Genetic exchange between endogenous and exogenous LINE-1 repetitive elements in mouse cells

XRCC4 in G1 suppresses homologous recombination in S/G2, in G1 checkpoint-defective cells

Non-homologous end joining (NHEJ) and homologous recombination (HR) are two pathways that can compete or cooperate for DNA double-strand break (DSB) repair. NHEJ was previously shown to act throughout the cell cycle whereas HR is restricted to late S/G2. Paradoxically, we show here that defect in XRCC4 (NHEJ) leads to over-stimulation of HR when cells were irradiated in G1, not in G2. However, XRCC4 defect did not modify the strict cell cycle regulation for HR (i.e. in S/G2) as attested by (i) the formation of Rad51 foci in late S/G2 whatever the XRCC4 status, and (ii) the fact that neither Rad51 foci nor HR (gene conversion plus single-strand annealing) events induced by ionizing radiation were detected when cells were maintained blocked in G1. Finally, both gamma-H2AX analysis and pulse field gel electrophoresis showed that following irradiation in G1, some DSBs reached S/G2 in NHEJ-defective cells. Taken together, our results show that when cells are defective in G1/S arrest, DSB produced in G1 and left unrepaired by XRCC4 can be processed by HR but in late S/G2.

Characterization of in vivo recombination activities in the mouse embryo

Homologous recombination makes use of sequence homology to repair DNA and to rearrange genetic material. In mammals, these processes have mainly been characterized using cultured cell systems. We have developed an assay that allows us to quantitatively analyze homologous recombination in vivo in the mouse embryo. Transgenic mouse lines were generated by microinjection into a fertilized mouse ovum of a vector containing two homologous LINE-1 (L1) sequences arranged as a direct repeat: these sequences can recombine with each other and with endogenous L1 sequences before, during or after integration of the vector into the genome. Using a plasmid rescue procedure, we determined the composition of the integrated vector array in several transgenic mice and their descendants. Homologous recombination frequencies were found to be strikingly high, involving 70% of integrated vectors in some arrays, with homologous deletions being five times more frequent than gene conversion without crossing-over. Interestingly, non-homologous recombination was found to be much less frequent. We also found that endogenous L1 sequences could be involved in homologous recombination events in the mouse embryo, and that the integrated arrays could be modified from generation to generation by homologous recombination between the integrated L1 sequences.

Preferential accessibility to specific genomic loci for the repair of double-strand breaks in human cells

The dynamic organization of the human genome in the nucleus is gaining recognition as a determining factor in its functional regulation. In order to be expressed, replicated or repaired, a genomic locus has to be present at the right place at the right time. In the present study, we have investigated the choice of a double-strand break (DSB) repair partner for a given genomic loci in an ATM-deficient human fibroblast cell line. We found that partner choice is restricted such that a given genomic locus preferentially uses certain sites in the genome to repair itself. These preferential sites can be in the vicinity of the damage site or megabases away or on other chromosomes entirely, while potential sites closer to the break along the length of the chromosome can be ignored. Moreover, there can be more than a 10-fold difference in usage between repair sites located only 10 kb apart. Interestingly, arms of a given chromosome are less accessible to one another than to other chromosomes. Altogether, these results indicate that the accessibility between genomic sites in the human genome during DSB repair is specific and conserved in a cell population.

CpG mutation rates in the human genome are highly dependent on local GC content

CpG dinucleotides mutate at a high rate because cytosine is vulnerable to deamination, cytosines in CpG dinucleotides are often methylated, and deamination of 5-methylcytosine (5mC) produces thymidine. Previous experiments have shown that DNA melting is the rate-limiting step in cytosine deamination. Here we show, through the analysis of human single-nucleotide polymorphisms (SNPs), that the mutation rate produced by 5mC deamination is highly dependent on local GC content. In fact, linear regression analysis showed that the log(10) of the 5mC mutation rates (inferred from SNP frequencies) had slopes of -3 when graphed with respect to the GC content of neighboring sequences. This is the ideal slope that would be expected if the correlation between CpG underrepresentation and GC content had been solely caused by DNA melting. Moreover, this same result was obtained regardless of the SNP locations (all SNPs versus only SNPs in noncoding intergenic regions, excluding CpG islands) and regardless of the lengths over which GC content was calculated (SNP sequences with a modal length of 564 bp versus genomic contigs with a modal length of 163 kb). Several alternative interpretations are discussed.

