Therapeutic Gene Targeting

Home , Electroporation, Gene knockout, Gene targeting

´ ˜ RJ Yanez and ACG Porter Gene Targeting Group, MRC Clinical Sciences Centre, Imperial College School of Medicine, Hammersmith Hospital, London W12 0NN, UK

Gene targeting is the use of homologous recombination to without dramatic improvements in targeting efficiency. Ex make defined alterations to the genome. One of the pos- vivo approaches might more realistically be considered, sible outcomes of gene targeting is the accurate correction but would benefit from progress in the isolation and growth of genetic defects, and this would make it the ideal method of somatic stem cells and improvements in targeting of gene therapy for single gene disorders. While gene tar- efficiency. We provide here a brief review of the challenges geting has been achieved both in human cell lines and in of gene therapy by gene targeting. This is followed by a nontransformed, primary human cells, its low efficiency has critical overview of recent developments in gene targeting been a major limitation to its therapeutic potential. Gene techniques, and in our understanding of the underlying pro- therapy by in vivo gene targeting is therefore impractical cesses of homologous and nonhomologous recombination.

Keywords: gene targeting; gene therapy; stem cells; homologous recombination; nonhomologous recombination; RAD genes

Introduction: ‘pros’ and ‘cons’ of gene require the use of different therapeutic DNA for different targeting for gene therapy patients with the same disease. Also, although it can be a high fidelity procedure,5 mutations associated with gene Genetic defects in human cells can be addressed by gene targeting have been observed.6 The main handicap, how- therapy in one of two ways: correction of the faulty gene ever, lies in the inefficiency of gene targeting which typi- (by gene targeting) or addition of a functional or thera- cally ranges from 10−5 to 10−7 targeted cells per trans- peutic gene (gene augmentation).1–3 Gene targeting is fected cell.7 The scale of this handicap is apparent when currently the only procedure that can produce pre- one considers that gene augmentation involving viral defined alterations in the genome of eukaryotic cells. As vectors can result in high percentages (Ͼ10%) of cells such, gene targeting is an attractive approach to gene incorporating the gene into the genome. therapy of genetic diseases because: (1) it can lead to the Two complementary strategies can be envisaged to accurate correction of the defect in the target locus of overcome this problem: (1) improvement of gene tar- interest; (2) it can correct both recessive and dominant geting techniques, leading to higher frequencies; and (2) mutations, unlike gene augmentation, which is restricted targeting of human somatic stem cells in culture, includ- to recessive defects; (3) it is not restricted by the size of ing their selective amplification for return to the donor. the mutated gene since only the damaged portion need Mouse gene knockout experiments are the paradigm of be considered; gene augmentation requires delivery of what can be achieved when stem cells are targeted in cul- the entire coding region which may be impractical for ture and returned to a host. Recent developments in the large genes; (4) actual correction of the faulty gene would characterisation and manipulation of human somatic permanently reverse the genetic disorder by restoring stem cells suggest that analogous procedures might production of the normal protein product; the corrected eventually be adopted for gene correction in specific gene would be under the control of its own regulatory tissues. Such developments will not be considered here sequences, thereby avoiding the potential problems of but have been reviewed elsewhere for hemopoietic cells8 nonphysiological gene expression and long-term gene and keratinocytes.9 In this review, we intend to provide inactivation frequently associated with gene aug- an overview of current advances in gene targeting and mentation procedures;4 and (5) it avoids the risk, associa- DNA recombination in higher eukaryotes, and of their ted with many gene augmentation approaches, of a possible practical applications in gene therapy. Gene tar- harmful mutation caused by the chance integration of the geting as an approach to gene therapy has been pre- therapeutic DNA at a sensitive genomic site such as an viously reviewed.3,10 Other reviews cover multiple oncogene. aspects of gene targeting.2,6,7,11–14 A potential drawback of gene correction by gene targeting is that it requires previous knowledge about the Practical aspects of gene targeting location and type of mutation to be corrected, and may Gene targeting relies on the capacity of cells to carry out homologous recombination (HR), a process in which two Correspondence: ACG Porter DNA molecules of similar sequence interact and undergo Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 150 either reciprocal exchange (as in meiotic recombination) sole proof that gene targeting has occurred, especially or unidirectional transfer (as in gene conversion or when nonclonal populations of cells are used. recombinational DNA repair) of genetic information. HR From the above it is clear that the overall targeting does not require that both DNA molecules are chromo- efficiency is determined by the efficiency of DNA deliv- somal, and this is exploited in gene targeting where a ery to the nucleus and of HR once it is there (these should chromosomal ‘target’ sequence is modified by HR with be maximised) and that, for practical reasons, the an exogenous DNA molecule, or targeting construct. efficiency of NHR should be minimised. Any protocol Depending on the design and use of the targeting con- that can improve one or more of these three key variables struct, a wide range of genetic modifications can be without adversely affecting the others is highly desirable. made.14 Factors that have been shown to improve gene tar- Targeting constructs are basically vectors that carry a geting frequencies are the use of long regions of homo- segment of DNA homologous to the target including, logy and isogenic DNA in the targeting construct. Tar- where desired, a genetic alteration to be introduced into geting efficiency increases exponentially with the length the target locus. HR between the target and this segment of homology up to values of around 14 kbp.15 DNA poly- can take the form of insertion or replacement events. morphisms seem to be detrimental15,16 and can be avoi- Insertion events involve a single reciprocal exchange, ded by using isogenic DNA, ie DNA isolated from cells resulting in the insertion of the entire construct into the of the same individual as those to be targeted. A DNA target locus and duplication of the region of homology. clone is only isogenic to one of the two alleles in a diploid Replacement events involve two reciprocal exchanges, or cell, and preferential targeting of the isogenic allele has a nonreciprocal gene conversion event, and result in been shown to occur in chicken DT40 cells.17 An alterna- replacement of a stretch of the target homology with a tive to the cumbersome procedure of making and screen- corresponding stretch from the targeting construct. Posi- ing isogenic genomic DNA libraries is amplification by tively selectable marker genes are usually included in tar- PCR of the target region of interest.17,18 It is also known geting constructs so that drugs can be used selectively to that linearisation of the targeting construct increases the 19–21 kill untransfected cells. The ways in which such events frequency of gene targeting. The choice of target can lead to repair of a target gene have been described locus and details of construct design appear to greatly elsewhere.10 influence gene targeting efficiencies but not in any clearly 7,11 Targeting constructs are delivered to the target cells predictable way. most often by electroporation, although other delivery methods can be used. Electroporation is convenient, but Gene therapy as an extension of existing it is not particularly efficient; typical frequencies of cells gene targeting experiments incorporating DNA in their genomes are 10−3 to 10−5 stable transfectants per electroporated cell. In mam- Gene targeting is best known for its widespread success malian cells, the great majority of stable transfectants are in generating strains of mice with defined genetic alter- the result of nonhomologous recombination (NHR, or ations (usually gene knockouts). The development of procedures for the handling and amplification of mouse random integration) of the construct into the genome; embryonic stem (ES) cells in culture, combined with the integration by HR is generally much less efficient. Thus gene targeting technology, has permitted the almost rou- ratios of gene targeting to nonhomologous recombination tine production of mouse mutants for genes of interest. (GT:NHR) in the range of 1:30–1:40 000 have been Established ES cell lines can thus be manipulated and reported.7,11 amplified in culture without losing their stem cell proper- Low GT:NHR ratios make gene targeting more com- ties and then reintroduced into the embryonic inner cell plex since the targeted cells have to be separated from mass where they contribute to the developing embryo.13 random integrants if purified populations of targeted The isolation of ES cell lines derived from rabbits22 and cells are required. Several selection and screening pro- 23 10,11 primates suggests that genetic modification in other ani- cedures have been devised for this purpose. Selection mals by gene targeting will be possible. procedures (eg ‘promoter-trap’ and ‘positive/negative Targeting in ES cells permits genetic modifications to selection’) rely on selectable marker genes carefully pos- be transferred to the germ line which, for ethical and/or itioned in the targeting construct to allow selective killing legal reasons, would be unacceptable in human therapy of nonhomologous recombinants. Targeted cells are in most societies.24 Acceptable