Theoretical mechanisms in targeted and random integration of transgene DNA

than eliminating or altering endogenous geneexpression.

Gene targeting is the key to surmount-ing these problems. To be of practical use,the process of gene targeting would com-prise the ability to efficiently direct the trans-gene to a specific, precisely defined genomicsite such that a particular sequence couldbe inserted or substituted at that target locus.This paper examines gene targeting at themechanistic and phenomenological levels.

1. INTRODUCTION

In animal genetic manipulation (trans-genesis) and also in gene therapy, the ran-dom nature of transgene integration posesseveral potential problems: (i) the transgenemay be expressed poorly or inappropriately;(ii) an essential gene may be disrupted, or anoncogene activated; (iii) the outcome willbe inconsistent between identically treatedcells; and (iv) transgenesis is largelyrestricted to adding new functions rather

Review

Theoretical mechanisms in targetedand random integration of transgene DNA

Kevin SMITH*

School of Science and Engineering, University of Abertay Dundee, Kydd Building,Bell Street, Dundee DD1 1HG, UK

(Received 4 September 2001; accepted 10 December 2001)

Abstract — The genetic manipulation of mammalian cells and animals would be greatly expeditedif gene targeting could be reliably achieved in the widest possible range of host cell types. Thispaper considers empirical evidence and theoretical considerations associated with transgene inte-gration, and concludes that utilisation of gene targeting in non-selective systems awaits furtherprogress in modelling homologous recombination.

transgenesis / gene targeting / gene therapy / integration / homologous recombination /concatenation

Reprod. Nutr. Dev. 41 (2001) 465–485 465© INRA, EDP Sciences, 2002

* Correspondence and reprintsE-mail: [email protected]

K. Smith

1.1. Cell types

Gene targeting is an achievable goal inmammalian cells. However, progress hasbeen limited by a lack of targeting effi-ciency. Studies on mammalian cells in vitrodemonstrate that the vast majority of inter-actions between transgene and endogenousDNA result in random rather than targetedintegrations. The reported ratio of random totargeted integration varies enormously, fromaround 1:4 to more than 1000000:1. In mostcases, the ratio is between 1 000:1 and10000:1 [4, 8, 13, 35, 43, 46]. It is note-worthy that the low efficiency of targetedintegration in mammalian cells is in markedcontrast to that which occurs when DNA istransfected into lower eukaryotes such asyeast: under appropriate empirical condi-tions, targeting is the norm and random inte-gration the exception for such organisms.

Due to the low efficiency of targeting, itis necessary to select for targeted outcomesagainst a background of random outcomes.Such selection is possible with cultured cells.Embryonic stem (ES) cells may be subjectedto selection such that targeted cells surviveat the expense of cells containing randomlyintegrated trangenes [55]. Because ES cellsare totipotent, surviving (targeted) cellswhen transferred into early embryos giverise to founder animals able to produce trans-genic offspring. The production of success-fully targeted mice by the ES cell route isnow fairly routine. However, this approachis at present limited until such time as EScells are developed for non-murine species.

Targeted outcomes cannot be selectedfor in zygotes. This does not, however, meanthat targeted events are impossible inzygotes. Very few studies have looked atgene targeting in zygotes, probably due tothe expense of gene transfer and analysis.A landmark study by Brinster et al. [6]involved the analysis of 506 transgenicfounder mice. These animals were producedby microinjecting zygotes from mice con-taining a deletion in the major histocom-

patibility (MHC) class II Eα gene. Thetransgene construct was based on sequencesfrom this gene and included the sequencesabsent in the host mice. A single mouse wasfound to have undergone targeted correc-tion of the Eα gene deletion. This studyshows that gene targeting is possible inzygotes. Although it is not possible to deter-mine an accurate frequency of gene target-ing in zygotes from this work, it appearsthat the rate of targeting (1 in 506 animals),although quite high compared with culturedcells (see above), is too low to permit theefficient use of gene targeting in zygotes.

Nuclear transfer (NT) is the latest methodfor introducing targeted changes into thegermline. NT involves replacing theoocyte’s genome with that from another cell.The genetic material from the donor cell is“reprogrammed” into totipotency by therecipient oocyte, such that the “recon-structed” egg is able to develop into a viableanimal [7]. NT per se is effective in a verybroad range of animal types including cattle,goats, mice and sheep [9]. Transgenes can beintroduced to donor cells in vitro, permit-ting the production of genetically modifiedanimals by NT [33]. Because selection canbe applied to cultured donor cells, NT can beused to produce gene-targeted transgenicanimals. Although in its infancy, the use ofNT for gene targeting certainly works, asdemonstrated by the recent generation ofthe first gene targeted sheep [26]. Themethod’s ability to work with many (pos-sibly all) animal types indicates that NTholds great promise as a tool for gene tar-geting.

Gene targeting has great potential in genetherapy, because it offers the ability to pre-cisely repair mutant genes to restore theirnormal functioning. Also, in contrast to genetherapy approaches involving randomly inte-grating transgenes, gene targeting is capableof correcting dominant, gain-of-functionmutations. In situ gene targeting in humansis a distant prospect, because selectioncannot be used in vivo. However, ex vivo

466

Mechanisms of integration of transgene DNA

transfected recombinases to enhance theefficiency of HR, the utility of selection-free gene targeting would be quite signifi-cant. Possible benefits might includecircumvention of the problem of non-avail-ability of non-murine ES cells (by allowinggene targeting to be used with zygotes), andthe use of gene targeting as part of in vivogene therapies. However, although severalcandidate genes for enhancing HR havebeen described, the use of such genes ingene targeting is at an early experimentalstage [46].

Novel alternative ways to improve tar-geting efficiencies have also been proposed.For example, triple helix-forming oligonu-cleotides (TFOs) are able to stimulate HR inmammalian cells [45]: it is possible thatTFOs could be used as agents to augmenttargeting. Similarly, endonuclease moleculesmay be able to deliver double-strandedbreaks to target DNA, leading to improvedrates of HR [37]. However, such approachesare at an early experimental stage, andthe possibilities presented by these appro-aches are beyond the scope of the presentdiscussion.

