Targeted mutagenesis in mammalian cells mediated by intracellular triple helix formation.

As an alternative to standard gene transfer techniques for genetic manipulation, we have investigated the use of triple helix-forming oligonucleotides to target mutations to selected genes within mammalian cells. By treating monkey COS cells with oligonucleotides linked to psoralen, we have generated targeted mutations in a simian virus 40 (SV40) vector contained within the cells via intracellular triple helix formation. Oligonucleotide entry into the cells and sequence-specific triplex formation within the SV40 DNA deliver the psoralen to the targeted site. Photoactivation of the psoralen by long-wavelength UV light yields adducts and thereby mutations at that site. We engineered into the SV40 vector novel supF mutation reporter genes containing modified polypurine sites amenable to triplex formation. By comparing the abilities of a series of oligonucleotides to target these new sites, we show that targeted mutagenesis in vivo depends on the strength and specificity of the third-strand binding. Oligonucleotides with weak target site binding affinity or with only partial target site homology were ineffective at inducing mutations in the SV40 vectors within the COS cells. We also show that the targeted mutagenesis is dependent on the oligonucleotide concentration and is influenced by the timing of the oligonucleotide treatment and of the UV irradiation of the cells. Frequencies of intracellular targeted mutagenesis in the range of 1 to 2% were observed, depending upon the conditions of the experiment. DNA sequence analysis revealed that most of the mutations were T.A-to-A.T transversions precisely at the targeted psoralen intercalation site. Several deletions encompassing that site were also seen. The ability to target mutations to selected sites within mammalian cells by using modified triplex-forming oligonucleotides may provide a new research tool and may eventually lead to therapeutic applications.     

Mutagenic SOS repair of site-specific psoralen damage in plasmid pBR322

The mutagenic repair of psoralen damage was examined by transforming Escherichia coli with psoralen-treated pBR322. Plasmid DNA randomly reacted with psoralen was repaired only when the E. coli was uvrA⁺ and recA⁺, and only when the cells were pre-irradiated with far-ultraviolet light. The recA dependence and requirement for pre-irradiation are characteristics of SOS repair.Psoralens were placed specifically near the BamHI site, in the tetracycline-resistance gene of pBR322, using a sulfhydryl-containing psoralen derivative. Repair of this damage also required pre-irradiation of the host cells. This repair was accompanied by a 4% frequency of mutagenesis to a tetraeycline-sensitive phenotype. Sequence analysis of these mutant plasmids revealed that 75% had mutations within the targeted region, while 25% had no sequence changes within 100 bases of the BamHI site. In up to five independent isolates only one kind of mutation was observed at each site, suggesting that mutagenic SOS repair is influenced by DNA structure at the site of the psoralen. Most mutations were transitions, primarily G-C to A-T changes. Some transitions occurred at sites where psoralen crosslinks could not have formed, and these may have arisen from the repair of psoralen monoadducts.     

Suppressible base substitution mutations induced by angelicin (isopsoralen) in the Escherichia coli lacI gene: implications for the mechanism of SOS mutagenesis.

Angelicin- plus near-UV-induced mutations were umuC dependent in Escherichia coli K-12. Angelicin, a monofunctional psoralen derivative, is believed to damage DNA almost exclusively at pyrimidine bases. To broaden our knowledge about the mutagenic specificity of SOS-dependent mutagens, we determined the mutational specificity of 233 suppressible lacI mutations induced by angelicin. More than 90% of the nonsense mutations arose via transversion substitutions. The three most frequently mutated sites were at A-T base pairs and accounted for more than one-third of all induced nonsense mutations. The two hottest sites were at the only occurrences of the 5'-TATA-3' tetranucleotide in lacI, a sequence expected to be a preferred binding site for a psoralen. Both A-T-to-T-A and A-T-to-C-G transversions were well induced by angelicin treatment, but the frequency of each transversion depended on the particular site. We also detected significant induction of transversion mutations at G-C sites. The induction of transversions by an SOS-dependent mutagen that generates lesions at pyrimidines supports the idea that DNA lesions influence the selection of bases that are incorporated via the process of SOS repair.     

Processing of targeted psoralen cross-links in Xenopus oocytes.

Psoralen cross-links have been shown to be both mutagenic and recombinagenic in bacterial, yeast, and mammalian cells. Double-strand breaks (DSBs) have been implicated as intermediates in the removal of psoralen cross-links. Recent work has suggested that site-specific mutagenesis and recombination might be achieved through the use of targeted psoralen adducts. The fate of plasmids containing psoralen adducts was evaluated in Xenopus oocytes, an experimental system that has well-characterized recombination capabilities and advantages in the analysis of intermediates in DNA metabolism. Psoralen adducts were delivered to a specific site by a triplex-forming oligonucleotide. These lesions are clearly recognized and processed in oocytes, since mutagenesis was observed at the target site. The spectrum of induced mutations was compared with that found in similar studies in mammalian cells. Plasmids carrying multiple random adducts were preferentially degraded, perhaps due to the introduction of DSBs. However, when DNAs carrying site-specific adducts were examined, no plasmid loss was observed and removal of cross-links was found to be very slow. Sensitive assays for DSB-dependent homologous recombination were performed with substrates with one or two cross-link sites. No adduct-stimulated recombination was observed with a single lesion, and only very low levels were observed with paired lesions, even when a large proportion of the cross-links was removed by the oocytes. We conclude that DSBs or other recombinagenic structures are not efficiently formed at psoralen adducts in Xenopus oocytes. While psoralen is not a promising reagent for stimulating site-specific recombination, it is effective in inducing targeted mutations.     