gene therapy therefore amenable to direct selection in the few cases when the requires modification of a patient’s somatic cells and target gene has a selectable phenotype (eg HPRT). Screen- more specifically, if this is to have long-term benefits, ing procedures are mostly based on polymerase chain those cells capable of both self-renewal and differen- reaction (PCR) assays or genomic Southern blots to detect tiation into the affected tissue-type (somatic stem cells). DNA fragments unique to targeted clones. Phenotypic The combination of inefficient gene targeting and low analysis by methods such as FACS or ELISA has also natural abundance of somatic stem cells makes in vivo been used to screen for targeting events inducing the gene targeting (delivery of the targeting construct expression of novel membrane epitopes or the synthesis directly to the affected tissue of a patient) an impractical of secreted proteins, respectively. Following any initial approach at present. An ex vivo strategy, on the other selection or screening procedure, Southern analysis of hand, is attractive because of its parallels with established genomic DNA from a purified clone is the most reliable methods for generating targeted mice. This would and widely accepted way to confirm that the desired tar- involve the growth and gene targeting of a patient’s geted modification has taken place. PCR assays, while somatic stem cells in culture, including the expansion of extremely valuable for initial screens, are prone to con- targeted derivatives for return to the patient. The limiting tamination or other artefacts making them unsuitable as factor in this approach is not the efficiency of gene Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 151 targeting but the methodology for the isolation, growth can undergo targeted modifications with GT:NHR ratios and manipulation of somatic stem cells. An intermediate (1:40–1:6550) and absolute targeting frequencies approach might therefore be envisaged where partially (1.2 × 10−8–2.7 × 10−5) similar to those described for purified stem cells are maintained and manipulated in murine cells.7,11 As for primary cells, retinal epithelial culture for a limited period. Here again, an improvement cells, foreskin keratinocytes and embryonic lung fibro- in targeting efficiency would be required but not as much blasts have been modified by gene targeting with absol- as for a purely in vivo approach. ute frequencies between 6 × 10−8 and 4.3 × 10−6.32,33 These results show that cell immortalisation is not required to Gene targeting in different cell types achieve normal frequencies of chromosomal gene targeting in human cells. From the preceding discussion, it is clearly important to From the above considerations it appears that gene tar- consider the extent to which the successes of gene targeting can be performed with similar efficiencies in a var- geting in mouse ES cells can be reproduced in human iety of cell types from different species, whether transfor- somatic cells. It is also of interest to assess the influence of med or primary, totipotent (ES cells) or monopotent cell-type and state of differentiation and immortalisation. (differentiated). Ex vivo gene targeting of nontransfor- Early gene targeting experiments were carried out in med, proliferating primary somatic cells therefore seems transformed somatic cell lines such as mouse L cells, to be a feasible strategy to introduce pre-defined modifi- hamster CHO cells and human bladder carcinoma EJ cations in the somatic genome. The next logical step cells. Targeting constructs often involved nonisogenic towards gene therapy by gene targeting would be the ex DNA and/or quite small regions of homology and tar- vivo correction of mutations in nontransformed somatic geting efficiencies were generally low. The great majority cells, if possible with stem cell properties, isolated from of gene targeting experiments, however, has been done in the appropriate tissue of patients with single gene mouse ES cells, making use of long regions of homology diseases. and/or isogenic DNA, and efficiencies have been rela- tively high. There is consequently a widespread impression that ES cells are more proficient at gene tar- Recent developments in gene targeting geting than primary or immortalised somatic cells. This methods may be so but there is no clear evidence to suggest it. Few studies have compared gene targeting frequencies in ES cells with those in other mouse cell lineages using the Delivery methods same targeting constructs. Two such studies suggested that F9 embryonic carcinoma cells25 and two pre-B cell Nonviral vectors: We have already mentioned that most lines (BASC6-C2 and 18–8tK−)26 supported gene targeting gene targeting experiments involve electroporation, a as efficiently as ES cells. In a third study,27 primary myo- convenient method with less tendency to result in mul- blasts gave an absolute targeting frequency higher than tiple integrations than calcium phosphate transfection. ES cells targeted with the same construct, but this Many alternative delivery methods exist (eg microinjec- reflected the greater transfection efficiency of the myo- tion, optoporation, polybrene, DMSO, DEAE-dextran, blasts, masking a lower GT:NHR ratio. The latter study many liposome formulations, viral vectors, gene gun, was also notable for showing that primary somatic cells polyamidoamine dendrimers, synthetic peptides and can support normal absolute frequencies of gene tar- combinations of some of these and other reagents) but geting with no measurable deterioration of morphology, few have been used for gene targeting and, to our knowl- karyotype or growth control. edge, there are few systematic comparisons of transfec- The only known case of a higher eukaryote cell line tion methods for gene targeting efficiency. with reproducibly high gene targeting efficiency is the High gene targeting frequencies (1–6.7 × 10−3) have chicken B cell line DT40.28 In the chicken, immunoglob- been obtained by microinjection of DNA constructs ulin light chain variation is generated by a peculiar mech- directly into the cell nucleus, but the GT:NHR ratios are anism, in which 25 pseudo-V genes can undergo gene still quite low (1:30–1:125).21,34 Microinjection also conversion with the only functional V locus. This process requires expertise and expensive equipment and the is maintained in some cell lines derived from chicken number of cells that can be conveniently microinjected is tumoral B cells, particularly the DT40 cell line.29 The limiting. Two reports indicate that, for targeting of the immunoglobulin light chain locus and other unrelated APRT35 and HPRT36 genes in CHO cells, electroporation loci can be efficiently modified by gene targeting in this can give higher absolute targeting frequencies than cell line with homologous recombinants accounting for calcium phosphate precipitation, even though the 6.5–100% of stably transfected clones.28,30 Absolute tar- two methods gave similar nontargeted transfection geting frequencies in DT40 cells do not appear to be efficiencies. unusually high (3.1–8.4 × 10−6)28 suggesting that NHR We have compared the efficiency of NHR and gene tar- might be inefficient. The DT40 cell line is currently being geting in the human fibrosarcoma cell line HT1080 using used to study recombination.17,29 DT40 cells can also electroporation and lipofectamine transfection. We used serve as hosts for the efficient targeted modification of an isogenic, replacement-type construct that can disrupt human chromosomes before their transfer by microcell the HPRT locus by inserting a hygromycin expression fusion back to mammalian cells for functional analysis.31 cassette in exon II. After optimising stable transfection Gene targeting has also been achieved in human cell frequencies to similar values by both methods, our pre- lines (Table 1) and, more importantly, in several types of liminary results indicate that gene targeting is 10-fold primary human cells (Table 2). Fibroblastoid, lymphoid, more efficient by electroporation than by lipofectamine ´ ˜ hepatic, epitheloid, myeloid, bladder and colon cell lines transfection (Yanez and Porter, unpublished). Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 152 Table 1 Gene targeting in human cell lines

Cell line (origin) Target gene Strategya Gene Ratio Ref. targeting GT:NHR frequencyb

EJ (bladder carcinoma) ␤-globin NE/I/4.6/No/CP 2.2 × 10−7 1:820, 1:2520 40 ⌬neo PS/R/5.7/Yes/CP NR 1:78 111 JM (T cell leukemia) CD4 PT+PA/I/3.5/No/EP 1.3 × 10−7 1:900 112 HT1080 (ﬁbrosarcoma) 6-16 PT+PA/I/3.5/No/EP 1.9 × 10−7 1:950 113 6-16 PT+PA/I/2/No/EP 3 × 10−7 1:1500 114 6-16 PT/R/6/No/EP у6.3 × 10−8 1:1700 44 − ´ ˜ 6-16 PT/R/6.2/Yes/EP 4–9.5 × 10 7 ND Yanez and Porter, unpublished CDC2 PS/I/10.5/Yes/EP 2 × 10−6 1:50 115 CDC2 PS/R/9/Yes/EP 7.6 × 10−7 1:230 115 HPRT PS+DS/R/6.9/No/EP NR NR 5 − ´ ˜ HPRT PS+DS/R/8.5/Yes/EP 3.1 × 10 6– 1:40–1:105 Yanez and Porter, 2.7 × 10−5 unpublished HeLa (cervical 6-16 PT+PA/I/3.5/No/EP 1.2 × 10−8 1:6550 113 epitheloid carcinoma) 6-16 PT/I-R/1-6/No/EP Ͻ10−8 Ͻ1:5000 44 HepG2 (hepatoma) apolipoprotein B PT+PA/R/8.9/No/EP р5 × 10−8 NR 116 CEM (T cell leukemia) CD43 PS+PA/R/2.7/No/EP р10−7 NR 117 PLB-985 (myeloid X-CGD PS+PA/R/4/No/EP NR 1:241 118 leukemia) DLD-1 (colon cancer) Ki-RAS PT/R/6.7/No/TR 7 × 10−7 NR 119 HCT116 (colon cancer) Ki-RAS PT/R/6.7/No/TR 4 × 10−7 NR 119 p21WAF1/CIP1 PT/R/8.6/No/LP NR NR 120 HT29 (colon carcinoma) CFTR PNS/R/4.3/No/EP NR NR 121

aGene targeting strategy: enrichment procedure (NE: no enrichment; PS: positive selection; PT: promoter-trap; PA: phenotypic analysis; DS: direct selection; PNS: positive-negative selection)/vector type (I: insertion-type; R: replacement-type)/length of sequence homology to target locus incorporated into targeting construct (kbp)/isogenicity of targeting construct to target locus (Yes or No)/transfection method (CP: calcium phosphate: EP: electroporation; TR: transfectam; LP: lipofectamine). bGene targeting frequency: targeted clones per transfected cell. NR: not reported; ND: not determined.