2. MODELS OF HR

HR has proved to be highly recalcitrant tomolecular/biochemical analysis. Indeed, theability to perform an entire recombinationreaction in the test tube remains anunachieved goal. Thus an understanding ofHR depends primarily on genetic data. Mostof this data has come from model systemsbased on lower eukaryotes such as mouldsand yeasts. From such data several plausiblehypotheses or models may be constructed.However, determining the most validmodel(s) by distinguishing between specificmechanistic details will require advances atthe level of precise molecular analysis. Thefollowing sections (2.1–2.6) review themajor models of HR.

targeting approaches are under develop-ment. Hatada et al. [17] used gene target-ing to correct a defective hypoxanthine phos-phoribosyltransferase (HPRT) gene inhematopoietic progenitor cells. Theapproach was similar to gene targeting withES cells or NT, in that selection was used toenrich for targeted outcomes. If similar suc-cesses can be obtained with pluripotentclonogenic cells such as hematopoietic stemcells (HSCs), it may be possible to returnsuch targeted cells to the body of the patientsuch that repopulation by the corrected cellsyields a therapeutic or curative outcome forhematological and other disorders.

1.2. Homologous recombination

Gene targeting depends upon homolo-gous recombination (HR). HR refers to anyprocess in which two similar DNAsequences interact and exchange geneticinformation. Cells have the inherent abilityof performing HR (the most obvious naturaloccurrence of HR is in meiotic recombina-tion; some DNA repair mechanismsundoubtedly use similar processes). Genetargeting seeks to harness HR such thattransgenes can be induced to undergorecombination with their homologousendogenous counterpart sequences.

Molecular biology has not yet elucidatedthe details of HR. It remains conceivablethat the efficiency of HR in mammals couldbe artificially enhanced in some way, suchthat selection would not be necessary. Forexample, if an appropriate mammalianrecombinase enzyme (or enzyme complex)was to be discovered, the relevant gene forthe recombinase might be co-introducedwith the transgene molecules in order toboost the rate of HR. The expression of therecombinase would have to be tightly con-trolled, however, since excessive produc-tion of known recombinase enzymes is fre-quently associated with recombinationabnormalities and cytotoxic effects. Never-theless, if it were to become possible to use

467

K. Smith

2.1. Simple “crossing-over”

In conceptual terms, the simplest mech-anism of HR would involve the followingevents: (1) pairing of duplex DNA on thebasis of homology; (2) double-strand cleav-age at a homologous position; (3) duplexDNA strands “crossing-over” each othersuch that the broken ends become juxta-posed; and (4) ligation (Fig. 1). This “cross-ing-over” model is able to account for the“swapping” of entire chromatid arms or seg-ments (swapped where two crossovers occurwithin a single chromatid) in meiotic recom-bination.

2.2. The Holliday model

In the “crossing-over” model, eachrecombined allele should segregate withequal frequency, since genetic informationis neither created nor destroyed in a recip-rocal “crossing-over” event. Genetic crosses,however, do not always show this 1:1 seg-regation outcome: at low frequency, “aber-rant” segregation is observed, where theratio is skewed [18, 19]. Tetrad analysis inmoulds such as neurospora illustrate thiswell. For example, in the cross between lociAB × ab, “crossing-over” should give equalnumbers of the progeny Ab and aB, repre-sented in the eight-spored ascus as a ratioof 4:4. When aberrant segregation occurs,outcomes such as 6:2 or 5:3 are observed[34]. Aberrant segregation indicates thatnonreciprocal transfer of genetic informa-tion may occur in HR, a process known asgene conversion. The “crossing-over” modelcannot account for gene conversion.

In 1964, Robin Holliday proposed amodel that is able to explain the existence ofgene conversion [18]. The Holliday modelinvolves the following events: (1) homol-ogy pairing, as per the “crossing-over”model; (2) single-strand cleavage (“nick-ing”) of both duplexes at a homologous site;(3) “strand invasion” of free ends betweenduplexes, such that the crossed strands unite

the duplexes in a structure called a Holli-day junction; (4) “branch migration” of theHolliday junction, moving the crossoverpoint away from its original position; and(5) “resolution” of the Holliday junction bysingle strand nicking (followed by ligationof free ends) (Fig. 2).

The special features of the Hollidaymodel are: (a) the Holliday junction; and(b) branch migration. The existence of theformer structure permits the latter processto occur. Holliday junctions are envisaged assymmetrical structures in which all four (sin-gle) strands of DNA are equivalent, withthe junction being in a state of rapid equi-librium in vivo (Fig. 3). In branch migra-tion, the bases on each side of the junctionexchange places. Since breakage of basepairing is balanced by formation of newbase pairing, the exchange process is ther-modynamically neutral: thus it may beenvisaged that the crossover point is able toeasily and quickly migrate along the pairedduplexes. As branch migration occurs, inits wake heteroduplex DNA will be formed.

Assuming that the homologous allelesundergoing HR are nonidentical, the het-eroduplex DNA formed by branch migra-tion will contain mismatched bases. Suchmismatches may be repaired. Any suchrepair would involve bases being replacedusing the opposite strand as a donor ofsequence information. The repair of mis-matched bases in heteroduplex DNA wouldexplain the occurrence of aberrant segrega-tion of the sort described above. Thus, theHolliday model is able to account for geneconversion.

Implications for transgenesis can bedrawn from the central features of the Hol-liday model, in that gene conversion mayoccur between transgene and endogenousDNA. Indeed, depending upon the “plane”of resolution of the Holliday junction, thetargeted transgene may not actually becomeintegrated into the target genome; howeverthe target sequences may have undergonegene conversion-mediated alteration. If this

468

Mechanisms of integration of transgene DNA 469

Figure 1. Simple Crossing-over Model.

K. Smith

process could be controlled (or its desiredoutcomes selected for), it would provide themeans to directly introduce subtle changesinto endogenous genes. To harness geneconversion in this way would be of majorimportance for gene therapy.

2.3. The Meselson-Radding model

Analysis of gene conversion genetic dataindicates that relative segregation ratios

often vary significantly from predicted val-ues, the latter values being based on theassumption that the direction of mismatchrepair is random. The apparent breakdownof randomness is most simply explained bypostulating the existence of asymmetric het-eroduplex DNA at the HR initiation region.

In 1975, Matthew Meselson and CharlesRadding proposed a model of HR that incor-porates asymmetric heteroduplex DNA [27].The Meselson-Radding model involves the

470

Figure 2. The Holliday Model.


mismatch repair is assumed to occurfollowing branch migration: thus themodel envisions both asymmetric and sym-metric heteroduplex DNA formation. TheMeselson-Radding model is able to accountfor virtually all data from meiotic recombi-nation. Additionally, the model simplifiesthe concept of initiation of recombination, inthat only a single nick is required. This con-trasts with the Holliday model, where thenuclear machinery must somehow be able toprecisely localise homologous sequenceson the recombining duplexes prior to nick-ing. Conceptually, single-stranded DNAseems to be an essential prerequisite for anysequence-level homology recognition, oth-erwise it is very difficult to envisage thetransient base-to-base interactions that pre-sumably must be an integral part of thesearch for homology. It is noteworthythat subsequent to the elucidation of theMeselson-Radding model, the E. colirecom-binase enzyme RecA was discovered: thisenzyme promotes strand exchange betweensingle-stranded DNA molecules and homol-ogous DNA duplex molecules. Morerecently, eukaryotic enzymes with abilitiessimilar to RecA have been described. Ofparticular interest is Rad51p, a homologue ofRecA: this enzymes also binds to single-stranded DNA and catalyses strandexchange [28, 39].

