Modifier screens in the mouse: Time to move forward with reverse genetics

Amajor preoccupation of biology in the last decade was the sequencing of the human and mouse genomes. Having identified their constituent genes, the imperative was to understand the physiological, cellular, and biochemical roles that these genes play and to determine how this knowledge can be harnessed to improve treatment for many diseases with limited therapeutic options. Although a great deal of knowledge about fundamental processes has been gained from genetic screens in lower organisms such as yeast, flies (1), and worms, it is the mouse that holds a special place because not only has it recently become possible to perform large-scale genetic screens (2, 3) but also its genome can be manipulated in a precise manner to model human diseases, so-called reverse genetics. In a landmark article published in a recent issue of PNAS, Carpinelli et al. (4) have shown that it is possible to take a mouse model of a human disease (congenital thrombocytopenia) and to identify mutations that result in an amelioration or cure. In doing so, they have demonstrated that genetic modifier screens, which have proven so powerful in lower organisms (5, 6), are also feasible in the mouse, and they have shown that this technology might be applied to the discovery and validation of targets for the treatment of human diseases. When one throws into the mix the questions their study raises about the regulation of blood cell formation, it is clear that this article will be of interest to a wide range of geneticists and cell biologists as well as to those in the biopharmaceutical industry.

Identifying Genetic Pathways

Platelet transfusions are an effective therapy for thrombocytopenia, although limited supplies, transfusion-related toxicities, and the development of platelet alloantibodies make alternate treatments urgently required. Unfortunately, the use of thrombopoietin (TPO), the lineage-specific regulator of platelet production, has been much less effective than erythropoietin for anemia or granulocyte colony-stimulating factor for chemotherapy-induced neutropenia. Carpinelli et al. (4) used mice nullizygous for Mpl, the receptor for TPO, to identify other regulators of platelet production. Mpl^-/- mice have 10% of normal platelet numbers (7). Male Mpl^-/- mice were treated with the powerful mutagen N-ethyl-N-nitrosourea (ENU), which most commonly induces AT to TA transversions. Studies utilizing seven different phenotypic traits (8), and more recently direct sequencing (9), predict that each sperm of an ENU-treated mouse will harbor mutations in ≈50 genes. ENU-treated Mpl^-/- mice were mated with female Mpl^-/- mice to generate 1,575 G₁ offspring, which at a rate of 50 gene mutations per mouse should provide ≈2-fold coverage of the genome's 35,000 genes. G₁ mice were screened for a dominant phenotype, in this case a suppressor of the heritable thrombocytopenia. They report five mutant G₁ mice with platelet counts at least 3 SD above the mean platelet count of Mpl^-/- mice, of which two of the animals passed the characteristic on to their progeny. The two hereditable mutants were intercrossed with the 129/Sv strain to generate a cohort of F₂ mice for mapping by simple sequence-length polymorphism (SSLP) analysis. Both mutations mapped to a 10-Mb region that contained the c-Myb gene along with a host of other known or putative genes. When a candidate gene approach was used, sequencing of c-Myb revealed point mutations in the DNA-binding or leucine-zipper domains. It remains to be directly proven that these mutants are hypomorphic alleles of c-Myb; however, the observations that heterozygosity for c-Myb (10) and a hypomorphic allele of c-Myb (11) elevate platelet count provide strong circumstantial evidence. Perhaps most remarkable, homozygous c-Myb mutants had similarly high platelet counts in the presence or absence of TPO signaling. Whether TPO elevates platelet numbers by down-regulating c-Myb expression remains to be determined.

One of the significant advantages of forward genetic screens is the ability to identify genes without any prior knowledge of the genetic regulation of the phenotype. c-Myb was first recognized as a regulator of hematopoiesis in 1991 after the generation of a knockout allele by using homologous recombination in embryonic stem cells (10). Like many other gene knockouts, embryonic lethality precluded more detailed analysis of c-Myb in adult hematopoiesis. Although v-Myb was implicated in the chicken as a regulator of megakaryocyte production (12), a further 12 years passed before a knockdown of c-Myb in the mouse demonstrated its role in platelet production (11). Indeed, the hypomorphic c-Myb allele was serendipitous because the neomycin selection cassette disrupted the expression of c-Myb. The study by Carpinelli et al. (4) identified c-Myb as a regulator of platelet production without any genetic bias within 2 years and almost certainly with fewer resources than the reverse genetic approach applied to c-Myb. It should be recognized that identification of the c-Myb mutants was made much simpler with the recent publication of the knockdown allele of c-Myb (11). Given that <20% of genes in the genome have an ascribed function, a more likely scenario for forward genetic screens is that a candidate gene approach cannot be used and more refined mapping would be required, followed by sequencing of known or putative genes and complementation studies to prove that the wild-type gene could rescue the phenotype. Indeed, two other suppressor mutants have been identified in this screen that do not map to c-Myb and, not unexpectedly, have not yet been reported (D. J. Hilton, personal communication). One of the strengths of ENU mutagenesis is the ability to identify hypomorphic alleles not possible by reverse genetics without intimate structure-function knowledge. The c-Myb mutants identify amino acids critical for c-MYB function that could not be easily predicted by structure analysis. This provides the basis for design of small molecules that target these regions of c-MYB for the treatment of thrombocytopenia in humans lacking functional Tpo or Mpl genes (13, 14). Perhaps more intriguing will be the potential of targeting c-MYB for the treatment or prevention of chemotherapy-induced thrombocytopenia.