Table 2 Gene targeting in primary human cells

Cell type (origin) Target gene Strategya Gene Ratio Ref. targeting GT:NHR frequencyb

␤ + × −8 Retinal pigmented 2 microglobulin PS PA/R/11.1/No/EP 6.7 10 –NR32 epithelial cells 4.3 × 10−6 (cadaver eye tissue) ␤ + × −7 Keratinocytes 2 microglobulin PS PA/R/7.4–11.1/No/EP 1.8–8.8 10 NR 32 (neonatal foreskin) Fibroblasts p21WAF1/CIP1 PT/R/7.8–8.6/No/EP 6 × 10−8 NR 33 (embryonic lung)

a,bAs in Table 1. NR, not reported.

Viral vectors: Viral vectors have not been widely used for infected cell. In all cases, however, targeting was associa- gene targeting because of restrictions in the size and ted with gene conversion into regions of nonhomology. design of the homologous DNA, potential interference by More recent studies involving adenoviral vectors have virally encoded proteins (eg retroviral integrase) and been more encouraging. Adenoviral vectors have the because simpler delivery methods have been adequate advantage of delivering their double-stranded DNA gen- for gene knockout studies. The high transfection fre- ome to the nucleus with high efficiencies (approaching quencies that can be achieved with viral vectors, how- 100%) and in a form that is naturally episomal (ie inef- ever, make them attractive as potential vectors for gene ficient at random integration). The ability of adenovirus targeting, particularly for in vivo applications. In an initial to infect nondividing cells is an additional attraction study,37 integration-deficient retroviral vectors were used given potential difficulties in stimulating division of to correct mutant neo genes previously integrated in chro- somatic stem cells. Targeting of the FGR locus with iso- mosomes of Rat-2 and human K421b cells. Stable correc- genic DNA in mouse ES cells using a replication-incom- tion was detected at frequencies close to 3 × 10−7 per petent adenoviral vector produced ratios of GT:NHR as Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 153 high as 1:2.5.38 The same targeting cassette gave a ratio may depend on cell type since in CHO cells gene tar- of 1:20 when transfected by electroporation. The absolute geting was stimulated by three to four orders of magni- targeting frequencies in the adenovirus experiments were tude48 while no stimulation was observed in mouse Ltk− around 10−5, in the range described for electroporation. cells.49 It is worth noting that target gene-specific DSBs An earlier report described the use of replication-com- may be mutagenic, as well as having a stimulatory effect petent adenovirus vectors for gene targeting of the APRT on HR. Thus, in an assay for I–SceI-induced loss of APRT locus in CHO cells.39 In this case the GT:NHR ratio gene function in CHO cells, I–SceI stimulated HR about ranged from 1:5 to 1:14, a 400-fold improvement over that 100-fold but NHR was stimulated more than 1000-fold, reported upon calcium phosphate transfection of a plas- and point mutations about 10-fold.50 For a practical appli- mid targeting construct. Again, the absolute targeting fre- cation of target-specific cleavage in single-step targeting quency was low, around 6 × 10−7, similar to that observed strategies, an alternative to engineered I–SceI recognition in plasmid transfections. These reports therefore show sites must be found. If oligonucleotides able to form triple encouraging ratios of GT:NHR, similar to those found DNA helices upon binding to their polypurine target routinely in chicken DT40 cells, even though no increase site51 could be linked to DSB-inducing agents, cleavage in the absolute targeting frequency has been achieved so of specific sequences in the target locus might be feasible. far. The improved ratios of GT:NHR may or may not be offset by the additional work of constructing a targeting Design of targeting constructs virus, but if absolute frequencies can be improved, adenoviral vectors could become the vectors of choice Optimising promoter-trap constructs: Promoter-trap- for gene therapy by gene targeting. based targeting involves DNA constructs containing promoterless selectable marker genes that become activated Manipulating target gene status by the target gene promoter following HR. These strategies can make use of the varying efficiency of different Transcription of target gene: Targeted modification of the selectable markers to generate resistance to their corre- ␤-globin gene in human EJ cells40 and of the adipsin and sponding selective drug. Several positively selectable adipocyte P2 genes in mouse ES cells41 established that markers have shown up to 200-fold differences in their transcription of the target gene is not required for gene ability to generate resistant colonies in rat-1 cells when targeting. It has been reported, however, that transcrip- transfected as promoterless constructs.52 Using a com- tion stimulates both extrachromosomal HR in CHO bined promoter-trap and positive-negative selection sys- cells42 and gene targeting in HT1080 cells.43 Both stra- tem, it has been shown that some markers may be too tegies involved recombination between two defective neo weak to render targeted, drug-resistant clones when the genes and made use of the dexamethasone-inducible level of expression of the target gene is low. Matching MMTV promoter to drive transcription of both substrates the expression level of the target gene (c-MYC), the (extrachromosomal HR) or of the target gene (gene strength of the selectable marker in the targeting con- targeting). HR was scored as the frequency of neo+ colon- struct and the selection stringency (drug concentration) ies, and stimulation upon transcription ranged from 3- to has led to Ͼ60-fold improvements in GT:NHR ratio. This Ͼ20-fold. In contrast, no report is available on the effect was due to selective loss of nonhomologous recombi- of transcription on gene targeting of endogenous mam- nants, with absolute targeting frequencies remaining malian genes. To address this question we have targeted unchanged.52 the IFN-inducible 6–16 gene in HT1080 cells. The promoter-trap targeting construct was used as previously DNA end-modifications: Modification of DNA ends in the described44 except that it was reconstructed using iso- transfected molecules might also be a way to reduce genic (HT1080) DNA. Transcription from the 6–16 pro- NHR and even to improve HR. Dideoxynucleotides moter was induced by IFN treatment and controls were added to the 3Ј ends of a linear plasmid transfected into performed to account for any nonspecific effects of IFN. monkey COS-1 cells increased the intramolecular We could detect no variation in targeting frequency that HR:end-joining (NHR) ratio about five-fold.53 To our correlated with the transcriptional status of the 6–16 gene knowledge, use of this treatment in targeting experiments ´ ˜ (Yanez and Porter, unpublished). Thus, the increase in has not been reported so far. A preliminary report does targeting frequencies observed with exogenous target state, however, that ligation of linear targeting constructs genes may not be a general phenomenon. to self-complementary oligonucleotides, to seal the DNA ends, enhances stability of the DNA in transfected cells Cleavage of target locus: It has been recently shown that and increases the frequency of gene targeting.54 introduction of a double-stranded break (DSB) in the genomic DNA of the target locus can increase targeting fre- Use of oligonucleotides and small DNA fragments: Two quencies. The strategies used to show this were similar recent reports describe dramatically improved gene in several cell lines. Defective marker genes were engin- targeting frequencies using either chimeric DNA–RNA eered in vitro by insertion of the recognition site for the oligonucleotides55 or small denatured DNA fragments.56 S. cerevisiae DNA endonuclease I–SceI. Stable transfec- The possibility of using small nucleic acids for gene tar- tants of the marker are sensitive to the selective drug, but geting is attractive. Sufficiently small molecules can be can become resistant after HR with a construct containing synthesised on a large scale, if necessary with chemical wild-type marker sequences. Cleavage at the I–SceI site, modifications to improve stability, and without the worry induced by expressing the I–SceI gene in vivo or electro- of biological contamination. Unfortunately, the logarith- porating purified enzyme, stimulate gene targeting 50- mic relationship between targeting efficiency and length fold to 104-fold variously in mouse ES, 3T3 and PCC-7 of homology suggests that simple nucleic acids with cells, and hamster CHO cells.45–48 The extent of the effect small regions of homology will not be efficient. Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 154 The stimulus for the experiments of the Kmiec labora- efficiency. Thus targeting frequencies in mouse ES cells tory was their observation that RNA–DNA hybrids were were found to keep an inverse linear correlation with cell more active in homologous pairing reactions than corre- doubling times, and differences of up to 14-fold were sponding DNA duplexes.57 Chimeric, self-complementary obtained by varying the amount of foetal calf serum.61 DNA–RNA oligonucleotides were therefore designed for Also, HR between co-injected plasmids peaks in early- correction of the single nucleotide mutation in the chromo- to mid-S phase in rat-20 cells62 suggesting that cell cycle somal ␤-globin gene of immortalised B cells from a patient synchronisation might be a way to improve gene tar- with sickle-cell anaemia.55 In these oligonucleotides, self- geting efficiencies. folding creates hairpin ends flanking 25 bp of homology In a recent report of gene targeting at the HPRT locus to the target locus. The noncoding strand in the oligonu- in ES cells, absolute targeting efficiencies were improved cleotide contains two 10-nt long 2Ј-O-methyl RNA from 0.001 to 20% by the use of an optimised gene tar- stretches flanking a 5 bp DNA sequence, with the wild- geting protocol. A key feature of this protocol appears type version of the mutated nucleotide in the central pos- to have been the plating of cells after electroporation at ition. The hairpin ends and modified ribonucleotides may unusually high densities and delaying selection for at protect the oligonucleotides from degradation. Oligonu- least 60 h.63 This dramatic improvement would be sig- cleotides were transfected using DOTAP liposomes. nificant for gene therapy if it can be reproduced at other Transfections were allowed to proceed for 6 h, after which loci and in other cell lines. the entire, mixed population of cells was harvested and assayed for corrected globin alleles by Southern analysis Poly(ADP ribose) polymerase (PARP) inhibition: PARP is and PCR. The intensities of the corrected and uncorrected an abundant nuclear enzyme activated by binding to a bands in the Southern analysis were similar, suggesting particular type of DNA lesion, single-strand breaks. that approximately 50% of the target alleles had been cor- PARP itself appears to be the main target for ADP-ribo- rected.55 This remarkable result clearly has huge impli- sylation, and this modification seems to precede DNA cations for the feasibility of gene therapy by gene tar- repair.64,65 It has been reported that inhibition of PARP geting. However, as has already been pointed out,58,59 reduces both random genomic integration of exogenous further controls are required to rule out the possibilities constructs66,67 and the integration of the retroviral gen- of screening artefacts and cell contamination. An unam- ome in the host cell DNA.68 This reduction in NHR led biguous demonstration of targeted correction would to the reasoning that 3-methoxybenzamide (3-MB)- involve the isolation of a clonal cell population bearing inhibition of PARP might be a way to enrich for targeted both a corrected ␤-globin allele (as judged by Southern clones in gene targeting experiments.36 The hypothesis analysis), and an otherwise patient-specific genotype (as was tested by targeting a hemizygous 3-bp deletion judged eg by DNA-fingerprint analyses). mutant of the mouse APRT gene in CHO cells. Treatment Small DNA fragments have been used to target the 3- of the cells with 3-MB led to a reduction in the frequency bp deletion (⌬F 508) of the cystic fibrosis transmembrane of both gene targeting and NHR, the effect being more regulator (CFTR) gene in transformed lung epithelial cells pronounced for the latter. Thus, the authors report an from a patient with cystic fibrosis.56 The targeting con- enrichment for targeted clones of up to 100-fold, even struct was a denatured 491 bp PCR product derived from though the absolute targeting frequency was reduced a functional CFTR gene. This DNA fragment was com- two- to 13-fold. Curiously, the enrichment was depen- bined with purified E. coli recA protein (see below) in dent on the DNA delivery method, being observed with some experiments before transfection by electroporation, calcium phosphate transfection but not with electropor- polyamidoamine dendrimers or gramicidin S-liposome ation. An undesirable consequence of using calcium complexes. Screening for CFTR correction was by allele- phosphate was the high frequency of concomitant ran- specific, PCR-based Southern blots and functional Cl− dom integration events (50–64%) observed in targeted conductance analyses, but genomic Southern blots of colonies.36 clonal cell populations were not carried out. The reason for the apparently high targeting efficiency, estimated as MSH2 gene inactivation: Non-isogenic sequences impair close to 10−2, remains unclear. The different transfection intra- and inter-chromosomal HR,69,70 and this has been procedures gave similar targeting frequencies, suggesting interpreted as a defensive mechanism against potentially that the high frequencies were not due to the choice of deleterious ectopic DNA recombination between transfection method or the use of recA protein. The use diverged sequences. Isogenic DNA is also more efficient of single-stranded vectors for gene targeting has been in gene targeting15,16 and this has recently been linked reported before, and did not seem to produce an to the action of DNA mismatch repair mechanisms. The improvement.43,60 It has also been reported that targeting mammalian gene MSH2 is a homologue of bacterial with approximately 700-bp-long targeting constructs in MutS, whose protein product binds to base mispairs and mouse ES cells47 and hamster CHO cells48 is undetectable loops of up to four unpaired nucleotides, triggering the (Ͻ6.3 × 10−7) unless the genomic DNA is linearised at the mutSL pathway of DNA mismatch repair in E. coli. The target locus. Thus, as one would predict from earlier human hMSH2 gene is defective in hereditary non- studies,15 high targeting frequencies are not generally polyposis colorectal cancer.71 The mouse homologue of obtained when small targeting constructs are used. MSH2 has been disrupted in ES cells, and extracts from homozygous MSH2−/− cells are defective in DNA mis- Treatment of host cells match-binding proteins in vitro. Gene targeting of the reti- noblastoma locus was found to occur