Introduction

Conventional gene therapy employing viral vectors still faces difficulties such as unregulated gene expression, transgene silencing, and insertional mutagenesis. Therefore, genetic manipulations allowing the precise replacement of mutated sequences at the endogenous locus are of great interest. They might be achieved by stimulating the cell's own repair machineries, for instance, by employing triple-helix-forming DNA oligonucleotides ^1,2. In an alternative attempt to minimize the introduction of exogenous DNAs into the target cells, recent research has shown that precise gene repair can be elicited purely with small single-stranded (ss) oligodeoxynucleotides (ODNs) (for a review see ³). The reaction steps involved in this nonviral approach are still not fully defined ⁴, but it has already been shown that single strandedness and an open topology of the repair molecule are highly advantageous ⁵ for the repair process to proceed via physical incorporation of the ssODN ⁶. Obviously, a productive interaction is a prerequisite that is less likely at single-copy loci. It was thus hypothesized that an additional trigger might be helpful in ssODN-directed gene repair. In this context, a DNA double-strand break (DSB) in the genome seemed attractive. This major threat leads to the activation of a multitude of pathways ⁷ vital to the survival of the affected cell. The cell can cope with DSBs in two different ways: the broken ends are either reconnected by an imprecise mechanism called nonhomologous end-joining (NHEJ) or, alternatively and less frequently, faithfully restored by homologous recombination (HR) ⁸. Attractively, the cellular DNA repair system also accepts virtually any extrachromosomal DNA as a template for homology-directed repair ⁹, which might play an important role in the evolution of eukaryotic genomes ¹⁰. The ability of a DSB to stimulate HR between homologous chromosomal segments was initially demonstrated using I-SceI ¹¹, an intron-encoded endonuclease derived from the mitochondria of Saccharomyces cerevisiae. This enzyme recognizes an 18-bp-long target sequence that can be engineered to allow defined DSB induction. Studies in yeast and vertebrate cells have demonstrated up to 1000-fold DSB-induced stimulation of HR between a donor plasmid and the target locus ¹². Furthermore, recombinant adeno-associated virus (rAAV) vectors have successfully served as gene repair templates in that intentional DSBs stimulated rAAV-mediated rates more than 100-fold ^13,14.

To avoid introducing large DNA fragments into cells, ssODNs are promising because their short lengths (10–100 nucleotides) facilitate high-quality production and reduce the likelihood of carrying promoter(-like) elements. Involvement of ssODNs in DSB repair in nonhuman systems has already been reported from work with Escherichia coli ¹⁵, the lower eukaryote S. cerevisiae ¹⁶, and Drosophila melanogaster ¹⁷.

The work presented here has been carried out to explore the ability of ssODNs to direct specific gene repair in human-derived cells in concert with an intentional DSB. A HEK-293-derived cell line has been developed. The single constitutively expressed reporter target locus consisted of an open reading frame (ORF) for LacZ followed by a spacer which translationally silenced the downstream enhanced green fluorescent protein (EGFP) ORF. By transient I-SceI expression from a cotransfected plasmid, a defined DSB could be induced in the spacer, which—upon DSB repair—might yield cells exhibiting EGFP-specific fluorescence. Since in another HEK-293-derived model system ssODNs of sense orientation have already resulted in targeted gene repair, the new model system utilized this type of ssODN to explore their influence on the natural repair process at the level of single living cells. Additionally, the question whether the ssODN becomes physically incorporated into the DSB site was addressed by employing a biotinylated variant.

Results

DSB repair in the presence of ssODNs

We generated the cell line 293LG32 to fulfill the following characteristics as a model system: (i) robust growth, (ii) high and reproducible transfectability, and (iii) a single-copy target locus to mimic the situation of a single mutated gene as found to be the cause for, e.g., X-linked diseases. In addition, the composite target locus (Fig. 1) allowed production of two different types of EGFP: (i) a fusion protein, when EGFP is linked in-frame via its N-terminus to LacZ (LacZ–EGFP; 145 kDa); this might result from endogenous repair events or by successful involvement of the modifying ODNs; and (ii) an independent protein entity (EGFP; 30 kDa), when translation initiation occurred at the EGFP-specific start codon embedded in a Kozak consensus sequence; this might happen particularly when functional promoter fragments get inserted into the break region ("promoter trap"). These two types of reporter proteins can be differentiated by fluorescence microscopy: EGFP on its own will be expressed throughout the entire cell, while the fusion product, due to its size, will be excluded from the cellular nucleus.

Figure 1.

System outline: target locus and ssODNs used in DSB repair. (A) Part of the target locus where a DSB specifically induced with the homing endonuclease I-SceI is to be repaired. The nontranscribed strand is shown. Black bar, LacZ ORF (length in nt: 3081); white bar, 50-nt spacer DNA with the I-SceI recognition site and stop codons in all reading frames; gray bar, EGFP ORF (length in nt: 720). The I-SceI and the PacI recognition sites are indicated. The LacZ stop codon and the EGFP start codon belong to the frames + 0 and + 2, respectively. As the target locus reference in (B), (C), and (D), the highlighted area is given at the nucleotide level. Start and stop codons are indicated by gray and black lines, respectively. The gray–white ruler depicts the LacZ-specific codon frame. Black and white triangles point at the I-SceI cutting positions in the nontranscribed strand and the transcribed strand, respectively. Note that positions of nucleotides refer to the numbering of the detailed target locus sequence. (B–D) Primary structures of the ssODNs. (B) DSB-shift with three features: LacZ stop codon "TAA" (nt 18–20) replaced with sense codon "GAA," the I-SceI recognition sequence (nt 21–38) carries a T A transversion (nt 33) to reduce the likelihood of an I-SceI-induced DSB after ODN-directed repair ⁴⁰, and the PacI recognition sequence (nt 39–46) has a 2-nt deletion (nt 43 and 44) enabling translation of a functional LacZ–spacer–EGFP fusion reporter protein (ORF: 3849 bp) upon ODN-directed repair. (C) DSB-deletion with 59 nt (here, nt 12–70) of the target locus deleted resulting in an in-frame fusion of both reporter ORFs (fusion ORF: 3792 bp). (D) DSB-insertion with three features: the LacZ stop codon and the I-SceI recognition site are as in DSB shift; additionally, between nt 45 and 46, a 10-nt insertion results in (i) destruction of the PacI recognition site, (ii) introduction of the NsiI and DraI restriction sites, and (iii) generation of a fusion ORF, i.e., LacZ ORF, spacer with insertion, and the EGFP ORF (fusion ORF: 3861 bp).