Identifying Loss-of-Function Mutations

The successful application of a modifier screen in the mouse has far-reaching implications. Until now, modifier screens have been restricted to lower organisms such as eye development of Drosophila melanogaster (15). The study by Carpinelli et al. (4) represents the first published large-scale modifier screen in mice, and this study eloquently demonstrates the advantages of modifier screens over screens performed in wild-type mice. As discussed above, the modifier screen can define genetic pathways and, in this case, demonstrates that c-MYB and MPL may lie in the same pathway, something that was not recognized in previous studies of c-Myb. In addition, modifier screens provide a more sensitive background to detect changes in phenotype. In this study, the low platelet count of the Mpl^-/- mice allowed the detection of a hypomorphic mutation of c-Myb in G₁ mice. Although ENU can induce activating mutations, loss-of-function mutations are more common and will be detectable only in mice homozygous for the mutation, i.e., G₃ mice. Although dominant screens require generation of 1,000-2,000 mice for saturation, a recessive screen requires analysis of at least 20 G₃ animals from each independent G₂ pedigree and, thus, is 20-fold more labor intensive than a dominant screen. For this reason, relatively few genome-wide recessive screens have been reported (16, 17). One method of simplifying recessive screens in the mouse is chromosome region-based screens (18). G₁ offspring from ENU-treated males are bred with mice carrying a chromosome deletion or inversion. G₂ offspring that inherit a mutation of a gene within the deleted chromosome will display the phenotype (19-21). Tagging the chromosome region with a coat color gene or fluorescent marker allows the tracking of the mutant chromosome. ENU-mutagenesis of embryonic stem (ES) cells containing chromosomal deletions also can identify recessive mutations in G₁ mice (22). The study by Carpinelli et al. demonstrates that a modifier screen can more readily identify loss-of-function mutations in G₁ mice. Heterozygous c-Myb mice on a wild-type genetic background have only a very mild increase in platelet count, and thus a screen using wild-type mice would not have identified this mutation in G₁ mice. Ironically, another group performing a genome-wide screen on a wild-type genetic background have also identified a mutant of c-Myb, but only in G₃ mice. Modifier screens are also more likely to detect genes that make the phenotype worse because the underlying mutation reduces the reserve of the organism for compensatory changes. For example, compensatory changes in Mpl^-/- mice result in normal red and white cell peripheral blood numbers despite a 10-fold reduction in hematopoietic stem cells numbers (23). Thus, genes regulating hematopoietic stem cells may be more likely detected in an ENU mutagenesis screen on an Mpl^-/- background.

Modifier Screens for Drug Discovery

The success of a modifier screen in mice paves the way for similar screens using mouse models of human disease, provided that they fulfil several prerequisites. The phenotype of the mouse model needs to be highly penetrant with little effect of genetic background to allow easy mapping of mutations. Ideally, the phenotype needs to be identifiable shortly after weaning to reduce the time animals are housed. Phenotypes that require >6 months of housing are less practical. The phenotypic screen also needs to be high-throughput to allow analysis of large numbers of mice each week. Assays such as the automated blood analysis used in this study (4) are particularly well suited. Single gene disorders would be the most applicable disease models. For example, modifiers of mouse models of hemoglobinopathies or inherited clotting disorders such as Factor V Leiden would be ideal. Mouse models of cancer such as P53 deficiency may also be fruitful for identifying genes that modify tumorigenesis. In contrast, the dissection of complex genetic traits such as diabetes by forward genetic approaches is more problematic because of effects of genetic background.

The power of ENU mutagenesis in lower organisms has been well demonstrated for the study of development. Although the majority of laboratories will continue to use the mouse for reverse genetics, the study of Carpinelli et al. (4) should provide others with the courage to pursue a similar approach in their biological process of interest. It is now time to apply modifier screens to physiologic and disease processes in the mouse, and ultimately to use this approach for the design of small molecules for therapeutic purposes.