as efficiently with a Growth and selection conditions: Careful optimisation of non-isogenic targeting construct as with an isogenic one cell growth conditions both before and after transfection in the MSH2−/− ES cells.72 If inhibitors of mismatch may have a drastic effect on measurable targeting repair become available, they could be used to block the Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 155 process temporarily in cells undergoing HR with non-iso- three genes.84–86 The RAD5184 and RAD5486 genes of genic constructs. Alternatively, proteins such as E. coli higher eukaryotes are expressed at particularly high lev- recA,73 RuvAB74 or the gene 4 helicase of bacteriophage els in spleen, thymus and testis, and at low levels in all T7,75 which allow HR to proceed through mismatches other tissues tested. This might also be the case for and nonhomologous patches, might usefully be incorpor- RAD52.85,87 Recent results suggest that these genes play ated into a targeting strategy. This could be particularly a role in proliferating cells, which are present at high lev- important in gene therapy procedures relying on gene els in immune tissues and gonads. However, the gonads targeting, when the use of a standard, non-isogenic con- support meiotic DNA recombination, and the thymus is struct would be most convenient. the tissue where V(D)J recombination (a site-specific type of NHR) of T cell receptor genes takes place. The possible DNA damage: Several types of DNA damage have been involvement of RAD51, RAD52 and RAD54 in these shown to increase HR in eukaryotic cells. These include processes is also under investigation. lesions caused by chemical carcinogens,76 UV The human RAD51 protein is a DNA-dependent irradiation,77 gamma irradiation78 and photoreactive mol- ATPase able to promote homologous pairing and strand ecules.79 Another type of DNA damage is the presence exchange in vitro, albeit less efficiently than its E. coli of uracil residues substituting for thymine, and this counterpart, recA.88 The disruption of the mouse homo- seems to be recombinogenic in some phage and bacterial logue of RAD51 produces an early embryonic lethal systems.80,81 We have used uracil-containing targeting phenotype. Additionally, no RAD51−/− ES or F9 cells constructs, prepared from dut ung E. coli strains, in our could be produced, and no cells could be grown from the previously mentioned HPRT and 6–16 targeting systems early embryos.89,90 A conditional RAD51−/− cell line has in HT1080 cells. The presence of uracil was detrimental been obtained from chicken DT40 cells upon expression to both gene targeting and NHR, the effect being more of the human RAD51 gene as a repressible transgene and ´ ˜ marked for the former (Yanez and Porter, unpublished). sequential targeting of the endogenous RAD51 loci. If the The mutagenic agents mentioned above essentially human transgene is repressed, RAD51−/− cells accumu- cause random DNA damage, which is not a desirable late chromosomal aberrations and arrest in the G2/M outcome in gene targeting procedures. Target DNA- phases of the cell cycle, and this is followed by DNA specific techniques are potentially more attractive. In this degradation and massive cell death. The authors con- respect, it has recently been shown that triplex-forming clude that DSBs generated during DNA replication are oligonucleotides bound to photoreactive molecules can repaired by RAD51-mediated recombinational repair, cause sequence-specific DNA damage around the target this being an essential role in higher eukaryotes.91 The site for the oligonucleotide, and this can induce HR.82 key importance of RAD51 is underscored by its interac- This could be a way to direct HR activities to the desired tion with three tumour suppressors: p53,92 and breast DNA region in gene targeting experiments. cancer BRCA193 and BRCA2.94,95 The biochemical information available for RAD52 indi- Recent advances in DNA recombination cates that the protein can promote annealing of homologous single-stranded DNA molecules, and strand transfer The past few years have brought considerable advances between homologous single-stranded and double- in our understanding of the mechanisms of recombi- stranded DNA substrates.96–98 Mouse RAD52−/− ES cells nation in eukaryotes and their roles in DSB repair. In show a mild reduction in the frequencies of gene tar- yeast, where transfected DNA integrates mostly by HR, geting, but are not hypersensitive to ionising radiation the predominant pathway for DSB repair involves HR. (DSBs). RAD52 knockout mice are viable, fertile and have Although an analogous pathway exists in higher eukary- no apparent defect in immunoglobulin gene rearrange- otes, it appears to have a more specialised role, and DSB ment (T Rijkers and A Pastink, personal communication). repair occurs preferentially by end-joining (NHR). The RAD52−/− chicken DT40 cells show decreased fre- key players in these two pathways have now been ident- quencies of gene targeting, but wild-type levels of gene ified: the RAD52 group of proteins for HR and a DNA- conversion (Y Yamaguchi-Iwai and S Takeda, personal dependent protein kinase and associated proteins for communication). This may indicate that RAD52 function NHR. If we could understand the way cells control the is required to different degrees in various HR processes. balance between these two pathways, we might be able Expression of the S. cerevisiae RAD52 gene in human to devise methods to reverse the bias towards NHR in HT1080 cells99 and of the human gene in monkey FSH2 gene targeting procedures. cells100 stimulates extrachromosomal and intrachromosomal HR, respectively, and confers increased resistance to HR proteins in higher eukaryotes ionising radiation. These results suggest that the cellular Genetic studies in bacteria and yeast have uncovered machinery for HR can be stimulated by over-expression some key players in HR pathways, and the search for of some of the relevant genes. homologous genes in higher eukaryotes has proved to be The RAD54 protein is a putative helicase of the extremely rewarding. In E. coli, recA mutants profoundly SNF2/SWI2 family.86 The mouse RAD54 gene knockout affect HR, inhibiting the process up to 104-fold. recA pro- has shown that this gene is dispensable for V(D)J recom- tein is a single-stranded DNA-dependent ATPase that bination and meiosis, RAD54−/− mice having normal B catalyses homologous pairing and strand exchange and T cell populations and producing offspring. How- between the DNA substrates in HR.73 In S. cerevisiae,HR ever, RAD54−/− mouse ES cells are hypersensitive to is mediated by the genes of the RAD52 epistasis group, DSB-causing agents, and exhibit reduced gene targeting of which RAD51 (the yeast homologue of recA), RAD52 frequencies compared with wild-type cells.101 RAD54−/− and RAD54 are particularly important.83 Humans and mutants of the chicken B cell line DT40 are also hypersen- other higher eukaryotes encode homologues of these sitive to ionising radiation, and have seven-fold reduced Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 156 gene conversion frequencies. In contrast, gene targeting of new data may lead to more efficient ways of DNA frequencies are reduced by at least two orders of magni- delivery for gene targeting, possible targets for inhibitors tude in RAD54−/− chicken cells, suggesting a high of undesirable NHR, and proteins that could be manipu- dependence on RAD54.17 These reports17,101 have pro- lated in targeting strategies. vided genetic evidence of the relationship between