Implications for transgenesis can bedrawn from the special features of theMeselson-Radding model. Firstly the con-cept of a single nick is encouraging in thatit suggests that incoming transgenemolecules may be able to search throughthe endogenous duplex DNA for homolo-gous sequences, without requiring any pre-liminary homology recognition/alignment(for meiotic recombination, it is conceptuallypossible to rely to an extent on the gross-level homology alignment of homologouschromosomes). Secondly, the single nickconcept suggests that it may be possible toenhance targeting frequencies by generat-ing a single nick in a transgene prior to itsintroduction to the cell. Finally, the concept

following events: (1) homology pairing, asper the previous models; (2) nicking of (onestrand of) one duplex; (3) strand invasion ofa free end from the nicked strand into theintact duplex; (4) “strand exchange” wherebythe invading strand undergoes progressivehomology pairing with one strand of theinvaded duplex (the other, displaced strandis assumed to play no further part in therecombination event); (5) Holliday junctionformation, as per the Holliday model;(6) branch migration; and (7) resolution(Fig. 4).

The special features of the Meselson-Radding model are: (a) the single nick(2, above); and (b) strand exchange(4, above). Essentially, this creates a sin-gle-stranded gap in one of the recombiningduplexes: the loss of sequence informationin this region generates the previously men-tioned asymmetric heteroduplex DNA. Mis-match repair, of the sort envisaged for theHolliday model, cannot occur within thegapped region. Instead, “gap repair” couldoccur, with sequence information beingdonated from the single-strand within thegapped region (Fig. 5). Gap repair is thusunidirectional, and therefore may accountfor the breakdown of randomness observed ingene conversion genetic data. The Meselson-Radding model fits with the observation thatmost gene conversion genetic data is indica-tive of random, bidirectional repair, because

471

Figure 3. Holliday junction envisioned as a sym-metrical structure.

K. Smith

of strand transfer/gap repair suggests that itmay be possible to ensure that the flow ofgene conversion information is in the desireddirection (at least for sequences close to thesingle nick), i.e. from transgene to endoge-nous gene.

2.4. The double-strand-breakrepair model

Yeast cells are amenable to transfectionby plasmids. Transfected plasmids undergo

gene targeting if they contain sequenceshomologous to yeast endogenous sequences.The Meselson-Radding model is unable toexplain certain findings from yeast trans-fection experiments [29, 40]. Firstly, intro-duction of a double-strand break in theregion of homology on the plasmid leads toproducts of HR that are identical to thosefrom unbroken (i.e. closed circular) plas-mids; however, the “frequency” of target-ing is markedly increased. Secondly, intro-duction to the plasmid of a gap (in the region

472

Figure 4. The Meselson-Radding Model.


The special features of the DSBR modelare: (a) the double-strand break/gap (2 and3, above); and (b) the generation of two Hol-liday junctions (7, above). As with theMeselson–Radding model, both gap repairand mismatch repair are envisaged, leadingto gene conversion.

Implications for transgenesis can bedrawn from the special features of the DSBRmodel, in addition to the implications aris-ing from the Meselson-Radding model.Firstly, the double-strand break concept sug-gests that it may be possible to enhance tar-geting frequencies by generating a break ina transgene prior to its introduction to thecell. Secondly, the concept of gap genera-tion/repair suggests that information flowat the region of the gap may go in the“wrong” direction, i.e. from endogenousgene to transgene, at least in cases wherethe transgene construct has been introducedas a linear molecule. In practice this neednot always be a problem, because the trans-gene can be designed such that although itsrecombining (homologous) element(s) mayundergo such conversion, the net result is atargeted integration of the desired (nonho-mologous) sequences.

2.5. Chimeric oligonucleotidesand HR

Chimeric oligonucleotides (COs) aresmall (ca. 50 bp), self-complementary DNA-RNA oligonucleotides with a double-hairpinconfiguration. These highly specializedtransgene molecules have been developedas tools for gene targeting [51]. By virtueof their inherent structure, COs cannot pro-duce large (several nucleotides) changes intargeted genes; however they are able togenerate base substitutions at preciselydefined genomic positions. An importantfeature of COs is that they have a higherstructural integrity than conventional trans-gene molecules, and are thus less prone todestruction within the nucleus prior to tar-geting. Following initial reports of success

of homology) yields the same HR productsas does the break, however the HR fre-quency is markedly elevated.

In 1983, Jack Szostak, Terry Orr-Weaver,Rodney Rothstein and Frank Stahl proposeda model of HR that accounts for the aboveobservations [40]. The Szostak, Orr-Weaver,Rothstein and Stahl, or double-strand-breakrepair (DSBR) model of HR involves thefollowing events: (1) homology pairing;(2) double-strand breakage of one duplex;(3) generation of a gap by exonucleaseaction on the double-stranded break;(4) strand invasion of a free end from thegapped strand into the intact duplex;(5) strand exchange, accompanied by dis-placement of a “D loop” of single-strandedDNA from the uncut duplex; (6) invadingstrand acting as primer for DNA synthesisusing nondisplaced strand as template, lead-ing to D loop enlargement; (7) D loop “inva-sion” into gap site on opposite duplex, form-ing asymmetric duplex DNA and generatingtwo Holliday junctions; (8) branch migra-tion (of both Holliday junctions); and(9) resolution (by cutting/ligation at bothHolliday junctions) (Fig. 6).

473

Figure 5. Gap Repair.

K. Smith474

Figure 6. Double Strand Break Repair model.


of transgenesis, deciding between the twomodels is important, in that the implicationsdiffer between the two models. These impli-cations include: (a) the form of DNA thatmay be used for the transgene; and (b) thedirection of information flow between trans-gene and endogenous gene (and the associ-ated transgene design implications).