Full figure and legend (182K)

Initially, we transfected 293LG32 cells with 1 g of pRK5.LHA-Sce1 without or with addition of 0.5 g of ssODN. With I-SceI expression alone, we measured a rather high background of living, EGFP-positive cells in FACS analyses (0.3%; data not shown), proving the functionality of pRK5.LHA-Sce1 and the competence of 293LG32 cells to heal DNA DSBs. In the presence of ssODNs, we observed no increase in the number of green fluorescent cells, suggesting that ODN-directed repair might not have occurred. However, microscopic analyses revealed an elevated proportion of EGFP-positive cells with nonfluorescent nuclei compared to transfections with pRK5.LHA-Sce1 only (data not shown; for exemplary photographs depicting the two different EGFP phenotypes, see Fig. 2A). Therefore, we concluded that the ssODN had influenced the cellular DSB repair. To set the stage for sensitive and quantitative FACS-based detection of DSB repair, we carried out titration experiments to reduce the I-SceI-specific background close to the autofluorescence level (data not shown), against which possible repair events could be detected.

Figure 2.

Analyses of ODN-directed gene repair with concomitant DSB. (A) Microscopic analyses of 293LG32 cells. Pictures were taken 2 days after transfection. EGFP-specific fluorescence was observed in the FITC channel. The fluorescent cell doublets are encircled with white lines in the corresponding phase-contrast pictures. Examples of the two EGFP phenotypes observed after transfection are shown. Note that in transfected cell cultures both phenotypes could be observed concurrently albeit in different relative proportions. (B) FACS analyses of gene repair rates. 293LG32 cells were transfected—in the presence or absence of pRK5.LHA-Sce1—with the ssODNs given on the abscissa. Each data point represents the average value from three independent experiments done in duplicate. Standard deviations are added to each bar. (C) FACS measurements analyzing time courses of the numbers of living, EGFP-positive cells in targeted cultures. 293LG32 cells were cotransfected with the ssODNs as given below the abscissa. The experiment was carried out two times and the data of one experiment are shown.

Full figure and legend (319K)

When we omitted the I-SceI expression plasmid from the transfection mixtures, any ssODN yielded at most a few EGFP-positive cells (Fig. 2B). We concluded that an ssODN on its own was not able to efficiently modify the single-copy target locus to express the fused LacZ-EGFP. This result confirmed earlier findings from targeting chromosomal loci in which either no ⁵ or very little repair activity ^6,18 had been detected.

Next, we induced a specific DSB within the region to which the ssODNs could hybridize and the respective gene repair/modification rates were measured by FACS. In the absence of ssODNs, expression of I-SceI resulted in 0.04% EGFP-positive cells (Fig. 2B). Transfection of DNA via calcium phosphate coprecipitation strongly depends on the amounts of the nucleic acids and whether they are of single- or double-stranded topology. Therefore, we defined the reference background relevant for the 293LG32-specific DSB repair activity experiments by transfecting a complete mixture of DNAs, i.e., pRK5.LHA-Sce1, pcDNA6/myc-His A, and the control ssODN DSB-scramble. This resulted in 0.07% EGFP-positive cells. We observed an almost threefold induction above the background repair rate in cotransfections with DSB-shift which introduced few sequence changes. Titration experiments with rAAV vectors yielded rates of 1.1% (Supplementary Fig. 1), and thus, the correction frequency of ssODN was about 20% compared to that of rAAV. Lower DSB repair rates were elicited by DSB-deletion and DSB-insertion (0.11 and 0.12%, respectively). This suggested that a 59-bp deletion as well as a 10-bp insertion might have been achieved using ssODNs.

Earlier work found that only a minor population of gene-corrected cells was actively growing after targeted repair ^5,19. Thus, it was interesting to evaluate the behavior of cells when DSB repair was involved. To this end, we transfected 293LG32 cells with pRK5.LHA-Sce1 and pcDNA6/myc-His A without or with ssODNs. After 3, 7, and 14 days, we measured the numbers of living, EGFP-positive cells (Fig. 2C). They rather tended to increase, attaining final levels of 0.32% for DSB-shift, 0.21% for DSB-deletion, and 0.15% for DSB-insertion, showing that the ODN-directed DSB repair allowed the cells to expand thereafter.