DSB Many old and new questions remain unanswered. We repair, HR and gene targeting in higher eukaryotes. do not understand why different loci appear to target at The first attempts to find practical applications for different frequencies, and whether this may be related to homologous recombination proteins in gene targeting epigenetic factors such as DNA methylation or chromatin procedures are already ongoing. Targeting constructs condensation. It is unknown how the proteins of HR are have been pre-incubated with purified E. coli recA pro- recruited to the DSB, and little is known about the tein before transfection by electroporation, polyamido- decision making that drives DSB repair to NHR or HR. amine dendrimers and gramicidin S–liposome com- In vivo gene therapy by gene targeting is not viable at plexes, and this treatment did not seem to increase present. However, several types of human tissue-specific targeting frequencies.56 Experiments in which the tar- stem cells can be isolated and subjected to ex vivo ampli- geting construct is pre-incubated with purified human fication and manipulation. This opens the possibility of RAD51 protein before transfection are reported to be selecting and expanding cells that have undergone gene underway.102 targeting. Nontransformed, proliferating human cells have already been modified ex vivo by gene targeting. The NHR in higher eukaryotes next step towards gene therapy by gene targeting in The analysis of several sets of mammalian cell lines human cells could be the actual correction of the DNA hypersensitive to ionising radiation has revealed four defect in nontransformed cells from patients with single- complementation groups of defects in DSB repair. Three gene disorders. of these have been mapped to the genes encoding the three subunits of DNA-dependent protein kinase (DNA- PK): Ku70, Ku80 and the catalytic subunit DNA-PKcs.103 Acknowledgements The fourth group corresponds to the XRCC-4 gene.104 Additional data on unpublished results by the authors V(D)J recombination is impaired in Ku, DNA-PKcs or are available upon request. We thank J-M Buerstedde for XRCC4-deficient cell lines, providing a genetic link stimulating discussions and we are grateful to T Rijkers, between that process and DSB repair. The Ku70-Ku80 A Pastink, E Sonoda, Y Yamaguchi-Iwai and S Takeda for dimer is a DNA-binding protein specific for DNA ends. allowing us to cite their unpublished observations and Upon DNA end-binding, Ku recruits DNA-PKcs to the providing preprints. We thank A Pastink, S Takeda, SP site of DNA damage, and this enzyme becomes acti- Jackson, SC West, ML Caparros, I Correa and J Itzhaki vated.103 On the other hand, XRCC-4 binds to and stimu- for critical reading of the manuscript. lates DNA ligase IV, which might be involved in the final steps of DNA end-joining.105,106 Using I–SceI-induced chromosomal DSBs, it has been shown that Ku80- References deficient CHO cells are impaired in end-joining, while they are proficient in gene targeting and DNA repair by 1 Anderson WF. Prospects for human gene therapy. Science 1984; HR.48 Additionally, Ku80-deficient CHO cells have 226: 401–409. reduced stable transfection frequencies under some con- 2 Capecchi MR. Altering the genome by homologous recombi- ditions.48,107 These data show that NHR can be specifi- nation. Science 1989; 244: 1288–1292. 3 Vega MA. Gene targeting in human gene therapy. In: Vega MA cally impaired in mammalian cells without affecting HR. 108 (ed). Gene Targeting. CRC Press: Boca Raton, 1994, pp 211–229. It has been recently shown that DNA-PKcs and the 4 Verma IM, Somia N. Gene therapy: promises, problems and 109 ataxia telangectasia gene product ATM interact consti- prospects. Nature 1997; 389: 239–242. tutively with the protein tyrosine kinase c-Abl. Func- 5 Zheng H et al. Fidelity of targeted recombination in human tional ATM is required for ionising radiation-induced fibroblasts and murine embryonic stem cells. Proc Natl Acad Sci activation of c-Abl, and DNA damage promotes a physi- USA 1991; 88: 8067–8071. cal association of c-Abl and p53.110 Thus, p53 seems to 6 Bertling WM. Gene targeting. In: Vega MA (ed). Gene Targeting. interact with both NHR and HR pathways, and it has CRC Press: Boca Raton, 1994, pp 1–44. been hypothesised that it might control the way DNA 7 Porter ACG. Designer genomes. Technique 1989; 1: 53–65. 8 Emerson SG. Ex vivo expansion of hematopoietic precursors, repair of DSBs is performed.92 progenitors, and stem cells: the next generation of cellular therapeutics. Blood 1996; 87: 3082–3088. Final remarks 9 Watt FM. Epidermal stem cells in culture. J Cell Sci Suppl 1988; 10: 85–94. The past few years have provided substantial advances 10 Porter A. Gene targeting as an approach to gene therapy. In: in our understanding of several aspects of gene targeting Lemoine NR, Cooper DN (eds). Gene Therapy. BIOS Scientific in cells of higher eukaryotes. Human cells can be modi- Publishers: Oxford, 1996, pp 169–190. fied by gene targeting at frequencies similar to those 11 Doi S, Campbell C, Kucherlapati R. Directed modification of described for other vertebrates. The ratio of GT:NHR can genes by homologous recombination in mammalian cells. In: be improved by using adenoviral vectors for DNA deliv- Grosveld F, Kollias G (eds). Transgenic Animals. Academic Press: London, 1992, pp 27–46. ery, and the absolute targeting frequency can be greatly 12 Hasty P, Bradley A. Gene targeting vectors for mammalian increased by induced DSBs in the target locus. The cells. In: Joyner AL (ed). Gene targeting. A Practical Approach. expected relationship between gene targeting and DNA IRL: Oxford, 1993, pp 1–31. repair by HR has been confirmed, and we now know a 13 Melton DW. Gene targeting in the mouse. BioEssays 1994; 16: handful of proteins involved in the process. This wealth 633–638. Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 157 14 Porter ACG, Dallman MJ. Gene targeting: techniques and targeting in mammalian cells by inhibition of poly(ADP- applications to transplantation. Transplantation 1997 64: 1227– ribosylation). Biochim Biophys Acta 1996; 1308: 241–250. 1235. 37 Ellis J, Bernstein A. Gene targeting with retroviral vectors: 15 Deng C, Capecchi MR. Reexamination of gene targeting fre- recombination by gene conversion into regions of nonhom- quency as a function of the extent of homology between the ology. Mol Cell Biol 1989; 9: 1621–1627. targeting vector and the target locus. Mol Cell Biol 1992; 12: 38 Mitani K et al. Gene targeting in mouse embryonic stem cells 3365–3371. with an adenoviral vector. Somat Cell Mol Genet 1995; 21: 16 te Riele H, Maandag ER, Berns A. Highly efficient gene tar- 221–231. geting in embryonic stem cells through homologous recombi- 39 Wang Q, Taylor MW. Correction of a deletion mutant by gene nation with isogenic DNA constructs. Proc Natl Acad Sci USA targeting with an adenovirus vector. Mol Cell Biol 1993; 13: 1992; 89: 5128–5132. 918–927. 17 Bezzubova O et al. Reduced X-ray resistance and homologous 40 Smithies O et al. Insertion of DNA sequences into the human recombination frequencies in a RAD54−/− mutant of the chromosomal ␤-globin locus by homologous recombination. chicken DT40 cell line. Cell 1997; 89: 185–193. Nature 1985; 317: 230–234. 18 Randolph DA et al. PCR-based gene targeting of the inducible 41 Johnson RS et al. Targeting of nonexpressed genes in embry- nitric oxide synthase (NOS2) locus in murine ES cells, a new onic stem cells via homologous recombination. Science 1989; and more cost-effective approach. Transgenic Res 1996; 5: 245: 1234–1236. 