At the moment, the fine molecular/biochemical details of HR are poorlyunderstood, therefore it is not yet possibleto definitively decide which model(Meselson-Radding or DSBR) is most accu-rate. Indeed, it has been suggested that “mul-tiple” recombination pathways may exist[34]. It is possible that different classes oforganisms have their own specific pathways;but it is also possible that each class oforganism is able to carry out HR by the useof more than one pathway. In simple sys-tems such as E. coli,many recombination-disabled mutants are known. However, notall mutants block all HR events: one muta-tion may block interplasmid recombinationbut leave conjugational recombination unaf-fected, for example. If such multiple path-ways exist in higher eukaryotic cells, it mayin principle be possible to influence the hostcell (for example by altering the structureof the transgene molecule) such that oneparticular pathway is used. Such “pathwayselection” could influence the outcome ofa targeting event, for example by ensuringthat information flow occurs in the desireddirection.

It is conceptually possible that indepen-dent, different HR pathways may be opera-tional during a single targeting event. This isprobably most likely with replacement(“ends-out” or Ω-type) transgenes, in whichthe homology regions are present at the endsof the transgene, separated by heterologousDNA. In such cases, an independent HRevent is expected at each end of the homol-ogously aligned transgene: such eventswould not necessarily be identical in nature.Indeed, it is also possible that HR may occurat one transgene end and nonhomologous

in 1996 [10, 52], COs have been used tocorrect various point mutations in a rangeof mammalian cells. However, reported ratesof gene repair have been highly variable(ranging from zero to ca. 40%), and occa-sional unexpected mutational effects havebeen found [11, 50]. These problems haveengendered a certain amount of controversyconcerning the use of COs in gene target-ing, particularly in the context of gene ther-apy.

The mechanism of CO action has notyet been fully elucidated. However, genecorrection experiments designed to explorethis question indicate that targeting by COsinvolves several of the fundamental featuresfound in the foregoing models of HR, suchas homology pairing and strand exchange[14, 15]. Thus, the action of COs representsa special case of HR. Extending from thefindings of Gamper et al, a provisionalmodel for the HR involved in CO may beenvisaged, as follows: (1) homology pair-ing, involving a stabilised displacement loopin which the entire CO resides; (2) strandinvasion of the free end of the CO into onetarget strand, leading to Holliday junctionformation; (3) branch migration; and(4) resolution (Fig. 7). If the CO includes asingle base difference compared with the tar-get site, the result of steps 1–7 above will bea mismatch. As with the previously describedmodels of HR, such mismatches may berepaired. In 50% of cases in which the mis-match is repaired, the target sequence shouldbe converted to that of the targeting CO.

2.6. Conclusion

The simple “crossing-over” and Holli-day models are inadequate as explanationsof HR. The Meselson-Radding model is ableto account for data from genetic crossesequally as successfully as the DSBR model,and has the advantage of being conceptu-ally the simpler of the two models. How-ever, only the DSBR model is able toaccount for yeast transfection data. In terms

475

K. Smith

integration at the other [1]. A further possi-bility, involving a variant form of DSBRhas been suggested by Li et al. [24]. In thismodel, the transgene ends are envisagedinvading the homology sites within the tar-get duplex, to produce one Holliday junc-tion at each site. Thus, the resulting structure(encompassing the entire transgene) wouldbe equivalent to a single unresolved DSBRintermediate (Fig. 8).

Finally, the special case of COs fits wellwith the concept of multiple HR pathways.Although the mechanism of targeting used

by COs has not yet been fully elucidated, itseems certain that key steps from the generalmodels of HR are utilised by COs to yieldgene correction outcomes.

3. TRANSGENE INTEGRATIONIN MAMMALIAN CELLS

This section addresses the issue of whathappens to transgene molecules upon intro-duction to the nucleus or pronucleus of thehost cell. First, the events that may lead to

476

Figure 7. Proposed model for gene targeting by chimeric oligonucleotides.


• Integration occurs at a frequency ofbetween 10–30% of cells in which trans-gene DNA is delivered to the nucleus;

• Integration occurs at one or, rarely, afew chromosomal sites per nucleus;

• Integrated DNA is usually present inthe form of a multicopy array;

• The vast majority of arrays consist ofhead-to-tail associations.

The precise molecular mechanisms ofrandom integration are not known. How-ever, experimental data such as that outlinedabove has allowed the construction of mod-els of random integration. A general modelis presented below (Sects. 3.1.1 and 3.1.2).

3.1.1. Concatenation

The fact that transgenes are usually pre-sent as arrays (concatemers) indicates that

non-targeted (i.e. random) integration areconsidered. Second, the phenomenology oftargeted integration is reviewed.

3.1. Random integration

The arrangements of exogenous DNAintegrated into the chromosomes of culturedmammalian somatic cells are very similarto the arrangements found in transgenic ani-mals [5, 12, 16]. Furthermore, this is trueregardless of whether the transgenic ani-mals have been derived from one-cellembryos or ES cells. This suggests that theunderlying molecular mechanisms of ran-dom integration are essentially the same inall cell types.

The main features of randomly integratedexogenous DNA are as follows:

477

Figure 8. Lin et al. (2001) model for mammalian gene replacement.

K. Smith

extrachromosomal events (concatenation)take place prior to chromosomal integra-tion.

Random end-to-end joining (ligation) oftransgene molecules should generate head-to-tail, head-to-head and tail-to-tail associ-ations in a ratio of 2:1:1. However, as men-tioned above (Sect. 3.1), the vast majority ofarrays take the form of head-to-tail associ-ations. Therefore, end-to-end joining can-not be an adequate explanation of concate-nation. However, rare head-to-head/tail-to-tail associations do occur, so ligationwould appear to be a possibility. The sim-plest explanation lies in the kinetics of freetransgene molecules: it must be stochasti-cally infrequent for any two transgenemolecules to meet together in an end-to-endfashion, and stay together long enough for amolecule of DNA ligase to unite them [2].

If end-to-end joining cannot explain themajority of concatenation events, anotherform of interaction between transgenemolecules must be operative. The mostlikely process is extrachromosomal HRbetween circular and linear molecules. Thismechanism can be shown, in formal geo-metric terms, to generate exclusivelyhead-to-tail concatemers [2].

A prerequisite for concatenation by extra-chromosomal HR is the co-existence of bothcircular and linear transgene molecules.Experimental data shows that head-to-tailarrays result (with equal frequency) fol-lowing the introduction of either circular orlinear molecules [5, 12]. Therefore it is nec-essary to postulate the existence of twonuclear processes: (1) circularisation (byligation of the free ends of a transgenemolecule); and (2) linearisation (by randomnuclease action).

Linearisation would generate circularlypermuted molecules. HR could then occurbetween circularly permuted and circularmolecules, or between circularly permutedand input linear molecules, or between indi-vidual (different) circularly permutedmolecules: in all cases the effect would be

the formation of a head-to-tail concatemer.Repeated rounds of HR would extend thearray (i.e. increase the number of transgenecopies therein) (Fig. 9).

