Introduction

The creation of novel phenotypes and identification of the mutations responsible has led to some of the most important advances in biology. In mammalian systems, phenotypes may be created either by modifying the germline or by modifying cultured cells, using either chemical or insertional mutagens. Germline mutagenesis can reveal phenotypes that are impossible to produce in culture (e.g., behavioral phenotypes or phenotypes related to development). However, many basic biological phenomena such as cell cycle regulation and cell death can be better explored using cultured cells. Germline mutagenesis is limited by the very low rate at which mutations can be produced (by present estimates, perhaps 100 times the background rate). In vitro mutagenesis in cultured cells can be expected to yield far higher rates of mutation than the germline approach. Because of the biological complexity of animal models and the time and expense involved in breeding, cells and cell lines are often preferred when cellular questions are at issue. Unfortunately, forward genetic studies in cultured mammalian cells, as presently practiced, have serious limitations (1,2,3,4,5,6,7,8,9,10).

Given that the cells used are diploid, only a single copy of each autosomal gene is generally destroyed by mutagenesis in vitro, and breeding cannot be used to achieve homozygosity for mutations as it can be in germline mutagenesis. In many instances, a dominant phenotype is not rendered and, indeed, cannot be rendered by mutation. This problem has led to develop antisense library approaches (5,9,11,12), and rescue methods have been developed to quickly identify which antisense cDNA is responsible for a given phenotype (9). But for many genes, antisense RNA expression is not effective in blocking expression. The same problem is present if a modified method, random homozygous knockout (RHKO), is used to produce a phenotype (8). In recent years, the use of small interfering RNA (siRNA) library has attracted a great deal of attention as a powerful tool for functional identification of genes (13,14,15). siRNA is clearly much more efficient than antisense RNA in knocking down gene expression; however, there are still some drawbacks. One study has shown that sequence identity of as few as 11 to 12 nucleotides between an interfering RNA and a messenger RNA (mRNA) may be sufficient for interference to occur (16). If this is true, then cross-reactivity, which is referred to as the off-target effect, could be a substantial problem (17). Although RNA interference (RNAi) targeting efficiency is much better than antisense, it is clear—as in the case of worms or flies—that different genes in mammalian cells are turned down with differing efficiencies. The potential interferon response to siRNA expression in mammalian cells could also interfere with some genetic screens (18,19,20). Recently, two groups of investigators generated embryonic stem (ES) cell libraries with genome-wide biallelic mutations (21,22). The increase in the rate of loss-of-hetero-zygosity (LOH) in Bloom's syndrome gene (Blm) deficient cells was used in their strategy to generate these biallelic mutations. Blm-deficient cells carrying heterozygous mutations segregate into homozygous daughters in vitro and in vivo, presumably through mitotic recombination between nonsister chromatids. One group used Blm knockout ES cells and another group developed a tetracycline-inducible system to transiently knockout Blm in ES cells. The rate of LOH in Blm-deficient ES cells is 4.2 × 10⁻⁴ and 2.3 × 10⁻⁴ determined by the two groups, respectively, which is approximately a 20-fold increase from the wild-type ES cells. In order to allow the LOH to occur in most of the loci, both groups passed the mutated cells a number of generations. As a result, the libraries contained approximately 5 × 10⁸ cells, and among them, 0.01%–0.1% cells containing biallelic mutations. Although the rates of LOH were quite low in these studies, considering the hypomorphic allele used by the first group and possible leaky inducible system used by the second group, the rate of mitotic recombination could have been higher, and therefore, using LOH to generate biallelic mutations can be a very hopeful approach. However, how the method will practically be applied to other cell types remains to be examined.

Retroviral insertion can create a single mutation in a cell, and the inserted retrovirus offers a tag with which to find the mutated gene with relative ease (23,24). In principle, insertional mutagenesis would be very effective if haploid mammalian cells could be created in culture. High rates of mutations can be most efficiently achieved by chemical mutagens because of their potency and because they can be applied to culture cells multiple times.