413–420. 42 Nickoloff JA, Reynolds RJ. Transcription stimulates homolo- 19 Smithies O, Koralewski MA, Song KY, Kucherlapati RS. Hom- gous recombination in mammalian cells. Mol Cell Biol 1990; 10: ologous recombination with DNA introduced into mammalian 4837–4845. cells. Cold Spring Harbor Symp Quant Biol 1984; 49: 161–170. 43 Thyagarajan B, Johnson BL, Campbell C. The effect of target 20 Jasin M, de Villiers J, Weber F, Schaffner W. High frequency site transcription on gene targeting in human cells in vitro. of homologous recombination in mammalian cells between Nucleic Acids Res 1995; 23: 2784–2790. endogenous and introduced SV40 genomes. Cell 1985; 43: 44 Porter AC, Itzhaki JE. Gene targeting in human somatic cells. 695–703. Complete inactivation of an interferon-inducible gene. Eur J 21 Thomas KR, Folger KR, Capecchi MR. High frequency tar- Biochem 1993; 218: 273–281. geting of genes to specific sites in the mammalian genome. Cell 45 Rouet P, Smih F, Jasin M. Introduction of double-strand breaks 1986; 44: 419–428. into the genome of mouse cells by expression of a rare-cutting 22 Graves KH, Moreadith RW. Derivation and characterization of endonuclease. Mol Cell Biol 1994; 14: 8096–8106. putative pluripotential embryonic stem cells from preimplan- 46 Choulika A, Perrin A, Dujon B, Nicolas JF. Induction of hom- tation rabbit embryos. Mol Reprod Dev 1993; 36: 424–433. ologous recombination in mammalian chromosomes by using 23 Thomson JA et al. Isolation of a primate embryonic stem cell the I–SceI system of Saccharomyces cerevisiae. Mol Cell Biol 1995; line. Proc Natl Acad Sci USA 1995; 92: 7844–7848. 15: 1968–1973. 24 Latchman DS. Germline gene therapy? Gene Therapy 1994; 1: 47 Smih F, Rouet P, Romanienko PJ, Jasin M. Double-strand 277–279. breaks at the target locus stimulate gene targeting in embryonic 25 Coll JL et al. Targeted disruption of vinculin genes in F9 and stem cells. Nucleic Acids Res 1995; 23: 5012–5019. embryonic stem cells changes cell morphology, adhesion, and 48 Liang F et al. Chromosomal double-strand break repair in locomotion. Proc Natl Acad Sci USA 1995; 92: 9161–9165. Ku80-deficient cells. Proc Natl Acad Sci USA 1996; 93: 8929– 26 Charron J et al. High-frequency disruption of the N-MYC gene 8933. in embryonic stem and pre-B cell lines by homologous recom- 49 Lukacsovich T, Yang D, Waldman AS. Repair of a specific bination. Mol Cell Biol 1990; 10: 1799–1804. double-strand break generated within a mammalian chromo- ´ some by yeast endonuclease I–SceI. Nucleic Acids Res 1994; 22: 27 Arbones ML, Austin HA, Capon DJ, Greenburg G. Gene tar- 5649–5657. geting in normal somatic cells: inactivation of the interferon- 50 Sargent RG, Brenneman MA, Wilson JH. Repair of site-specific gamma receptor in myoblasts. Nat Genet 1994; 6: 90–97. double-strand breaks in a mammalian chromosome by hom- 28 Buerstedde JM, Takeda S. Increased ratio of targeted to random ologous and illegitimate recombination. Mol Cell Biol 1997; 17: integration after transfection of chicken B cell lines. Cell 1991; 267–277. 67: 179–188. 51 Sun JS, Garestier T, Helene C. Oligonucleotide directed triple 29 Bezzubova OY, Buerstedde JM. Gene conversion in the chicken helix formation. Curr Opin Struct Biol 1996; 6: 327–333. immunoglobulin locus: a paradigm of homologous recombi- 52 Hanson KD, Sedivy JM. Analysis of biological selections for nation in higher eukaryotes. Experientia 1994; 50: 270–276. high-efficiency gene targeting. Mol Cell Biol 1995; 15: 45–51. 30 Takata M et al. Tyrosine kinases Lyn and Syk regulate B cell 53 Chang XB, Wilson JH. Modification of DNA ends can decrease receptor-coupled Ca2+ mobilization through distinct pathways. end-joining relative to homologous recombination in mam- EMBO J 1994; 13: 1341–1349. malian cells. Proc Natl Acad Sci USA 1987; 84: 4959–4963. 31 Dieken ES et al. Efficient modification of human chromosomal 54 Nehls M et al. Two genetically separable steps in the differen- alleles using recombination-proficient chicken/human tiation of thymic epithelium. Science 1996; 272: 886–889. microcell hybrids. Nat Genet 1996; 12: 174–182. 55 Cole-Strauss A et al. Correction of the mutation responsible for 32 Williams SR, Ousley FC, Vitez LJ, DuBridge RB. Rapid detec- sickle cell anemia by an RNA–DNA oligonucleotide. Science tion of homologous recombinants in nontransformed human 1996; 273: 1386–1389. cells. Proc Natl Acad Sci USA 1994; 91: 11943–11947. 56 Kunzelmann K et al. Gene targeting of CFTR DNA in CF epi- 33 Brown JP, Wei W, Sedivy JM. Bypass of senescence after dis- 3 CIP1/WAF1 thelial cells. Gene Therapy 1996; : 859–867. ruption of p21 gene in normal diploid human fibro- 57 Kotani H et al. RNA facilitates RecA-mediated DNA pairing blasts. Science 1997; 277: 831–834. and strand transfer between molecules bearing limited regions 34 Zimmer A, Gruss P. Production of chimaeric mice containing of homology. Mol Gen Genet 1996; 250: 626–634. embryonic stem (ES) cells carrying a homoeobox Hox 1.1 allele 58 Thomas KR, Capecchi MR. Recombinant DNA technique and mutated by homologous recombination. Nature 1989; 338: sickle cell anemia research (letter). Science 1997; 275: 1404. 150–153. 59 Stasiak A, West SC, Egelman EH. Sickle cell anemia research 35 Nairn RS et al. Targeting vector configuration and method of and a recombinant DNA technique (letter). Science 1997; 277: gene transfer influence targeted correction of the APRT gene 460–461. in Chinese hamster ovary cells. Somat Cell Mol Genet 1993; 19: 60 Fujioka K, Aratani Y, Kusano K, Koyama H. Targeted recombi- 363–375. nation with single-stranded DNA vectors in mammalian cells. 36 Waldman BC, O’Quinn JR, Waldman AS. Enrichment for gene Nucleic Acids Res 1993; 21: 407–412. Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 158 61 Udy GB, Parkes BD, Wells DN. ES cell cycle rates affect gene roles of double-strand breaks. Trends Biochem Sci 1995; 20: targeting frequencies. Exp Cell Res 1997; 231: 296–301. 387–391. 62 Wong EA, Capecchi MR. Homologous recombination between 84 Shinohara A et al. Cloning of human, mouse and fission yeast coinjected DNA sequences peaks in early to mid-S phase. Mol recombination genes homologous to RAD51 and recA. Nat Cell Biol 1987; 7: 2294–2295. Genet 1993; 4: 239–243. 63 Templeton NS, Roberts DD, Safer B. Efficient gene targeting in 85 Muris DF et al. Cloning of human and mouse genes homolo- mouse embryonic stem cells. Gene Therapy 1997; 4: 700–709. gous to RAD52, a yeast gene involved in DNA repair and ´ 64 de Murcia G, Menissier de Murcia J. Poly(ADP-ribose) poly- recombination. Mutat Res 1994; 315: 295–305. merase: a molecular nick-sensor. Trends Biochem Sci 1994; 19: 86 Kanaar R et al. Human and mouse homologs of the Saccharo- 172–176. myces cerevisiae RAD54 DNA repair gene: evidence for func- 65 Lindahl T, Satoh MS, Poirier GG, Klungland A. Post-trans- tional conservation. Curr Biol 1996; 6: 828–838. lational modification of poly(ADP-ribose) polymerase induced 87 Shen Z et al. The human and mouse homologs of the yeast by DNA strand breaks. Trends Biochem Sci 1995; 20: 405–411. RAD52 gene: cDNA cloning, sequence analysis, assignment to 66 Farzaneh F et al. ADP-ribosylation is involved in the inte- human chromosome 12p12.2-p13, and mRNA expression in gration of foreign DNA into the mammalian cell genome. mouse tissues. Genomics 1995; 25: 199–206. Nucleic Acids Res 1988; 16: 11319–11326. 