Gene repeat expansion and contraction by spontaneous intrachromosomal homologous recombination in mammalian cells

Homologous recombination (HR) is important in repairing errors of replication and other forms of DNA damage. In mammalian cells, potential templates include the homologous chromosome, and after DNA replication, the sister chromatid. Previous work has shown that the mammalian recombination machinery is organized to suppress interchromosomal recombination while preserving intrachromosomal HR. In the present study, we investigated spontaneous intrachromosomal HR in mouse hybridoma cell lines in which variously numbered tandem repeats of the mu heavy chain constant (C mu) region reside at the haploid, chromosomal immunoglobulin mu heavy chain locus. This organization provides the opportunity to investigate recombination between homologous gene repeats in a well-defined chromosomal locus under conditions in which recombinants are conveniently recovered. This system revealed several features about the mammalian intrachromosomal HR process: (i) the frequency of HR was high (recombinants represented as much as several percent of the total of recombinants and non-recombinants); (ii) the recombination process appeared to be predominantly non-reciprocal, consistent with the possibility of gene conversion; (iii) putative gene conversion tracts were long (up to 13.4 kb); (iv) the recombination process occurred with precision, initiating and terminating within regions of shared homology. The results are discussed with respect to mammalian intrachromosomal HR involving interactions both within and between sister chromatids.

Strand invasion and DNA synthesis from the two 3' ends of a double-strand break in Mammalian cells

Analysis of the crossover products recovered following transformation of mammalian cells with a sequence insertion ("ends-in") gene-targeting vector revealed a novel class of recombinant. In this class of recombinants, a single vector copy has integrated into an ectopic genomic position, leaving the structure of the cognate chromosomal locus unaltered. Thus, in this respect, the recombinants resemble simple cases of random vector integration. However, the important difference is that the two paired 3' vector ends have acquired endogenous, chromosomal sequences flanking both sides of the vector-borne double-strand break (DSB). In some cases, copying was extensive, extending >16 kb into nonhomologous flanking DNA. The results suggest that mammalian homologous recombination events can involve strand invasion and DNA synthesis by both 3' ends of the DSB. These DNA interactions are a central, predicted feature of the DSBR model of recombination.

Mechanisms involved in targeted gene replacement in mammalian cells

The "ends-out" or omega (Omega)-form gene replacement vector is used routinely to perform targeted genome modification in a variety of species and has the potential to be an effective vehicle for gene therapy. However, in mammalian cells, the frequency of this reaction is low and the mechanism unknown. Understanding molecular features associated with gene replacement is important and may lead to an increase in the efficiency of the process. In this study, we investigated gene replacement in mammalian cells using a powerful assay system that permits efficient recovery of the product(s) of individual recombination events at the haploid, chromosomal mu-delta locus in a murine hybridoma cell line. The results showed that (i) heteroduplex DNA (hDNA) is formed during mammalian gene replacement; (ii) mismatches in hDNA are usually efficiently repaired before DNA replication and cell division; (iii) the gene replacement reaction occurs with fidelity; (iv) the presence of multiple markers in one homologous flanking arm in the replacement vector did not affect the efficiency of gene replacement; and (v) in comparison to a genomic fragment bearing contiguous homology to the chromosomal target, gene targeting was only slightly inhibited by internal heterology (pSV2neo sequences) in the replacement vector.

Potential gene conversion and source genes for recently integrated Alu elements

Alu elements comprise >10% of the human genome. We have used a computational biology approach to analyze the human genomic DNA sequence databases to determine the impact of gene conversion on the sequence diversity of recently integrated Alu elements and to identify Alu elements that were potentially retroposition competent. We analyzed 269 Alu Ya5 elements and identified 23 members of a new Alu subfamily termed Ya5a2 with an estimated copy number of 35 members, including the de novo Alu insertion in the NF1 gene. Our analysis of Alu elements containing one to four (Ya1-Ya4) of the Ya5 subfamily-specific mutations suggests that gene conversion contributed as much as 10%-20% of the variation between recently integrated Alu elements. In addition, analysis of the middle A-rich region of the different Alu Ya5 members indicates a tendency toward expansion of this region and subsequent generation of simple sequence repeats. Mining the databases for putative retroposition-competent elements that share 100% nucleotide identity to the previously reported de novo Alu insertions linked to human diseases resulted in the retrieval of 13 exact matches to the NF1 Alu repeat, three to the Alu element in BRCA2, and one to the Alu element in FGFR2 (Apert syndrome). Transient transfections of the potential source gene for the Apert's Alu with its endogenous flanking genomic sequences demonstrated the transcriptional and presumptive transpositional competency of the element.