Sequence analyses of target loci after DSB repair

The FACS analyses suggested successful ssODN involvement in DSB repair. For a decisive proof, we carried out sequence analyses. To this end, we sorted living, EGFP-positive 293LG32 cells and PCR amplified DNA fragments encompassing the DSB region. As shown in Fig. 3A, the undigested amplification products differed according to the actual ssODN. The DSB-shift-specific product ran—as expected—around 500 bp, showing a rather defined single band similar to the unmodified 293LG32 locus (lane A). The main products derived from loci targeted with DSB-deletion and DSB-insertion were shorter or slightly larger, as predicted by the respective ssODN sequence. Both controls without relevant ssODN (no ODN, DSB-scramble) showed main products shorter than 500 bp, suggesting as the major DSB repair pathway NHEJ, with loss of nucleotides from the targeted chromosomal site. Since the gene-modifying ODNs had been designed to eliminate the I-SceI recognition site, restriction analysis with I-SceI allowed scrutiny of the entire fragment populations for sites retained after DSB repair. The 293LG32 target locus-derived 505-bp PCR control product was completely digested (Fig. 3A; lanes A and B; note that the small digestion products are visible only in Fig. 3B). In the other pairs of lanes, hardly any differences in the product patterns were observed, suggesting that most loci had lost the I-SceI site. Longer exposure of the whole gel (Fig. 3B) corroborated this finding by showing weak signals for the digestion products (289 bp, 216 bp) in lanes 2, 4, and 8. With complete digestion of the long internal control DNA, the occurrence of these products might be explained by incomplete inactivation of the I-SceI recognition site, in the case of DSB-shift, and/or alleles derived from contaminating EGFP-negative cells. The small products found directly above the "primer dimer" signal in the DSB-deletion sample (lanes 5 and 6) were somewhat smaller than the I-SceI-specific fragments. Since they also occurred in the undigested material, they were most likely nonspecific amplification products.

Figure 3.

Sequence analyses of transgene loci from EGFP-positive cells. (A, B) Restriction analysis. The same agarose gel stained with ethidium bromide (shown in reverse contrast) is displayed. While (B) presents the entire gel, the less exposed part in (A) focuses on fragments of 400–600 bp typical of nontargeted or successfully modified loci. Genomic DNA of EGFP-positive cells (experiment I) was PCR amplified and digested in vitro with I-SceI. The locus-derived PCR product of 505 bp gave rise to a 289- and a 216-bp fragment (small black triangles), while the internal digestion control (linearized 2499-bp plasmid; big gray triangle) yielded a 1518- and a 981-bp fragment (small gray triangles). Open triangle, "primer dimer" band; M, 100 bp ladder size standard. Numbers to the right indicate fragment length in base pairs. (C) Survey of sequences. Subcloned PCR products (see (A) and (B)) derived from EGFP-positive cells were sequenced. The totaled numbers of clones are given on top of the columns. The resulting sequences have been grouped into six categories: correction—correct sequence of particular ssODN, partial correction—only part of the particular ssODN sequence, NHEJ—processed DSB ends rejoined, original sequence—as nontargeted transgene locus, ODN sequence addition—additional, unwanted DNA regions derived from transfected ssODNs, plasmid sequence addition—additional DNA segments derived (mostly) from cotransfected plasmid DNA. The percentages of sequences found for each category are written on the respective column. I and II, independent experimental runs.

Full figure and legend (303K)

Fig. 3C interprets the sequencing results. The primary sequences found in the differently targeted 293LG32 loci (the alignments of which are available as Supplementary Tables) have been grouped as indicated. Most notably, the majorities of the alleles recovered from EGFP-positive cells after transfections with DSB-shift, DSB-deletion, and DSB-insertion carried sequences identical with the respective ssODN, while the reference background of the DSB repair activity was most likely due to intrinsic NHEJ repair (see sample DSB-scramble). An exception was sample DSB-insertion with low repair rates from experiment II. We observed a reduced percentage of ODN-specific sequences together with sequences assigned to the NHEJ group. This was in agreement with the respective agarose gel also showing increased amounts of shorter PCR products typical of NHEJ (data not shown).

These findings demonstrate that exogenously added ssODNs are able to interact with the DSB repair machinery, and point mutations as well as deletions or insertions can be achieved in a human-derived chromosome. While NHEJ constituted the major DSB repair pathway when relevant ssODNs were absent or inefficient, cotransfection of modifying ssODNs yielded a major proportion of faithfully modified target loci.

We also made some observations with disturbing implications. Partial corrections occurred, e.g., in 3 cases among 21 scrutinized sequences of DSB-insertion samples. Here, the 10-bp insertion was always positioned as intended; however, either the upstream point mutations (Fig. 1D) had not been introduced, a 3-nucleotide deletion had occurred—possibly due to a synthesis error of the ssODN, or a deletion reminiscent of NHEJ happened 5' of the DSB (see also Supplementary Table 4). Another critical result was the unwanted addition into the DSB site of DNA sequences derived either from the ssODNs or from cotransfected plasmids. Two examples in which ODN sequences had been patched between the DSB ends most likely via microhomologies are (i) the 3' half of the DSB-deletion sequence upstream of the EGFP ORF (Supplementary Table 3) and (ii) the unrelated DSB-scramble sequence (trimmed on both sides by a few nucleotides) between a duplication of the 4 bp that separate the I-SceI cut positions in the transcribed and nontranscribed strands (Supplementary Table 5). Double-stranded extrachromosomal DNA sequences originated mainly from the cotransfected fill-in plasmid, and they were found in samples transfected with DSB-insertion and DSB-scramble. About 60% of these sequences contained segments of the cytomegalovirus (CMV) immediate early (IE) promoter/enhancer. This percentage might exaggerate the real proportion due to the bias introduced during analysis, i.e., selection of EGFP-positive cells.

