Search

Gene targeting was used to introduce nonselectable genetic changes into chromosomal loci in mouse embryo-derived stem cells. The nonselectable markers were linked to a selectable marker in both insertion- and replacement-type vectors, and the transfer of the two elements to the Hprt locus was assayed. When insertion vectors were used as substrates, the frequency of transfer was highly dependent upon the distance between the nonselectable marker and the double-strand break in the vector. A marker located close to the vector ends was frequently lost, suggesting that a double-strand gap repair activity is involved in vector integration. When replacement vectors were used, cotransfer of a selectable marker and a nonselectable marker 3 kb apart was over 50%, suggesting that recombination between vector and target often occurs near the ends of the vector. To illustrate the use of replacement vectors to transfer specific mutations to the genome, we describe targeting of the delta F508 mutation to the CFTR gene in mouse embryo-derived stem cells.

Mutations were targeted to the Hprt locus in murine embryonic stem cells by using sequence replacement vectors. When the vector was designed such that the mutated sequences were flanked on both sides by several kilobases of DNA homologous to the target locus, replacement of chromosomal sequences with the exogenous DNA occurred with precision. If, on the other hand, the target-homologous DNA on one arm of the vector was reduced to below 1 kb in length, the fidelity of recombination was diminished.

Heteroduplexes were prepared from two plasmids, pRH4-14/TK and pRH5-8/TK, containing different amber mutations in the neomycin resistance gene (Neor). The Neor gene was engineered to be expressed in both bacterial and mammalian cells. A functional Neor gene conferred kanamycin resistance to bacteria and resistance to the drug G418 to mammalian cells. In addition, the plasmids contained restriction site polymorphisms which did not confer a selectable phenotype but were used to follow the pattern of correction of mismatched bases in the heteroduplexes. In a direct comparison of the efficiency of transforming mouse LMtk- cells to G418r, the injection of heteroduplexes of pRH4-14/TK-pRH5-8/TK was 10-fold more efficient than the coinjection of pRH4-14/TK and pRH5-8/TK linear plasmid DNA. In fact, injection of 5 to 10 molecules of heteroduplex DNA per cell was as efficient in transforming LMtk- cells to G418r as the injection of 5 to 10 molecules of linear plasmid DNA per cell containing a wild-type Neor gene. To determine the pattern of mismatch repair of the injected heteroduplexes, plasmids were "rescued" from the G418r cell lines. From this analysis we conclude that the generation of wild-type Neor genes from heteroduplex DNA proceeds directly by correction of the mismatched bases, rather than by alternative mechanisms such as recombination between the injected heteroduplexes. Our finding that a cell can efficiently correct mismatched bases when confronted with preformed heteroduplexes suggests that this experimental protocol could be used to study a wide range of DNA repair mechanisms in cultured mammalian cells.

We have examined the mechanism of homologous recombination between plasmid molecules coinjected into cultured mammalian cells. Cell lines containing recombinant DNA molecules were obtained by selecting for the reconstruction of a functional Neor gene from two plasmids that bear different amber mutations in the Neor gene. In addition, these plasmids contain restriction-length polymorphisms within and near the Neor gene. These polymorphisms did not confer a selectable phenotype but were used to identify and categorize selected and nonselected recombinant DNA molecules. The striking conclusion from this analysis is that the predominant mechanism for the exchange of information between coinjected plasmid molecules over short distances (i.e., less than 1 kilobase) proceeds via nonreciprocal homologous recombination. The frequency of homologous recombination between coinjected plasmid molecules in cultured mammalian cells is extremely high, approaching unity. We demonstrate that this high frequency requires neither a high input of plasmid molecules per cell nor a localized high concentration of plasmid DNA within the nucleus. Thus, it appears that plasmid molecules, once introduced into the nucleus, have no difficulty seeking each other out and participating in homologous recombination even in the presence of a vast excess of host DNA sequences. Finally, we show that most of the homologous recombination events occur within a 1-h interval after the introduction of plasmid DNA into the cell nucleus.