Several cultured cell studies have demon-strated that linear DNA molecules are cir-cularised by intracellular ligation, and anumber of similar studies have indicatedthat circular DNA molecules are randomlycleaved [3]. Thus, circularly permutedmolecules will be produced following intro-duction of exogenous DNA molecules.

Concatenation of circularly permutedmolecules can most simply be explained bythe following events: (1) homology pairing;(2) exposure of single-strand substrates atthe end of each duplex; (3) formation of aduplex between exposed complementarystrands; and (4) resolution (by repair of theduplex) (Fig. 10).

The fact that extrachromosomal HRoccurs with high frequency amongst indi-vidual transgene molecules contrasts sharplywith the low frequency of HR betweentransgenes and endogenous chromosomalsequences (see Sect. 1.1). There wouldappear to be two possible explanations, asfollows: (a) the “free” nature of the inter-acting transgene molecules in some wayenables HR to proceed very efficiently;and/or (b) the free (non-telomeric) ends ofthe interacting transgene molecules are verygood substrates for recombinase enzymeactivities. However, the molecular/bio-chemical details of HR remain to be eluci-dated (Sect. 2); thus it is not yet possible togive a precise explanation for the contrastthat exists between extrachromosomal HRand the HR that underlies gene targeting.

3.1.2. Illegitimate recombination

Transgene or transgene array integrationonly occurs in a minority of surviving trans-fected cells, suggesting that (nontargeted)integration is the result of a rare intranu-clear event. The simplest model wouldsuppose that the rare intranuclear event is

478

Mechanisms of integration of transgene DNA 479

Figure 9. Model for concatenation.

K. Smith

chromosomal (double-strand) breakage fol-lowed by end-joining between the transgeneends and the chromosomal broken ends.Certainly, the frequency of DNA integra-tion is increased by irradiation of the trans-fected cells [30].

An alternative mode of chromosomalintegration could be “illegitimate” recom-bination between very poorly matchedsequences. Studies of the nucleotidesequences at exogenous-endogenous DNAjunctions have shown illegitimate recombi-nation in a number of cases, although theoverall number of studies is small [2].

Interestingly, a diverse range of chro-mosomal sequence disturbances has beenfound in junctional studies. These include, inorder of frequency: deletions, duplications,inversions and more complex rearrange-ments including the appearance of sequencesfrom elsewhere in the genome or even ofunknown origin. Bishop [2] has proposedan explanation of these observations that

sees exposed single-stranded ends of trans-gene molecules initiating recombination byinvading DNA duplexes.

It is now known that the majority oftransgenic founder animals are mosaics.This suggests that integration occurs dur-ing DNA replication. Wilkie and Palmiter[48] have proposed that the free ends of thetransgene initiate recombination by invadinga replication “eye”.

It may be that, in a situation analogousto that of deciding between alternative path-ways of HR (see Sect. 2.5), more than oneintegration route is possible for randomlyintegrated transgenes. As with gene target-ing, a full understanding of the mechanismsof random transgene integration awaits fur-ther study at the molecular/biochemicallevel. Such understanding is important fromthe perspective of gene targeting, becauseit may be that targeting frequencies couldbe enhanced by somehow blocking “back-ground” (i.e. random) integration events.

3.2. Targeted integration

The utility of gene targeting as a means ofgene therapy requires systematic study intothe mechanisms of the process. Given themany variables involved in any gene tar-geting experiment (i.e. cell type, transfec-tion method, transgene design, target site,etc.), it is unsurprising that progress towardsa detailed understanding has been relativelyslow.

Nevertheless, various studies have pro-vided important insights into several aspectsof mammalian gene targeting. Initial studiesinvolved the use of artificially introducedselectable target sites in mammalian celllines, such that rare targeting events could berecovered. Later studies have used targetsof natural loci in mammalian cell lines, EScells and mammalian zygotes. Variousobservations and inferences from such stud-ies are considered in the following sections(3.2.1–3.2.8).

480

Figure 10. Model for nonconservative HRbetween circularly permuted (extrachromoso-mal) transgene molecules.


3.2.2. Transgene sequences

Gene targeting is dependent on HR,which is in turn dependent on shared homol-ogy between recombining DNA sequences.The question is, how much homology isrequired for optimal efficiency of gene tar-geting? There is at present no completeanswer to this question, because systematicstudies are lacking, and comparison betweenseparate studies is very problematic due tothe existence of several variables other thanthe extent of homology. Such variablesinclude other transgene sequences, the phys-ical state of the transgene, the cell type used,the target gene and the actual nature (ratherthan simply the extent) of homology. Nev-ertheless, several studies have provided apartial answer. Thomas and Capecchi [41],targeting the HPRT gene in ES cells, foundthat targeting efficiency appeared to bestrongly dependent upon the degree ofhomology possessed by the transgene (andshared with the target locus). Specifically, anincrease in homology from 4 kb to 9.1 kbcorrelated with 40-fold increase in the rate oftargeting, as measured by the ratio of tar-geted:random integration. Shulman et al.[36], targeting an immunoglobulin gene inhybridoma cell lines, varied the extent ofhomology from 1.2 to 9.5 kb. Again thedegree of homology correlated with target-ing efficiency, with a 25-fold increase seenover the range from 2.5 to 9.5 kb. Noincrease in targeting efficiency was observedbetween 1.2 and 2.5 kb. From these studiesit can be concluded that the frequency ofgene targeting is roughly proportional to theextent of homology shared by the transgeneand its target locus. However, it is notablethat the effects of very large (> 9.5 kb)homologies are not known.

Besides homology length, base pair vari-ation may affect the rate of targeting. This isevident from experiments comparing iso-genic and nonisogenic transgenes. Thehomology region(s) in an isogenic trans-gene is derived from the same (syngenic)laboratory animal strain as the target

3.2.1. Transfection method

In methods such as co-precipitation, theexogenous DNA molecules must somehowmigrate through the cytoplasm of the hostcell in order to reach the nucleus. Of theDNA that survives this journey, a substan-tial proportion sustains some degree ofendonucleolytic or exonucleolytic damage.In contrast, virtually no damage occurs toDNA delivered directly into the nucleus bymicroinjection [22, 47].