A double-strand break in a chromosomal LINE element can be repaired by gene conversion with various endogenous LINE elements in mouse cells

A double-strand break (DSB) in the mammalian genome has been shown to be a very potent signal for the cell to activate repair processes. Two different types of repair have been identified in mammalian cells. Broken ends can be rejoined with or without loss or addition of DNA or, alternatively, a homologous template can be used to repair the break. For most genomic sequences the latter event would involve allelic sequences present on the sister chromatid or homologous chromosome. However, since more than 30% of our genome consists of repetitive sequences, these would have the option of using nonallelic sequences for homologous repair. This could have an impact on the evolution of these sequences and of the genome itself. We have designed an assay to look at the repair of DSBs in LINE-1 (L1) elements which number 10(5) copies distributed throughout the genome of all mammals. We introduced into the genome of mouse epithelial cells an L1 element with an I-SceI endonuclease site. We induced DSBs at the I-SceI site and determined their mechanism of repair. We found that in over 95% of cases, the DSBs were repaired by an end-joining process. However, in almost 1% of cases, we found strong evidence for repair involving gene conversion with various endogenous L1 elements, with some being used preferentially. In particular, the T(F) family and the L1Md-A2 subfamily, which are the most active in retrotransposition, appeared to be contributing the most in this process. The degree of homology did not seem to be a determining factor in the selection of the endogenous elements used for repair but may be based instead on accessibility. Considering their abundance and dispersion, gene conversion between repetitive elements may be occurring frequently enough to be playing a role in their evolution.

Double-strand break-induced recombination in eukaryotes

Genetic recombination is of fundamental importance for a wide variety of biological processes in eukaryotic cells. One of the major questions in recombination relates to the mechanism by which the exchange of genetic information is initiated. In recent years, DNA double strand breaks (DSBs) have emerged as an important lesion that can initiate and stimulate meiotic and mitotic homologous recombination. In this review, we examine the models by which DSBs induce recombination, describe the types of recombination events that DSBs stimulate, and compare the genetic control of DSB-induced mitotic recombination in budding and fission yeasts.

Ectopic gene targeting exhibits a bimodal distribution of integration in murine cells, indicating that both intra- and interchromosomal sites are accessible to the targeting vector

Ectopic gene targeting is an alternative outcome of the gene targeting process in which the targeting vector acquires sequences from the genomic target but proceeds to integrate elsewhere in the genome. Using two-color fluorescent in situ hybridization analysis, we have determined the integration sites of the gene targeting vector with respect to the target locus in a murine fibroblast line (LTA). We found that for ectopic gene targeting the distribution of integration sites was bimodal, being either within 3 Mb of the target or on chromosomes distinct from the chromosome carrying the target locus. Inter- and intrachromosomal sites appeared to be equally accessible to the targeting vector, with site-specific variations. Interestingly, interphase analysis indicated that vector sequences which had integrated ectopically in chromosomes other than the target colocalized with the target locus at a significant frequency compared to that of colocalization to random unlinked loci. We propose that ectopic gene targeting could be used to determine which chromosomal domains within the genome are accessible to a given genetic locus. Thus, recombination access mapping may present a new paradigm for the analysis of DNA accessibility and interaction within the genome.

LINE-I element insertion at the t(11;22) translocation breakpoint of a desmoplastic small round cell tumor

A t(11;22)(p13;p12) chromosomal translocation, juxtaposing the Wilms' tumor (WT1) and Ewing's sarcoma (EWS) genes, is the cytogenetic hallmark of desmoplastic small round cell tumor (DSRCT), a primitive multiphenotypic sarcoma arising in serosal tissues. Chimeric transcripts generated by this rearrangement encode an aberrant transcription factor that fuses the 5' region of EWS with a 3' WT1 segment. We describe the insertion of a LINE-I DNA mobile genetic element at the genomic breakpoint of a DSRCT chromosomal translocation. A 480 bp heterologous DNA segment with homology to the LINE-I DNA consensus sequence was located between EWS intron 8 and WT1 exon 8 in the productively rearranged allele. Sequence homology corresponded to the LINE-I ORF-2, which encodes a protein with reverse-transcriptase activity. The heterologous inserted fragment was not evident in the germline of normal tissue from the patient, suggesting that transposition occurred in somatic cells, possibly during the process of chromosomal rearrangement. This case represents the first example of LINE-I DNA transposition at the fusion site of a tumor-associated chromosomal rearrangement.