Fate of ssODNs during targeted DSB repair

The understanding of the mechanism(s) involved in ODN-directed gene repair is still in its infancy. Replication is important ^19,20,21 in systems in which no defined DSB is induced. Compatible with this is the recent finding that ssODNs can be physically incorporated into the targeted site likely upon hybridization to the lagging strand ⁶. To test whether physical incorporation also happens in this DSB repair model, we transfected 293LG32 cells with a mixture of DSB-shift—representing the modification ODNs—and DSB-scramble. We linked the biotin label for subsequent capture on streptavidin-coated beads either to the relevant ssODN (to monitor possible physical incorporation) or to the unrelated ssODN (to control nonspecific signals). The capture procedure and the ensuing multiplex PCR are depicted in Fig. 4A. By including an internal control—also derived from an EGFP locus (293mEGFP-M12), but coamplified by a different specific primer pair—we verified the functionality of the assay even when there might be no signal from the target locus due to the lack of physical incorporation of the correction ssODN. The respective control amplification reactions (Fig. 4B, lanes 8–12) demonstrated that our assay is able to detect 50 targeted alleles steadily. Since we always worked with genomic DNA (gDNA) samples from sorted EGFP-positive cells containing about 500 targeted alleles and have never seen a specific signal, the assay sensitivity allowed us to conclude that 90% of the targeted loci recovered from EGFP-positive cells did not carry physically incorporated ODNs. Thus, at least the vast majority of loci were corrected by ssODNs which may have served as an information template being finally displaced.

Figure 4.

Capture assay analyzing the fate of an ssODN during DSB repair. (A) Scheme of the capture procedure followed by multiplex PCR. The simultaneous capturing of a control fragment and a gDNA fragment is shown. Nontranscribed and transcribed strands are depicted as black and gray lines, respectively. The biotin label is symbolized by the appended black polygon. Sequence alterations carried by the ssODNs are shown as black and white triangles. The multiplex PCR employs common backward primer gfp7 and forward primers #48-LacZ-TC and gfpM. (B) Multiplex PCR analysis of bead-captured DNAs. 293LG32 cells were cotransfected with the ssODN mixtures indicated, and gDNAs were isolated from EGFP-positive (EGFP⁺) and EGFP-negative (EGFP^-) fractions. Lanes 1–6, samples each with added internal positive control. Lanes 1 and 7, negative controls without gDNA. Lane 2, sample "untreated"—derived from nontransfected 293LG32 cells. Lanes 8–10, positive control for the multiplex PCR format detecting captured gDNA fragments. The gray triangle points to the 292-bp PCR product made from the 293mEGFP-M12-specific internal control (IC), the black triangle marks the 505-bp product derived from the 293LG32-specific test fragment (TF). Lanes 11–13, bead-free PCR controls presenting expected products. Lane M, 100 bp ladder size standard. Numbers to the right refer to the lengths in base pairs of the marked DNA fragments. Open triangle, "primer dimer" band. PCR products were made visible with ethidium bromide and are shown in reverse contrast.

Full figure and legend (340K)

Discussion

Targeted gene repair at chromosomal single-copy loci is still rather inefficient. In an attempt to improve efficacy, the work herein evaluated the influence of a defined DSB on ODN-mediated gene modifications. Employing a new model system, it has been demonstrated that ssODNs reproducibly elicit gene modification/repair if a DSB is concomitantly induced. Notably, an ssODN that introduced two separate point mutations and a 2-bp deletion resulted in up to 96% faithfully corrected chromosomal single-copy loci among the 0.3% cells having acquired EGFP fluorescence. ODN-directed modifications aiming at a 59-bp deletion or a 10-bp insertion at the target site were slightly less efficient, and the proportions of correctly modified sequences were lower. Concerning the mechanism, the ssODN served as an information template. This is in contrast to results from targeted gene repair in cells without DSB induction in which the ssODN was physically incorporated into the chromosome.

Influence of a defined DSB on ODN-directed gene modification

Earlier work has shown that, employing long double-stranded plasmid DNA, chromosomal DSBs induced with I-SceI were corrected at rates of 5% ²² and 10% ²³, and a DSB induced in the human IL2R gene by a pair of highly specifically designed zinc-finger nucleases (ZFNs) was corrected at rates of about 20% ²⁴. With 0.3% EGFP-positive cells among living populations, gene modification rates for ssODN-directed DSB repair were 20–70 times lower and, therefore, possibly useful in clinical applications only when precisely modified cells could gain a growth advantage. Reasons for the reduced efficiencies might be: (i) Long double-stranded DNA might reach effective concentrations in the cellular nucleus more efficiently than ssODNs. (ii) The number of DSBs induced during targeting, i.e., the total number of cells adequately expressing I-SceI, and the occurrence of repeated breaks at the site are unknown. Since DSB induction was achieved by cotransfection of an expression plasmid for I-SceI, defining the efficiency of induction would require the quantification of the expression profile of I-SceI (based on the plasmid's uptake and expression kinetics) in relation to the biological functionality and intracellular availability of the ssODNs. It is conceivable that—instead of cotransfection—a time window might be helpful, in which first the plasmid DNA would be applied, allowing I-SceI to reach effective concentrations when the ssODNs gain access to the target site. (iii) I-SceI might not be the DSB inducer of choice in the context of short ssODNs, because it strongly binds to the hydrolyzed DNA ends, which may cause persistence of the cleaved products ²⁵. How this complex influences the functionality of the ssODN is unclear. (iv) The structure of the modifying ODN could possibly be optimized. The ssODNs in this work were of sense orientation, had backbones with standard phosphodiester bonds, and were 80 or 96 nucleotides long. Thus, further experiments should evaluate whether antisense orientation, increased nuclease resistance, different lengths, and (partially) double-stranded topology might be advantageous.

Concerning the mechanism, earlier work has reported that purely ODN-mediated gene repair is achieved by physical incorporation of the correcting ODN into the targeted chromosomal site during replication ⁶. For the DSB-accompanied gene modification, the experiments showed that the ssODN was undetectable physically at the site of repair after imprinting its sequence information. Thus, ODN-mediated gene repair/modification—at least in HEK-293-derived systems—can proceed via different cellular pathways (Fig. 5A). Fig. 5B presents a hypothetical scheme of the repair pathway. It is based on annealing of the ssODN to a single-stranded region in the target region followed by several steps of new DNA synthesis involving also a template switch of the DNA polymerase. Thus, S-phase replication is not necessarily involved; however, local DNA synthesis seems to allow information transfer from the exogenously added ssODN to the mammalian genome.