88 Baumann P, Benson FE, West SC. Human RAD51 protein pro- 67 Waldman BC, Waldman AS. Illegitimate and homologous motes ATP-dependent homologous pairing and strand transfer recombination in mammalian cells: differential sensitivity to an reactions in vitro. Cell 1996; 87: 757–766. inhibitor of poly(ADP-ribosylation). Nucleic Acids Res 1990; 18: 89 Tsuzuki T et al. Targeted disruption of the RAD51 gene leads 5981–5988. to lethality in embryonic mice. Proc Natl Acad Sci USA 1996; ¨ 93: 6236–6240. 68 Gaken JA et al. Efficient retroviral infection of mammalian cells 90 Lim DS, Hasty P. A mutation in mouse RAD51 results in an is blocked by inhibition of poly(ADP-ribose) polymerase early embryonic lethal that is suppressed by a mutation in p53. activity. J Virol 1996; 70: 3992–4000. Mol Cell Biol 1996; 16: 7133–7143. 69 Waldman AS, Liskay RM. Dependence of intrachromosomal 91 Sonoda E et al. RAD51 deficient vertebrate cells accumulate recombination in mammalian cells on uninterrupted hom- chromosomal breaks prior to cell death. EMBO J 1997 (in ology. Mol Cell Biol 1988; 8: 5350–5357. press). 70 Radman M, Wagner R. Mismatch recognition in chromosomal ¨ 92 Sturzbecher HW et al. p53 is linked directly to homologous interactions and speciation. Chromosoma 1993; 102: 369–373. recombination processes via RAD51/RecA protein interaction. 71 Modrich P. Mismatch repair, genetic stability, and cancer. EMBO J 1996; 15: 1992–2002. Science 1994; 266: 1959–1960. 93 Scully R et al. Association of BRCA1 with RAD51 in mitotic 72 de Wind N et al. Inactivation of the mouse MSH2 gene results and meiotic cells. Cell 1997; 88: 265–275. in mismatch repair deficiency, methylation tolerance, hyper- 94 Sharan SK et al. Embryonic lethality and radiation hypersensi- recombination, and predisposition to cancer. Cell 1995; 82: tivity mediated by RAD51 in mice lacking Brca2. Nature 1997; 321–330. 386: 804–810. 73 Camerini-Otero RD, Hsieh P. Homologous recombination pro- 95 Mizuta R et al. RAB22 and RAB163/mouse BRCA2: proteins teins in prokaryotes and eukaryotes. Annu Rev Genet 1995; 29: that specifically interact with the RAD51 protein. Proc Natl Acad 509–552. Sci USA 1997; 94: 6927–6932. 74 Parsons CA, Stasiak A, West SC. The E. coli RuvAB proteins 96 Ogawa T et al. RecA-like recombination proteins in eukaryotes: branch migrate Holliday junctions through heterologous DNA functions and structures of RAD51 genes. Cold Spring Harbor sequences in a reaction facilitated by SSB. EMBO J 1995; 14: Symp Quant Biol 1993; 58: 567–576. 5736–5744. 97 Mortensen UH, Bendixen C, Sunjevaric I, Rothstein R. DNA 75 Kong D, Griffith JD, Richardson CC. Gene 4 helicase of bac- strand annealing is promoted by the yeast Rad52 protein. Proc teriophage T7 mediates strand transfer through pyrimidine Natl Acad Sci USA 1996; 93: 10729–10734. dimers, mismatches, and nonhomologous regions. Proc Natl 98 Reddy G, Golub EI, Radding CM. Human Rad52 protein pro- Acad Sci USA 1997; 94: 2987–2992. motes single-strand DNA annealing followed by branch 76 Wang YY, Maher VM, Liskay RM, McCormick JJ. Carcinogens migration. Mutat Res 1997; 377: 53–59. can induce homologous recombination between duplicated 99 Johnson BL et al. Elevated levels of recombinational DNA chromosomal sequences in mouse L cells. Mol Cell Biol 1988; repair in human somatic cells expressing the Saccharomyces cer- 8: 196–202. evisiae RAD52 gene. Mutat Res 1996; 363: 179–189. 77 Tsujimura T et al. Frequency of intrachromosomal homologous 100 Park MS. Expression of human RAD52 confers resistance to recombination induced by UV radiation in normally repairing ionizing radiation in mammalian cells. J Biol Chem 1995; 270: and excision repair-deficient human cells. Proc Natl Acad Sci 15467–15470. USA 1990; 87: 1566–1570. 101 Essers J et al. Disruption of mouse RAD54 reduces ionizing 78 Mudgett JS, Taylor WD. Recombination between irradiated radiation resistance and homologous recombination. Cell 1997; shuttle vector DNA and chromosomal DNA in African green 89: 195–204. monkey kidney cells. Mol Cell Biol 1990; 10: 37–46. 102 Bertolotti R. Recombinase mediated gene therapy: strategies based on Lesch–Nyhan mutants for gene repair/inactivation 79 Saffran WA, Cantor CR, Smith ED, Magdi M. Psoralen dam- using human RAD51 nucleoprotein filaments. Biogenic Amines age-induced plasmid recombination in Saccharomyces cerevisiae: 1996; 12: 487–498. dependence on RAD1 and RAD52. Mutat Res 1992; 274: 1–9. 103 Jackson SP, Jeggo PA. DNA double-strand break repair and 80 Hays JB, Duncan BK, Boehmer S. Recombination of uracil- V(D)J recombination: involvement of DNA-PK. Trends Biochem containing lambda bacteriophages. J Bacteriol 1981; 145: 306– Sci 1995; 20: 412–415. 320. 104 Li Z et al. The XRCC4 gene encodes a novel protein involved 81 Dianov GL, Saparbaev MK, Mazin AV, Salganik RI. The chemi- in DNA double-strand break repair and V(D)J recombination. cal mutagen dimethyl sulphate induces homologous recombi- Cell 1995; 83: 1079–1089. nation of plasmid DNA by increasing the binding of RecA pro- 105 Grawunder U et al. Activity of DNA ligase IV stimulated by tein to duplex DNA. Mutat Res 1991; 249: 189–193. complex formation with XRCC4 protein in mammalian cells. 82 Faruqi AF et al. Recombination induced by triple-helix-targeted Nature 1997; 388: 492–495. DNA damage in mammalian cells. Mol Cell Biol 1996; 16: 106 Critchlow SE, Bowater RP, Jackson SP. Mammalian DNA 6820–6828. double-strand break repair protein XRCC4 interacts with DNA 83 Shinohara A, Ogawa T. Homologous recombination and the ligase IV. Curr Biol 1997; 7: 588–598. Gene therapy by gene targeting ´˜ RJ Yanez and ACG Porter 159 107 Moore PD et al. Homologous recombination in a Chinese ham- 115 Itzhaki JE, Gilbert CS, Porter AC. Construction by gene tar- ster X-ray-sensitive mutant. Mutat Res 1986; 160: 149–155. geting in human cells of a ‘conditional’ CDC2 mutant that 108 Kharbanda S et al. Functional interaction between DNA-PK rereplicates its DNA. Nat Genet 1997; 15: 258–265. and c-Abl in response to DNA damage. Nature 1997; 386: 116 Farese RV Jr, Flynn LM, Young SG. Modification of the apolipo- 732–735. protein B gene in HepG2 cells by gene targeting. J Clin Invest 109 Shafman T et al. Interaction between ATM protein and c-Abl 1992; 90: 256–261. in response to DNA damage. Nature 1997; 387: 520–523. 117 Manjunath N et al. Targeted disruption of CD43 gene enhances 110 Brown L, McCarthy N. DNA repair. A sense-abl response? T lymphocyte adhesion. J Immunol 1993; 151: 1528–1534. Nature 1997; 387: 450–451. 118 Zhen L et al. Gene targeting of X chromosome-linked chronic 111 Song K-Y et al. Accurate modification of a chromosomal plas- granulomatous disease locus in a human myeloid leukemia cell mid by homologous recombination in human cells. Proc Natl line and rescue by expression of recombinant gp91phox. Proc Acad Sci USA 1987; 84: 6820–6824. Natl Acad Sci USA 1993; 90: 9832–9836. 112 Jasin M, Elledge SJ, Davis RW, Berg P. Gene targeting at the 119 Shirasawa S, Furuse M, Yokoyama N, Sasazuki T. Altered CD4 Genes Dev 4 human locus by epitope addition. 1990; : growth of human colon cancer cell lines disrupted at activated 157–166. Ki-RAS. Science 1993; 260: 85–88. 113 Itzhaki JE, Porter AC. Targeted disruption of a human inter- 120 Waldman T, Kinzler KW, Vogelstein B. p21 is necessary for the feron-inducible gene detected by secretion of human growth p53-mediated G1 arrest in human cancer cells. Cancer Res 1995; hormone. Nucleic Acids Res 1991; 19: 3835–3842. 55: 5187–5190. 114 Itzhaki JE et al. Targeted breakage of a human chromosome 121 Montrose-Rafizadeh C et al. Gene targeting of a CFTR allele in mediated by cloned human telomeric DNA. Nat Genet 1992; 2: 283–287. HT29 human epithelial cells. J Cell Physiol 1997; 170: 299–308.