Gene targeting using a damaged trans-gene is unlikely to be desirable in gene ther-apy. Beyond this concern, it may be the casethat damage sustained by incoming trans-gene molecules renders them less able toundergo HR. The reasons for this are notknown, but the effect seems to exist. Forexample, separate studies were conductedby Lin et al. [25] and Thomas et al. [42]both involving gene targeting of artificiallyintroduced defective genes in mouse fibrob-lasts. Lin et al. reported a ratio of randomintegration to gene targeting of 100000:1whereas Thomas et al. reported a ratio of100:1. The major difference between thetwo sets of studies lies in the method oftransfection, with Lin et al. using CaPO4co-precipitation and Thomas et al. usingmicroinjection. Similarly, in a systematicstudy using the adenine phosphoribosyl-transferase (APRT) locus in Chinese ham-ster ovary cells, Vasquez et al. [46] com-pared the targeting efficiencies associatedwith various transfection methods. In theseexperiments, mass-delivery methods (ele-croporation, co-precipitation, liposomes)yielded an average ratio of random integra-tion to gene targeting of 200000:1 (range2400:1 to 350000:1) compared with a ratioof 1:15 for microinjection. In addition tothe possibility that HR frequency is reduceddue to DNA damage associated with themass-transfection methods, it has been sug-gested that the larger numbers of transgenemolecules delivered by the mass–transfec-tion methods may overwhelm the HRmachinery [46].

481

K. Smith

animal. Therefore, isogenic trangenes con-tain homology blocks that are geneticallyidentical (or virtually identical) to the tar-get homology regions. By contrast, a homol-ogy stretch in a nonisogenic transgene willtypically be interrupted by a number ofslight sequence divergences, such as base-pair mismatches and small deletions/ inser-tions. In a series of experiments designedto compare isogenic and nonisogenic trans-genes, Riele et al. [31] reported a 20-foldimprovement in targeting efficiency when anisogenic transgene was used, yielding aremarkably favourable ratio of random totargeted integration (approximately 1:4).The target site was the retinoblastoma sus-ceptibly gene (Rb) in an ES cell line derivedfrom mouse strain 129. The isogenic andnonisogenic transgene constructs contained17 kilobases of homology, derived respec-tively from (a) mouse strain 129 and(b) mouse strain BALB/c. Similar resultswere obtained from a systematic study byVan Deursen and Wieringa [44], in whichthe creatine kinase M gene (CKM) in EScells was targeted with transgenes sharing9 kb of homology with the target site. Inthese experiments, an increase in targetingefficiency of approximately 25-fold wasobserved when isogenic transgenes wereused. Thus, the use of isogenic DNA intransgenes appears to hold promise forimproving gene–targeting efficiencies. How-ever, it has not been established whether theoutcomes described above are applicable toother target genes in other cell types. More-over, there exists a dearth of systematicknowledge concerning the nature, frequencyand extent of heterologies that may affecttargeting efficiencies. Nevertheless, it is rea-sonable to conclude that, all other factorsbeing equal, transgenes employing perfecthomology are likely to yield better targetingefficiencies in comparison with those usinginterrupted homology.

3.2.3. Physical state of the transgene

In keeping with yeast data (see Sect. 2.4),all studies agree that linearization of the

transgene (in the region of homology)greatly enhances targeting efficiency (see[21] for example). This finding is supportiveof the DSBR model (Sect. 2.4) as an expla-nation of the mechanism of gene targeting(but see also Sect. 2.6). Beyond lineariza-tion, stripping the transgene ends to exposearound 200 nucleotides of single-strandedDNA appears to further enhance targeting[46]. Although the underlying mechanismis not understood, the finding that single-stranded transgene tails enhance targetingfits well with the notion that HR involvessingle-stranded DNA ends invading targetduplex DNA [38].

3.2.4. Transgene copy number

A targeting transgene molecule presum-ably has to “search” through the hostgenome until it “finds” its target sequence.Therefore, it might be expected that target-ing efficiency would be enhanced by increas-ing the number of transgene molecules intro-duced to the host cell. An increasingcytotoxic effect is observed where increas-ingly large quantities of DNA are microin-jected into the nucleus. However, the tar-geting efficiency can still be obtained, bycalculating the proportion of surviving cellsthat have been successfully targeted. How-ever, no study has demonstrated a correla-tion between transgene copy number andtargeting frequency. This area has not beenextensively researched, but at least two stud-ies have positively determined that thereappears to be no relationship whatsoeverbetween the number of copies introducedand the efficiency of [32, 41]. The inferencemust be that the initial search for homologydoes not seem to be the rate-limiting stepfor targeting. This conclusion is also sup-ported by experiments involving amplifica-tion of the target site (see Sect. 3.2.8):an increased target copy number doesnot appear to enhance the frequency of tar-geting. Indeed, if Vasquez et al. [46] arecorrect in postulating that too manytransgene molecules may overwhelm the

482


copies, located in three clusters on differ-ent chromosomes. Gene targeting rates werethe same in both cell lines. Similarly,Thomas et al. [41] used three cell lines con-taining integrated target plasmid sequencespresent as one copy, four dispersed copies orfive tandem copies. Again, the rates of genetargeting were similar in all three lines.Such results infer, as suggested previously(Sect. 3.2.4), that the initial search forhomology does not appear to be therate–limiting step for targeting. Interest-ingly, recent (target amplification) experi-ments in yeast have suggested that the fre-quency of gene targeting does dependon the number of target copies. Indeed,Wilson et al. [49] report a linear relation-ship between target site copy number andthe rate of targeting in yeast. The reason forthis difference between yeast cells and mam-malian cells remains to be established.

4. CONCLUDING REMARKS

Although major progress in model build-ing has been made in recent decades, exten-sive biochemical analysis of the molecularmechanism(s) of HR will be required if genetargeting is to be better understood. On adifferent level, the phenomenology of genetargeting also requires extensive systematicanalysis.

Absolute frequencies of gene targetingin mammalian cells remain low, and theratio of targeted to random integration isstill heavily weighted in favour of the lat-ter. Until the frequencies are improved, thepotential use of gene targeting in non-selec-tive systems will be limited. Such improve-ment is likely to depend upon a moredetailed understanding of gene targeting,which is in turn dependent upon systematicanalysis of the sort described above.

ACKNOWLEDGEMENTS

Artwork by 1st Class Media: http://www.1stclass.uk.com.

HR machinery (see Sect. 3.2.1 above), itmay turn out to be the case that an inverserelationship exists between transgene copynumber and targeting efficiency.

3.2.5. Position of target site

The position of the target site within thegenome does not strongly influence the fre-quency of HR. For example, twelve inde-pendent recipient cell lines were producedby Thomas et al. [41], each line containinga defective neomycin-resistance gene inte-grated at a different chromosomal position.Introduction of targeting DNA constructsgave similar gene targeting frequencies inall twelve lines.