Figure 5.

Mechanistic aspects of targeted gene modification. (A) Survey of various gene modification protocols. The flow chart contrasts the different fates of ssODNs during successful gene modification in the presence or absence of an intentionally induced DSB. (B) Hypothetical scheme for DSB repair in the presence of ssODNs. The modifying ssODN of sense orientation is symbolized by a short black line with a black triangle denoting the specific sequence alteration envisaged. The region to which the ssODN can hybridize is indicated by the vertical dotted lines. The long black and gray lines depict the nontranscribed and the transcribed strands of the target locus, respectively. Jagged open flash lines indicate the I-SceI cut positions. Newly synthesized DNA is symbolized by horizontal dotted lines.

Full figure and legend (199K)

Adverse effects during ODN-directed gene modification in the presence of a DSB

In this work, detection of gene modification events profited from the induction of a DSB close to the targeted mutation. Nevertheless, it must be born in mind that a DSB threatens the life of a cell, and, therefore, a multitude of responses are elicited to seal the broken chromosome ends. Free DNA ends are notorious for undergoing recombination ²⁶. More recent work with mouse fibroblasts utilizing I-SceI has revealed that virtually any extrachromosomal DNA molecule may be patched into DSBs in a mammalian genome ⁹. Thus, insertional mutagenesis is a concern, as had to be experienced in gene therapy trials employing a retroviral construct ^27,28. Furthermore, rAAV-based vectors, also under development for targeted gene repair protocols ²⁹, do integrate into chromosomal breakage sites ³⁰. Therefore, the behaviors of the DNA molecules involved in this work had been evaluated. Even though the majority of EGFP-positive cells carried faithfully modified loci, several types of unwanted sequence additions—most likely via microhomologies—were observed originating either from ssODN or from plasmid sequences. Due to the relatively short lengths, the ssODN-derived events might not have changed the endogenous expression of the nearby ORFs. However, dramatic insertional mutagenesis resulting in efficient EGFP expression independent of the upstream LacZ ORF occurred after patching of promoter sequences derived from the CMV IE promoter/enhancer region of cotransfected plasmid DNAs. These "promoter trap" events demand that if defined DSBs are intended in gene repair protocols, the respective nuclease should not be expressed from DNA constructs but shall be provided either by specific mRNA or by transfer of the protein itself. Here, fusion of the nucleases to membrane-translocating peptides might be useful ^31,32.

There have been earlier reports from Kmiec and co-workers that cellular responses induced by DNA-damaging agents enhance gene repair activity ^33,34. Focusing now on a defined DSB close to the targeted mutation, the work herein suggests that the DSB itself rather than a general up-regulation of DNA repair pathways might enhance targeted gene correction. However, this has not yet been formally proven, and therefore, an additional effect from a general up-regulation of pathways cannot be excluded.

Since I-SceI is useful only in engineered model systems, targeted gene repair aiming at clinical applications requires the development of new tools such as chimeric ZFNs which—theoretically—allow DSB induction at any given site in the mammalian genome ^23,24,35. The propensity of DNAs—be they of single- or double-stranded topology—to participate in DSB repair in the mammalian genome demands thorough analyses of possible side effects of targeted gene repair. Thus, DNA-based expression vectors encoding ZFNs should not be used. Additionally, since cells also suffer from "physiologic" DSBs, e.g., during replication fork movements, studies are needed to minimize or possibly inhibit insertional mutagenesis.

Materials and methods

Plasmids and oligodeoxynucleotides

Plasmid pRK5.LHA-Sce1 (5548 bp) allows expression of the I-SceI endonuclease under control of the CMV IE promoter/enhancer. It was generated by subcloning the PCR-amplified coding sequence of I-SceI contained on plasmid rAAV.Sce ¹³ into plasmid pRK5 ³⁶. Plasmid pCMV.LacZssGFP contains—under control of the CMV IE promoter/enhancer—the repair target locus consisting of the LacZ ORF followed by the EGFP ORF (with its own start codon in a Kozak consensus sequence). The EGFP ORF was translationally decoupled from the LacZ ORF by a spacer region harboring multiple stop codons and the 18-bp-long recognition site for I-SceI (Fig. 1). It was generated by subcloning the PCR-amplified coding sequence of LacZ from plasmid pCMV (Clontech) into plasmid pEGFP-N3 (Clontech). The spacer region was inserted as a double-stranded oligonucleotide into the intergenic BamHI site. Plasmid pcDNA6/myc-His A (Invitrogen Corp.) served as fill-in for the transfection mixes. Fig. 1 presents the main repair ODNs. Primary sequences and their respective positioning against the target locus are given in Figs. 1B, 1C, and 1D. Additionally, DSB-shift-38BdT carries as its residue 38 a thymidine labeled with biotin via an amino-modifier-C6 spacer. For negative control purposes, a scrambled sequence version of DSB-shift, the DSB-scramble, was synthesized (5'-AAATGCGGTCGGACCATGACCAATACGCCCAATATTTATCTGAGGGTACGAGTAACGCC
ACAGTTAGTTTCCTCTACGAA-3'). DSB-scramble does not target the transgene, and therefore, the LacZ ORF is assumed to remain unchanged (ORF: 3081 bp). DSB-scramble-37BdT was generated in analogy to ssODN DSB-shift-38BdT. All ssODNs for transfections were obtained from Thermo Electron Corp. They comprised at least 82% full-length product as measured by capillary electrophoresis. Functionality tests involving the labeled ODN DSB-shift-38BdT revealed gene repair rates roughly 50% of that of the unlabeled version DSB-shift. All ssODNs involved in transfection experiments were redissolved in LiChrosolv (Merck) to a working concentration of 0.5 g/l. The following primers were used in PCR analyses: gfpM, 5'-AGGTGTCGACGCGTATTCGCTAGCGCTACCGGTCGCsC-3' (s: phosphorothioate linkage); gfp7, 5'-GAAGTCGTGCTGCTTCATGTGG-3'; #48-LacZ-TC, 5'-TCAGCCGCTACAGTCAACAG-3' (Thermo Electron Corp.).