Homologous junctions formed between a vector and human genomic repetitive LINE-1 elements as a result of one-sided invasion

Studies on homologous recombination in mammalian cells between an exogenous DNA molecule containing a double-strand break and a homologous genomic sequence have indicated that there were at least two distinct types of homologous recombination processes, one that involved the formation of two homologous junctions and another that involved the formation of one homologous junction and one illegitimate junction. Both types of events are produced in gene targeting experiments. We have proposed a model to account for the later process called one-sided invasion. One-sided invasion has now been reported in numerous species belonging to different phyla and appears to be a universal mechanism. It has also been observed in normal human germ cells. The role of one-sided invasion is still unknown. Using a recombination assay between LINE-1 elements from the human genome and exogenous LINE-1 sequences, we have characterized the process of homologous junction formation in one-sided invasion. We found that at each of the homologous junctions, variable lengths of the vector L1 sequences had been replaced by genomic L1 sequences. We also found a homologous junction that involved three partners, suggesting that the homologous end could be released and become available for a second round of interaction.

Gene conversion as a secondary mechanism of short interspersed element (SINE) evolution

The Alu repetitive family of short interspersed elements (SINEs) in primates can be subdivided into distinct subfamilies by specific diagnostic nucleotide changes. The older subfamilies are generally very abundant, while the younger subfamilies have fewer copies. Some of the youngest Alu elements are absent in the orthologous loci of nonhuman primates, indicative of recent retroposition events, the primary mode of SINE evolution. PCR analysis of one young Alu subfamily (Sb2) member found in the low-density lipoprotein receptor gene apparently revealed the presence of this element in the green monkey, orangutan, gorilla, and chimpanzee genomes, as well as the human genome. However, sequence analysis of these genomes revealed a highly mutated, older, primate-specific Alu element was present at this position in the nonhuman primates. Comparison of the flanking DNA sequences upstream of this Alu insertion corresponded to evolution expected for standard primate phylogeny, but comparison of the Alu repeat sequences revealed that the human element departed from this phylogeny. The change in the human sequence apparently occurred by a gene conversion event only within the Alu element itself, converting it from one of the oldest to one of the youngest Alu subfamilies. Although gene conversions of Alu elements are clearly very rare, this finding shows that such events can occur and contribute to specific cases of SINE subfamily evolution.

Integration of a vector containing a repetitive LINE-1 element in the human genome

Mammalian cells contain numerous nonallelic repeated sequences, such as multicopy genes, gene families, and repeated elements. One common feature of nonallelic repeated sequences is that they are homeologous (not perfectly identical). Our laboratory has been studying recombination between homeologous sequences by using LINE-1 (L1) elements as substrates. We showed previously that an exogenous L1 element could readily acquire endogenous L1 sequences by nonreciprocal homologous recombination. In the study presented here, we have investigated the propensity of exogenous L1 elements to be involved in a reciprocal process, namely, crossing-overs. This would result in the integration of the exogenous L1 element into an endogenous L1 element. Of over 400 distinct integration events analyzed, only 2% involved homologous recombination between exogenous and endogenous L1 elements. These homologous recombination events were imprecise, with the integrated vector being flanked by one homologous and one illegitimate junction. This type of structure is not consistent with classical crossing-overs that would result in two homologous junctions but rather is consistent with one-sided homologous recombination followed by illegitimate integration. Contrary to what has been found for reciprocal homologous integration, the degree of homology between the exogenous and endogenous L1 elements did not seem to play an important role in the choice of recombination partners. These results suggest that although exogenous and endogenous L1 elements are capable of homologous recombination, this seldom leads to crossing-overs. This observation could have implications for the stability of mammalian genomes.