3.2.6. Recombination hotspots

Targeting the endogenous β2-microglob-ulin gene in ES cells, Zijlstra et al. [54]achieved a very high frequency of gene tar-geting (a ratio of about 1: 25 targeting torandom integration). Other investigators forthe β2-microglobulin target gene [20] andfor the Hox 3.1 target gene [23] havereported similarly high targeting frequen-cies. Such studies support the existence ofrecombination “hotspots”. However, thesequences involved in such hotspots remainto be elucidated.

3.2.7. Target gene activity

There is no evidence that the level ofexpression of the target gene correlates withthe frequency of gene targeting (see [20]for example).

3.2.8. Target copy number

As noted in Sect. 3.2.4, experimentalamplification of the target site does notappear to enhance the frequency of targeting.For example, Zheng and Wilson [53] usedtwo mammalian cell lines, one of whichcontained 2 target gene copies. The secondline contained around 800 target gene

483

K. Smith

REFERENCES

[1] Berinstein N., Pennell N., Ottaway C.A.,Shulman M.J., Gene Replacement with One-Sided Homologous Recombination, Mol. Cell.Biol. 12 (1992) 360–367.

[2] Bishop J.O., Chromosomal insertion of foreignDNA, Reprod. Nutr. Dev. 36 (1996) 607–618.

[3] Bishop J.O., Smith P., Mechanism of Chromo-somal Integration of Microinjected DNA, Mol.Biol. Med. 6 (1989) 283–298.

[4] Bollag R.J., Waldman A.S., Liskay R.M.,Homologous Recombination in Mammalian-Cells, Annu. Rev. Genet. 23 (1989) 199–225.

[5] Brinster R.L., Chen H.Y., Trumbauer M.E.,Yagle M.K., Palmiter R.D., Factors Affectingthe Efficiency of Introducing Foreign DNA intoMice by Microinjecting Eggs, Proc. Natl. Acad.Sci. USA 82 (1985) 4438–4442.

[6] Brinster R.L., Braun R.E., Lo D., AvarbockM.R., Oram F., Palmiter R.D., Targeted Cor-rection of a Major Histocompatibility Class–Ii E-Alpha-Gene by DNA Microinjected into MouseEggs, Proc. Natl. Acad. Sci. USA 86 (1989)7087–7091.

[7] Campbell K.H.S., McWhir J., Ritchie W.A.,Wilmut I., Sheep cloned by nuclear transferfrom a cultured cell line, Nature (1996) 64–65.

[8] Capecchi M.R., Altering the Genome by Homol-ogous Recombination, Science 244 (1989)1288–1292.

[9] Clark A.J., Burl S., Denning C., Dickinson P.,Gene targeting in livestock: a preview, Trans-genic Res. 9 (2000) 263–275.

[10] Cole-Strauss A., Yoon K., Xiang Y., Byrne B.C.,Rice M.C., Gryn J., Holloman W.K., KmiecE.B., Correction of the Mutation Responsiblefor Sickle Cell Anemia by an RNA-DNAOligonucleotide, Science 273 (1996) 1386–1388.

[11] Cole-Strauss A., Gamper H., Holloman W.K.,Munoz M., Cheng N., Kmiec E.B., Targetedgene repair directed by the chimeric RNA/DNAoligonucleotide in a mammalian cell-free extract,Nucleic Acids Res. 27 (1999) 1323–1330.

[12] Folger K.R., Wong E.A., Wahl G., CapecchiM.R., Patterns of Integration of DNA Micro-Injected into Cultured Mammalian Cells – Evi-dence for Homologous Recombination betweenInjected Plasmid DNA-Molecules, Mol. Cell.Biol. 2 (1982) 1372–1387.

[13] Frohman M.A., Martin G.R., Cut, Paste, andSave – New Approaches to Altering SpecificGenes in Mice, Cell 56 (1989) 145–147.

[14] Gamper H.B., Cole-Strauss A., Metz R., ParekhH., Kumar R., Kmiec E.B., A plausible Mecha-nism for Gene Correction by Chimeric Oligonu-cleotides, Biochemistry-US 39 (2000)5808–5816.

[15] Gamper H.B., Parekh H., Rice M.C., BrunerM., Youkey H., Kmiec E.B., The DNA strand of

chimeric RNA/DNA oligonucleotides can directgene repair/conversion activity in mammalianand plant cell-free extracts, Nucleic Acids Res.28 (2000) 4332–4339.

[16] Gordon J.W., Ruddle F.H., DNA-MediatedGenetic Transformation of Mouse Embryos andBone Marrow – a Review, Gene 33 (1985)121–136.

[17] Hatada S., Nikkuni K., Bentley S.A., Kirby S.,Smithies O., Gene correction in hematopoieticprogenitor cells by homologous recombination,Proc. Natl. Acad. Sci. USA 97 (2000)13807–13811.

[18] Holliday R., A mechanism for gene conversionin fungi, Genet. Res. (1964) 282–304.

[19] Holliday R., Genetic recombination in fungi,in: Peacock J.W., Brook R.D. (Eds.), Replicationand recombination of genetic material, Can-berra, Australian Academy of Science, 1968,pp. 157–174.

[20] Koller B.H., Smithies O., Inactivating the Beta-2-Microglobulin Locus in Mouse EmbryonicStem-Cells by Homologous Recombination,Proc. Natl. Acad. Sci. USA 86 (1989)8932–8935.

[21] Kucherlapati R.S., Eves E.M., Song K.Y., MorseB.S., Smithies O., Homologous Recombinationbetween Plasmids in Mammalian Cells can beEnhanced by Treatment of Input DNA, Proc.Natl. Acad. Sci. USA 81 (1984) 3153–3157.

[22] Lebkowski J.S., Dubridge R.B., Antell E.A.,Greisen K.S., Calos M.P., Transfected DNA isMutated in Monkey, Mouse, and Human Cells,Mol. Cell. Biol. 4 (1984) 1951–1960.

[23] Lemouellic H., Lallemand Y., Brulet P., Tar-geted Replacement of the Homeobox Gene Hox-3.1 by the Escherichia-Coli LacZ in MouseChimeric Embryos, Proc. Natl. Acad. Sci. USA87 (1990) 4712–4716.

[24] Li J., Read L.R., Baker M.D., The Mechanism ofMammalian Gene Replacement Is Consistentwith the Formation of Long Regions of Het-eroduplex DNA Associated with Two Cross-ing-Over Events, Mol. Cell. Biol. 21 (2001)501–510.

[25] Lin F.L., Sperle K., Sternberg N., Recombina-tion in Mouse L-Cells between DNA Introducedinto Cells and Homologous ChromosomalSequences, Proc. Natl. Acad. Sci. USA 82(1985) 1391–1395.