Cell lines

Cell clone 293LG32 carries a single-copy gene modification target locus (confirmed by Southern blot analysis; data not shown) and was generated by stably transfecting HEK-293 cells with plasmid pCMV.LacZssGFP, followed by selection in 0.5 mg/ml Geneticin (Gibco-Invitrogen Corp.). The parental cell line HEK-293 ^37,38, cultured as described previously ⁵, served as negative control. In the case of 293LG32 cells, growth medium was supplemented with 0.5 mg/ml Geneticin to select for the presence of the target locus. Cell line 293mEGFP-M12, providing an internal control for the bead capture experiments, has been described before ⁵.

Transfection

ODNs and plasmid DNAs were transfected following the calcium phosphate coprecipitation method described earlier ⁵. Generally, 1 10⁵ cells were seeded into a 24-well the day before transfection. Full transfection mixtures for ODN-directed DSB repair contained 0.03 g of pRK5.LHA-Sce1, 1.2 g of pcDNA6/myc-His A (as fill-in vector), and 0.5 g of the respective ssODN in 60 l of 1 precipitation buffer. Control transfections in 293LG32 cells with ODNs but omitting pRK5.LHA-Sce1 were set up with 1.2 g of pcDNA6/myc-His A and 0.5 g of the respective ssODN. After adding 8.6 l of 1 M CaCl₂, precipitates were formed for 20–45 min at room temperature. Cells were transfected about 22 h after seeding. For the time course experiment analyzing the numbers of gene-corrected cells during continuous expansion followed by sequence analyses, cell cultures and transfection mixtures were scaled up 5-fold and were applied two times in single 6-wells. To provide sufficient numbers of EGFP-positive cells for sorting in the context of the bead assay, each transfection was run in three T75 flasks, each receiving 45-fold amount of cells and transfection materials. Since here a mixture of two ssODNs was transfected, the total input amount of ssODNs was equally split between them. Contamination with EGFP-encoding plasmids leading to false-positive FACS measurements was precluded by adding an aliquot of each transfection mixture to a single 24-well culture of parental HEK-293 cells. No or at most single EGFP-positive events were detected among 5 10⁴ events (data not shown).

FACS analyses

Three days after transfection, samples were analyzed on a FACSort flow cytometer (Becton–Dickinson (BD)) as described before ⁵. Gene modification rates were measured as the percentages of 7-AAD-negative, i.e., living, EGFP-positive cells among 5 10⁴ cells. They were obtained by running each transfection setup in duplicate three times independently. The data of the duplicates were transformed into arithmetic means to even out variations in pipetting and cell seeding. From these three values, again the arithmetic means were calculated and plotted standard deviations. Since transfection controls showed that at least 50% of the treated cells received DNA (data not shown), no normalization according to transfection efficiencies was done.

Cell sorting

Cells were sorted with a BD FACSAria cell sorter (BD Biosciences) for two different purposes: (i) DNA sequence analyses of transgene loci derived from living, EGFP-positive cells (taken from the time-course experiment),or (ii) analysis of the fate of transfected ssODNs in living, EGFP-positive and EGFP-negative cells with a bead-capture assay. Sample preparations and FACS equipment have been described earlier ⁶.

For (i), two independent experimental runs (I, II) were carried out. About 2 weeks after transfection, cells were sorted to final purity of 97% (data not shown). These materials were taken to collect eight aliquots of 500 cells from each transfection.

For (ii), two independent experiments were performed and cells were collected 3 days after transfection. For each experimental setting, cells of three T75-flasks had been combined for sorting. Final purity reached 80.3–96.8% for the EGFP-positive fractions (containing 7 10³ to 2.2 10⁴ cells; data not shown) and 99.99% for the EGFP-negative cells (containing about 2 10⁵ cells; data not shown).

Sequence analyses

Sorted cells in 3 l of PBS (Invitrogen Corp.) were lysed by adding 1 l of lysis buffer ³⁹ containing 1:10 diluted proteinase K (Sigma; stock concentration 25.1 mg protein/ml). The reaction was run at 56°C for 1 h. Proteinase K was inactivated by incubation at 96°C for 15 min. These lysates were used as template input for PCR amplifications. Fragments (roughly 500 bp) spanning the DSB region (LacZ 3' terminus–spacer–EGFP–5' terminus) were generated with Pfu polymerase and primers #48-LacZ-TC and gfp7 for 45 cycles according to the supplier's instructions. PCR products were digested with I-SceI (New England Biolabs). Furthermore, they were subcloned with the TOPO cloning kit (Invitrogen Corp.). Isolation of bacterial clones followed—with the exception of sample "no ODN"—the blue/white selection scheme. Sequence analyses were carried out with primer gfp7 using the BigDye Terminator v1.1 cycle sequencing kit (Applied Biosystems) and an ABI Prism 3100 genetic analyzer (Applied Biosystems).

Preparation of gDNA for capture experiments

Sorted cell sediments were used to isolate gDNA using the Wizard Genomic DNA Purification Kit (Promega Corp.). Some modifications compared to the supplier's instructions were employed as described ⁶.

Capture assay and analysis of biotinylated DNA fragments

The basic procedure to capture biotinylated DNA fragments with Dynabeads M-280 streptavidin (Dynal Biotech) was performed as described ⁶ with adaptation for the new model cell system: An internal positive control in the form of a biotinylated test fragment was added to each reaction to monitor capturing and multiplex PCR amplification.