Integration of a vector containing a repetitive LINE-1 element in the human genome

Mammalian cells contain numerous nonallelic repeated sequences, such as multicopy genes, gene families, and repeated elements. One common feature of nonallelic repeated sequences is that they are homeologous (not perfectly identical). Our laboratory has been studying recombination between homeologous sequences by using LINE-1 (L1) elements as substrates. We showed previously that an exogenous L1 element could readily acquire endogenous L1 sequences by nonreciprocal homologous recombination. In the study presented here, we have investigated the propensity of exogenous L1 elements to be involved in a reciprocal process, namely, crossing-overs. This would result in the integration of the exogenous L1 element into an endogenous L1 element. Of over 400 distinct integration events analyzed, only 2% involved homologous recombination between exogenous and endogenous L1 elements. These homologous recombination events were imprecise, with the integrated vector being flanked by one homologous and one illegitimate junction. This type of structure is not consistent with classical crossing-overs that would result in two homologous junctions but rather is consistent with one-sided homologous recombination followed by illegitimate integration. Contrary to what has been found for reciprocal homologous integration, the degree of homology between the exogenous and endogenous L1 elements did not seem to play an important role in the choice of recombination partners. These results suggest that although exogenous and endogenous L1 elements are capable of homologous recombination, this seldom leads to crossing-overs. This observation could have implications for the stability of mammalian genomes.

One-sided invasion events in homologous recombination at double-strand breaks

Two classes of homologous recombination mechanism for repair of double-strand breaks (DSBs) have been described in eukaryotes so far. One is conservative and has been explained by the double-strand break repair model (Szostak et al., 1983), whereas the other one is non-conservative and has been explained by the single-strand annealing model (Lin et al., 1984). Here, we will review data supporting the existence of another homologous recombination mechanism for double-strand break repair. We will present the one-sided invasion model that we have proposed to explain this mechanism and discuss its potential implication in various homologous recombination events.

Genome organization of repetitive elements in the rodent, Peromyscus leucopus

To document the frequency and distribution of repetitive elements in Peromyscus leucopus, the white-footed mouse, a cosmid genomic library was examined. Two thousand thirteen randomly chosen recombinants, with an average insert size of 35 kb and representing 2.35% of the haploid genome of P. leucopus, were screened with probes representing microsatellites, tandem repeats, and transposable elements. Of the four dinucleotides, (GT)n was present in 87% of the clones, (CT)n was present in 59% of the clones, and (AT)n and (GC)n each was represented in our sample by a single clone (0.05%). (TCC)n was present in 8% of the clones. Of the tandem repeats, the 28S ribosomal probe and the (TTAGGG)n telomere probe were not represented in the library, whereas a heterochromatic fragment was present in 9% of the clones. A transposable element, mys, was estimated to occur in 4700 copies, whereas a long interspersed element (LINE) was estimated to occur in about 41,000 copies per haploid genome. LINE and mys occurred together in the same clones more frequently than expected on the basis of chance. Hybridizing the library to genomic DNA from P. leucopus, Reithrodontomys fulvescens, Mus musculus, and human produced general agreement between phylogenetic relatedness and intensity of hybridization. However, dinucleotide repeats appeared to account for a disproportionately high number of positive clones in the more distantly related taxa.

End extension repair of introduced targeting vectors mediated by homologous recombination in mammalian cells

We have studied the mechanism of targeted recombination in mammalian cells using a hemizygous adenine phosphoribosyltransferase-deficient (APRT-) Chinese hamster ovary (CHO) cell mutant as a recipient. Three structurally different targeting vectors with a 5' or a 3', or both, end-deleted aprt sequence, in either a closed-circular or linear form, were transfected to the cells with a mutated aprt gene by electroporation. APRT-positive (APRT+) recombinant clones were selected and analyzed to study the gene correction events of the deletion mutation. Some half of 58 recombinant clones obtained resulted from corrections of the deleted chromosomal aprt gene by either gene replacement or gene insertion, a mechanism which is currently accepted for homologous recombination in mammalian cells. However, the chromosomal sequence in the remaining half of the recombinants remained uncorrected but their truncated end of the aprt gene in the incoming vectors was corrected by extending the end beyond the region of homology to the target locus; the corrected vector was then randomly integrated into the genome. This extension, termed end extension repair, was observed with all three vectors used and was as far as 4.6-kilobase (kb) or more long. It is evident that the novel repair reaction mediated by homologous recombination, in addition to gene replacement and gene insertion, is also involved in gene correction events in mammalian cells. We discuss the model which may account for this phenomenon.