[26] McCreath K.J., Howcroft J., Campbel K.H.S.,Colman A., Schnieke A.E., Kind A.J., Produc-tion of gene-targeted sheep by nuclear transferfrom cultured somatic cells, Nature 405 (2000)1066–1069.

[27] Meselson M., Radding C.M., A general modelfor genetic recombination, Proc. Natl. Acad.Sci. USA. 73 (1975) 358–361.

484


Sites in the Mammalian Genome, Cell 44 (1986)419–428.

[43] Thomas K.R., Deng C.X., Capecchi M.R., High-Fidelity Gene Targeting in Embryonic Stem-Cells by Using Sequence Replacement Vectors,Mol. Cell. Biol. 12 (1992) 2919–2923.

[44] Van Deursen J., Wieringa B., Targeting of theCreatine Kinase-M Gene in Embryonic Stem-Cells Using Isogenic and Nonisogenic Vectors,Nucleic Acids Res. 20 (1992) 3815–3820.

[45] Vasquez K.M.,Wilson J.H., Triplex-directedmodification of genes and gene, activity, TrendsBiochem. Sci. 23 (1998) 4–9.

[46] Vasquez K.M., Marburger K., Intody Z., WilsonJ.H., Manipulating the mammalian genome byhomologous recombination, Proc. Natl. Acad.Sci. USA 98 (2001) 8403–8410.

[47] Wake C.T., Gudewicz T., Porter T., White A.,Wilson J.H., How Damaged Is the Biologically-Active Subpopulation of Transfected DNA, Mol.Cell. Biol. 4 (1984) 387–398.

[48] Wilkie T.M., Palmiter R.D., Analysis of theIntegrant in Myk-103 Transgenic Mice in WhichMales Fail to Transmit the Integrant, Mol. Cell.Biol. 7 (1987) 1646–1655.

[49] Wilson J.H., Leung W.Y., Bosco G., De D., Thefrequency of gene targeting in yeast depends onthe number of target copies, Proc. Natl Acad.Sci. USA 91 (1994) 177.

[50] Xiang Y., Cole-Strauss A., Yoon K., Gryn J.,Kmiec E.B., Targeted gene conversion in amammalian CD34(+)-enriched cell populationusing a chimeric RNA/DNA oligonucleotide,J. Molec. Med. 75 (1997) 829–835.

[51] Ye S., Cole-Strauss A., Frank B., Kmiec E.B.,Targeted gene correction: a new strategy formolecular medicine, Mol. Med. Today 4 (1998)431–437.

[52] Yoon K., Cole-Strauss A., Kmiec E.B., Tar-geted gene correction of episomal DNA in mam-malian cells mediated by a chimeric RNA/DNAoligonucleotide, Proc. Natl. Acad. Sci. USA 93(1996) 2071–2076.

[53] Zheng H., Wilson J.H., Gene Targeting in Nor-mal and Amplified Cell-Lines, Nature 344(1990) 170–173.

[54] Zijlstra M., Li E., Sajjadi F., Subramani S.,Jaenisch R., Germ-Line Transmission of a Dis-rupted Beta-2-Microglobulin Gene Produced byHomologous Recombination in EmbryonicStem-Cells, Nature 342 (1989) 435–438.

[55] Zimmer A., Manipulating the Genome byHomologous Recombination in EmbryonicStem-Cells, Annu. Rev. Neurosci. 15 (1992)115–137.

[28] Ogawa T., Yu X., Shinohara A., Egelman E.H.,Similarity of the Yeast Rad51 Filament to theBacterial Reca Filament, Science 259 (1993)1896–1899.

[29] Orrweaver T.L., Szostak J.W., Fungal Recom-bination, Microbiol. Rev. 49 (1985) 33–58.

[30] Perez C.F., Botchan M.R., Tobias C.A., DNA-Mediated Gene Transfer Efficiency Is Enhancedby Ionizing and Ultraviolet-Irradiation of RodentCells In vitro, Radiat. Res. 104 (1985) 200–213.

[31] Riele H.T., Maandag E.R., Berns A., HighlyEfficient Gene Targeting in Embryonic Stem-Cells through Homologous Recombination withIsogenic DNA Constructs, Proc. Natl. Acad.Sci. USA 89 (1992) 5128–5132.

[32] Rommerskirch W., Graeber I., Grassmann M.,Grassmann A., Homologous Recombination ofSV40 DNA in Cos7 Cells Occurs withHigh–Frequency in a Gene Dose IndependentFashion, Nucleic Acids Res. 16 (1988) 941–952.

[33] Schnieke A.E., Kind A.J., Ritchie W.A., MycockK., Scott A.R., Ritchie M., Wilmut I., ColmanA., Campbell K.H.S., Human factor IX trans-genic sheep produced by transfer of nuclei fromtransfected fetal fibroblasts, Science 278 (1997)2130–2133.

[34] Sedivy J.M., Joyner A.L., Gene Targeting,Oxford University Press USA, Publ., New York,1993.

[35] Sedivy J.M., Dutriaux A., Gene targeting andsomatic cell genetics – a rebirth or a coming ofage? Trends Genet. 15 (1999) 88–90.

[36] Shulman M.J., Nissen L., Collins C., Homolo-gous Recombination in Hybridoma Cells –Dependence on Time and Fragment Length,Mol. Cell. Biol. 10 (1990) 4466–4472.

[37] Smih F., Rouet P., Romanienko P.J., Jasin M.,Double-strand breaks at the target locus stimu-late gene targeting in embryonic stem cells,Nucleic Acids Res. 23 (1995) 5012–5019.

[38] Sun H., Treco D., Szostak J.W., Extensive 3’-Overhanging, Single-Stranded-DNA Associ-ated with the Meiosis-Specific Double-StrandBreaks at the Arg4 Recombination InitiationSite, Cell 64 (1991) 1155–1161.

[39] Sung P., Catalysis of ATP-Dependent Homol-ogous DNA Pairing and Strand Exchange byYeast Rad51 Protein, Science 265 (1994)1241–1243.

[40] Szostak J.W., Orrweaver T.L., Rothstein R.J.,The Double-Strand-Break Repair Model forRecombination, Cell 33 (1983) 25–35.

[41] Thomas K.R., Capecchi M.R., Site-DirectedMutagenesis by Gene Targeting in MouseEmbryo-Derived Stem-Cells, Cell 51 (1987)503–512.

[42] Thomas K.R., Folger K.R., Capecchi M.R.,High-Frequency Targeting of Genes to Specific

485

Documents

Theoretical mechanisms in targeted and random integration of transgene DNA