293LG32 cells had been transfected with mixtures of DSB-shift and DSB-scramble (with or without biotin label) and sorted into EGFP-positive and EGFP-negative fractions. Genomic DNAs were isolated and digested with AflIII and BsrGI yielding a 1067-bp-long target locus-specific fragment. Completeness of digestion was validated by running in parallel an aliquot of the gDNA spiked with 100 ng of a control plasmid. Digested gDNA was purified with the QIAquick PCR purification kit (Qiagen), and recovery was controlled by a 35-cycle PCR (data not shown). Digested gDNA (4.5 ng, corresponding to 500 cells) was mixed with 500 molecules of the biotinylated molecule IC ⁶, the internal control specific for the unrelated multicopy transgene locus of 293mEGFP-M12 cells. DNA mixtures were incubated with 12.5 g of streptavidin-coated magnetic beads overnight at 4°C. Captured DNA fragments were coamplified by a multiplex 35-cycle PCR involving Taq DNA polymerase in the presence of 1 Q-solution (Qiagen) and the common reverse primer gfp7. The forward primers #48-LacZ-TC and gfpM were specific for the 293LG32 target locus and the 293mEGFP-M12-derived IC, respectively. Five microliters of the 50-l PCRs was run on an agarose gel. Test fragment TF (Fig. 4B) served as positive control for amplifying the targeted region of the 293LG32 locus. It was generated by running a 1-cycle PCR with the forward primer DSB-shift-38BdT on a template made by PCR with primers #48-LacZ-TC and gfp7.

