenotyped hundreds of technological limitations currently associ- an inbred line (129/SvEvTac or C57BL/6J), knockout mice using a variety of techniques. ated with generating conditional targeted and evaluating the alternative of using F1 ES Lessons learned from these projects include the mutations on a large scale and in a cost- cells and tetraploid aggregation to provide need for redundancy at each step to mitigate effective manner can be overcome. potential time savings, merits additional sci- pipeline bottlenecks and the need for robust entific review and discussion23,24. ES cells informatics systems to track the production, A combination of methods should be used should be converted into mice at a rate con- analysis, maintenance and distribution of thou- Various methods can be used to create sistent with project funding and the ability of sands of targeting constructs, ES cells and mice. mutated alleles, including gene targeting, the worldwide scientific community to ana- gene