X rays induce interallelic homologous recombination at the human thymidine kinase gene

We have developed a human lymphoblast cell line for the study of interchromosomal homologous recombination at the endogenous thymidine kinase (tk) gene on chromosome 17 (M. B. Benjamin, H. Potter, D. W. Yandell, and J. B. Little, Proc. Natl. Acad. Sci. USA 88:6652-6656, 1991). This cell line (designated 6:86) carries unique heterozygous frameshift mutations in exons 4 and 7 of its endogenous tk alleles and can revert to TK+ by frame-restoring mutations, gene conversion, or reciprocal recombination. Line 6:86 reverts spontaneously to TK+ at a frequency of 10(-7) to 10(-8), and exposures to X-irradiation or the frameshift mutagen ICR-191 induce increased reversion frequencies in a dose-dependent manner. Another cell line (designated 4:2) carries a homozygous exon 7 frameshift and is not expected to revert through mechanisms other than frame-restoring mutation. Line 4:2 reverts to TK+ at a lower spontaneous frequency than does 6:86 but can be induced with similar kinetics by ICR-191. In contrast to line 6:86, however, X rays did not induce detectable reversion of line 4:2. We have characterized a number of 6:86-derived revertants by means of restriction fragment length polymorphism analysis at tk and linked loci, single-strand conformation polymorphisms, and direct transcript sequencing. For X rays, most revertants retain both original mutations in the genomic DNA, and a subset of these frameshift-retaining revertants produce frameshift-free message, indicating that reversion is the result of reciprocal recombination within the tk gene. Frame-restoring point mutations, restoration of original sequences, and phenocopy reversion by acquisition of aminopterin resistance were also found among X-ray-induced revertants, whereas the ICR-191-induced revertants examined show only loss of the exon 7 frameshift.

X rays induce interallelic homologous recombination at the human thymidine kinase gene

We have developed a human lymphoblast cell line for the study of interchromosomal homologous recombination at the endogenous thymidine kinase (tk) gene on chromosome 17 (M. B. Benjamin, H. Potter, D. W. Yandell, and J. B. Little, Proc. Natl. Acad. Sci. USA 88:6652-6656, 1991). This cell line (designated 6:86) carries unique heterozygous frameshift mutations in exons 4 and 7 of its endogenous tk alleles and can revert to TK+ by frame-restoring mutations, gene conversion, or reciprocal recombination. Line 6:86 reverts spontaneously to TK+ at a frequency of 10(-7) to 10(-8), and exposures to X-irradiation or the frameshift mutagen ICR-191 induce increased reversion frequencies in a dose-dependent manner. Another cell line (designated 4:2) carries a homozygous exon 7 frameshift and is not expected to revert through mechanisms other than frame-restoring mutation. Line 4:2 reverts to TK+ at a lower spontaneous frequency than does 6:86 but can be induced with similar kinetics by ICR-191. In contrast to line 6:86, however, X rays did not induce detectable reversion of line 4:2. We have characterized a number of 6:86-derived revertants by means of restriction fragment length polymorphism analysis at tk and linked loci, single-strand conformation polymorphisms, and direct transcript sequencing. For X rays, most revertants retain both original mutations in the genomic DNA, and a subset of these frameshift-retaining revertants produce frameshift-free message, indicating that reversion is the result of reciprocal recombination within the tk gene. Frame-restoring point mutations, restoration of original sequences, and phenocopy reversion by acquisition of aminopterin resistance were also found among X-ray-induced revertants, whereas the ICR-191-induced revertants examined show only loss of the exon 7 frameshift.

Modulation of gene activity by consecutive gene targeting of one creatine kinase M allele in mouse embryonic stem cells

The cytosolic creatine kinases (CK's; EC 2.7.3.2) BB, BM and MM are dimeric isoenzymes which have an important role in energy metabolism and display characteristic tissue- and stage-specific patterns of expression in mammals. To study the functional role of the distribution of the CK isoenzymes we have focussed on the modulation of expression of the genes encoding the individual B and M subunits, starting at the muscle creatine kinase (CKM) gene which is transcriptionally inactive during early embryogenesis. Using repeated rounds of gene targeting in mouse embryonic stem (ES) cells, two types of mutant cell lines were obtained. First, we generated a cell line in which insertion of a neomycin resistance (neor) gene had disrupted one of the CKM alleles. Subsequently, from this cell line, following introduction of an insertion type vector designed for replacement of the muscle specific CKM-enhancer by the constitutively acting polyoma virus enhancer PyF441, several independent doubly targeted clones were isolated which all had insertions in the previously neo-disrupted CKM allele. In some of these ES clones, the targeted enhancer replacement resulted in gene correction and functional activation of the silent CKM gene. Dimerisation between the ectopically expressed CKM subunits and CKB subunits which are normally present at high levels in ES cells, led to the formation of the BM isoform of CK in these clones.