References

1. Kalish, J. M. and Glazer, P. M. (2005). Targeted genome modification via triple helix formation. Ann. N.Y. Acad. Sci. 1058: 151–161. | Article | PubMed | ChemPort |
2. Knauert, M. P. and Glazer, P. M. (2001). Triplex forming oligonucleotides: sequence-specific tools for gene targeting. Hum. Mol. Genet. 10: 2243–2251. | Article | PubMed | ISI | ChemPort |
3. Andersen, M. S., Sorensen, C. B., Bolund, L. and Jensen, T. G. (2002). Mechanisms underlying targeted gene correction using chimeric RNA/DNA and single-stranded DNA oligonucleotides. J. Mol. Med. 80: 770–781. | Article | PubMed | ISI | ChemPort |
4. Igoucheva, O., Alexeev, V. and Yoon, K. (2004). Mechanism of gene repair open for discussion. Oligonucleotides 14: 311–321. | Article | PubMed | ISI | ChemPort |
5. Radecke, F., Radecke, S. and Schwarz, K. (2004). Unmodified oligodeoxynucleotides require single-strandedness to induce targeted repair of a chromosomal EGFP gene. J. Gene Med. 6: 1257–1271. | Article | PubMed | ChemPort |
6. Radecke, S., Radecke, F., Peter, I. and Schwarz, K. (2005). Physical incorporation of a single-stranded oligodeoxynucleotide during targeted repair of a human chromosomal locus. J. Gene Med. 8: 217–228.
7. Petrini, J. H. and Stracker, T. H. (2003). The cellular response to DNA double-strand breaks: defining the sensors and mediators. Trends Cell Biol. 13: 458–462. | Article | PubMed | ISI | ChemPort |
8. Lieber, M. R., Ma, Y., Pannicke, U. and Schwarz, K. (2003). Mechanism and regulation of human non-homologous DNA end-joining. Nat. Rev. Mol. Cell Biol. 4: 712–720. | Article | PubMed | ISI | ChemPort |
9. Lin, Y. and Waldman, A. S. (2001). Promiscuous patching of broken chromosomes in mammalian cells with extrachromosomal DNA. Nucleic Acids Res. 29: 3975–3981. | PubMed | ChemPort |
10. Lin, Y. and Waldman, A. S. (2001). Capture of DNA sequences at double-strand breaks in mammalian chromosomes. Genetics 158: 1665–1674. | PubMed | ISI | ChemPort |
11. Rouet, P., Smih, F. and Jasin, M. (1994). Introduction of double-strand breaks into the genome of mouse cells by expression of a rare-cutting endonuclease. Mol. Cell. Biol. 14: 8096–8106. | PubMed | ISI | ChemPort |
12. Johnson, R. D. and Jasin, M. (2001). Double-strand-break-induced homologous recombination in mammalian cells. Biochem. Soc. Trans. 29: 196–201. | Article | PubMed | ISI | ChemPort |
13. Porteus, M. H., Cathomen, T., Weitzman, M. D. and Baltimore, D. (2003). Efficient gene targeting mediated by adeno-associated virus and DNA double-strand breaks. Mol. Cell. Biol. 23: 3558–3565. | Article | PubMed | ISI | ChemPort |
14. Miller, D. G., Petek, L. M. and Russell, D. W. (2003). Human gene targeting by adeno-associated virus vectors is enhanced by DNA double-strand breaks. Mol. Cell. Biol. 23: 3550–3557. | Article | PubMed | ISI | ChemPort |
15. Mandecki, W. (1986). Oligonucleotide-directed double-strand break repair in plasmids of Escherichia coli: a method for site-specific mutagenesis. Proc. Natl. Acad. Sci. USA 83: 7177–7181. | Article | PubMed | ChemPort |
16. Storici, F., Durham, C. L., Gordenin, D. A. and Resnick, M. A. (2003). Chromosomal site-specific double-strand breaks are efficiently targeted for repair by oligonucleotides in yeast. Proc. Natl. Acad. Sci. USA 100: 14994–14999. | Article | PubMed | ChemPort |
17. Banga, S. S. and Boyd, J. B. (1992). Oligonucleotide-directed site-specific mutagenesis in Drosophila melanogaster. Proc. Natl. Acad. Sci. USA 89: 1735–1739. | Article | PubMed | ChemPort |
18. Dekker, M., Brouwers, C. and te Riele, H. (2003). Targeted gene modification in mismatch-repair-deficient embryonic stem cells by single-stranded DNA oligonucleotides. Nucleic Acids Res. 31: E27. | Article | PubMed | ChemPort |
19. Olsen, P. A., Randol, M. and Krauss, S. (2005). Implications of cell cycle progression on functional sequence correction by short single-stranded DNA oligonucleotides. Gene Ther. 12: 546–551. | Article | PubMed | ChemPort |
20. Brachman, E. E. and Kmiec, E. B. (2005). Gene repair in mammalian cells is stimulated by the elongation of S phase and transient stalling of replication forks. DNA Repair (Amsterdam) 4: 445–457. | ChemPort |
21. Wu, X. S., et al. (2005). Increased efficiency of oligonucleotide-mediated gene repair through slowing replication fork progression. Proc. Natl. Acad. Sci. USA 102: 2508–2513. | Article | PubMed | ChemPort |
22. Porteus, M. H. and Baltimore, D. (2003). Chimeric nucleases stimulate gene targeting in human cells. Science 300: 763. | Article | PubMed | ISI |
23. Alwin, S., et al. (2005). Custom zinc-finger nucleases for use in human cells. Mol. Ther. 12: 610–617. | Article | PubMed | ChemPort |
24. Urnov, F. D., et al. (2005). Highly efficient endogenous human gene correction using designed zinc-finger nucleases. Nature 435: 646–651. | Article | PubMed | ChemPort |
25. Kuhne, C., Tjornhammar, M. L., Pongor, S., Banks, L. and Simoncsits, A. (2003). Repair of a minimal DNA double-strand break by NHEJ requires DNA-PKcs and is controlled by the ATM/ATR checkpoint. Nucleic Acids Res. 31: 7227–7237. | Article | PubMed | ISI | ChemPort |
26. Orr-Weaver, T. L., Szostak, J. W. and Rothstein, R. J. (1981). Yeast transformation: a model system for the study of recombination. Proc. Natl. Acad. Sci. USA 78: 6354–6358. | Article | PubMed | ChemPort |
27. Hacein-Bey-Abina, S., et al. (2003). A serious adverse event after successful gene therapy for X-linked severe combined immunodeficiency. N. Engl. J. Med. 348: 255–256. | Article | PubMed | ISI |
28. Hacein-Bey-Abina, S., et al. (2003). LMO2-associated clonal T cell proliferation in two patients after gene therapy for SCID-X1. Science 302: 415–419. | Article | PubMed | ISI | ChemPort |
29. Cathomen, T. (2004). AAV vectors for gene correction. Curr. Opin. Mol. Ther. 6: 360–366. | PubMed | ChemPort |
30. Miller, D. G., Petek, L. M. and Russell, D. W. (2004). Adeno-associated virus vectors integrate at chromosome breakage sites. Nat. Genet. 36: 767–773. | Article | PubMed | ISI | ChemPort |
31. Brooks, H., Lebleu, B. and Vives, E. (2005). Tat peptide-mediated cellular delivery: back to basics. Adv. Drug Delivery Rev. 57: 559–577. | Article | ChemPort |
32. Lindsay, M. A. (2002). Peptide-mediated cell delivery: application in protein target validation. Curr. Opin. Pharmacol. 2: 587–594. | Article | PubMed | ISI | ChemPort |
33. Ferrara, L. and Kmiec, E. B. (2004). Camptothecin enhances the frequency of oligonucleotide-directed gene repair in mammalian cells by inducing DNA damage and activating homologous recombination. Nucleic Acids Res. 32: 5239–5248. | Article | PubMed | ChemPort |
34. Ferrara, L., Parekh-Olmedo, H. and Kmiec, E. B. (2004). Enhanced oligonucleotide-directed gene targeting in mammalian cells following treatment with DNA damaging agents. Exp. Cell Res. 300: 170–179. | Article | PubMed | ChemPort |
35. Porteus, M. H. (2006). Mammalian gene targeting with designed zinc finger nucleases. Mol. Ther. 13: 438–446. | Article | PubMed | ISI | ChemPort |
36. Cathomen, T. and Weitzman, M. D. (2000). A functional complex of adenovirus proteins E1B-55kDa and E4orf6 is necessary to modulate the expression level of p53 but not its transcriptional activity. J. Virol. 74: 11407–11412. | Article | PubMed | ISI | ChemPort |
37. Shaw, G., Morse, S., Ararat, M. and Graham, F. L. (2002). Preferential transformation of human neuronal cells by human adenoviruses and the origin of HEK 293 cells. FASEB J. 16: 869–871. | PubMed | ChemPort |
38. Toth, K. and Wold, W. S. (2002). HEK? No!. Mol. Ther. 5: 654. | Article | PubMed | ChemPort |
39. Hopert, A., Uphoff, C. C., Wirth, M., Hauser, H. and Drexler, H. G. (1993). Specificity and sensitivity of polymerase chain reaction (PCR) in comparison with other methods for the detection of mycoplasma contamination in cell lines. J. Immunol. Methods 164: 91–100. | Article | PubMed | ChemPort |
40. Colleaux, L., D'Auriol, L., Galibert, F. and Dujon, B. (1988). Recognition and cleavage site of the intron-encoded omega transposase. Proc. Natl. Acad. Sci. USA 85: 6022–6026. | Article | PubMed | ChemPort |

Appendices

Appendix A

Supplementary data

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ymthe.2004.02.011.

Acknowledgements

We thank Katja Heinrich, Tatjana Kersten, and Eva-Maria Rump for their assistance in sequencing. This work was supported by the Institute for Clinical Transfusion Medicine and Immunogenetics Ulm and by Grant CA311/1 from the German Research Foundation (T.C.).