Repairing the Sickle Cell mutation. II. Effect of psoralen linker length on specificity of formation and yield of third strand-directed photoproducts with the mutant target sequence

Three identical deoxyoligonucleotide third strands with a 3′-terminal psoralen moiety attached by linkers that differ in length (N = 16, 6 and 4 atoms) and structure were examined for their ability to form triplex-directed psoralen photoproducts with both the mutant T residue of the Sickle Cell β-globin gene and the comparable wild-type sequence in linear duplex targets. Specificity and yield of UVA (365 nm) and visible (419 nm) light-induced photoadducts were studied. The total photoproduct yield varies with the linker and includes both monoadducts and crosslinks at various available pyrimidine sites. The specificity of photoadduct formation at the desired mutant T residue site was greatly improved by shortening the psoralen linker. In particular, using the N-4 linker, psoralen interaction with the residues of the non-coding duplex strand was essentially eliminated, while modification of the Sickle Cell mutant T residue was maximized. At the same time, the proportion of crosslink formation at the mutant T residue upon UV irradiation was much greater for the N-4 linker. The photoproducts formed with the wild-type target were fully consistent with its single base pair difference. The third strand with the N-4 linker was also shown to bind to a supercoiled plasmid containing the Sickle Cell mutation site, giving photoproduct yields comparable with those observed in the linear mutant target.     

Covalent crosslinks introduced via a triple helix-forming oligonucleotide coupled to psoralen are inefficiently repaired.

Triple helix-forming oligonucleotides (TFOs) represent potentially powerful tools to artificially modulate gene activity. In particular, they can be used to specifically introduce a lesion into a selected target sequence: interstrand crosslinks and monoadducts can be introduced via TFOs coupled to psoralen. The efficiency of these strategies depends on the cell ability to repair these lesions, an issue which is still controversial. Here we show, using psoralen-coupled TFOs and the yeast as a convenient cellular test system, that interstrand crosslinks are quantitatively poorly repaired, resulting in an efficient modification of target gene activity. In addition, these lesions result in the introduction of mutations in a high proportion of cells. We show that these mutations are generated by the Error-Prone Repair pathway, alone or in combination with Nucleotide Excision Repair. Taken together, these results suggest that TFOs coupled to psoralen could be used to inactivate a gene with significant efficiency.     

Psoralen photo-cross-linking by triplex-forming oligonucleotides at multiple sites in the human rhodopsin gene.

Targeting DNA damage by triplex-forming oligonucleotides (TFOs) represents a way of modifying gene expression and structure and a possible approach to gene therapy. We have determined that this approach can deliver damage with great specificity to sites in the human gene for the G-protein-linked receptor rhodopsin, mutations of which can lead to the genetic disorder autosomal dominant retinitis pigmentosa. We have introduced DNA monoadducts and interstrand cross-links at multiple target sites within the gene using TFOs with a photoactivatable psoralen group at the 5'-end. The extent of formation of photoadducts (i.e., monoadducts and cross-links) was measured at target sites with a 5'-ApT sequence at the triplex-duplex junction and at a target site with 5'-ApT and 5'-TpA sequences located four and seven nucleotides away, respectively. To improve psoralen reactivity at more distant sites, psoralen moieties were attached to TFOs with nucleotide "linkers" from two to nine nucleotides in length. High-affinity binding was maintained with linkers of up to 10 nucleotides, but affinities tended to decrease somewhat with increasing linker length due to faster dissociation kinetics. DNase I footprinting indicated little, if any, interaction between linkers and the duplex. Psoralen-TFO conjugates formed DNA cross-links with high efficiency (56-65%) at 5'-ApT sequences located at triplex junctions. At a 5'-ApT site four nucleotides away, the efficiency varied with linker length; a four-nucleotide linker gave the highest efficiency. Duplexes with 5'-TpA and 5'-ApT sites two nucleotides away, in otherwise identical sequences, were cross-linked with efficiencies of 56 and 38%, respectively. These results indicate that TFO-linker-psoralen conjugates allow simultaneous, efficient targeting of multiple sites in the human rhodopsin gene.     

Recombinational and mutagenic repair of psoralen interstrand cross-links in Saccharomyces cerevisiae

Psoralen photoreacts with DNA to form interstrand cross-links, which can be repaired by both nonmutagenic nucleotide excision repair and recombinational repair pathways and by mutagenic pathways. In the yeast Saccharomyces cerevisiae, psoralen cross-links are processed by nucleotide excision repair to form double-strand breaks (DSBs). In yeast, DSBs are repaired primarily by homologous recombination, predicting that cross-link and DSB repair should induce similar recombination end points. We compared psoralen cross-link, psoralen monoadduct, and DSB repair using plasmid substrates with site-specific lesions and measured the patterns of gene conversion, crossing over, and targeted mutation. Psoralen cross-links induced both recombination and mutations, whereas DSBs induced only recombination, and monoadducts were neither recombinogenic nor mutagenic. Although the cross-link- and DSB-induced patterns of plasmid integration and gene conversion were similar in most respects, they showed opposite asymmetries in their unidirectional conversion tracts: primarily upstream from the damage site for cross-links but downstream for DSBs. Cross-links induced targeted mutations in 5% of the repaired plasmids; all were base substitutions, primarily T --> C transitions. The major pathway of psoralen cross-link repair in yeast is error-free and involves the formation of DSB intermediates followed by homologous recombination. A fraction of the cross-links enter an error-prone pathway, resulting in mutations at the damage site.     

Mutagenic potential of DNA-peptide crosslinks mediated by acrolein-derived DNA adducts

Current data suggest that DNA-peptide crosslinks are formed in cellular DNA as likely intermediates in the repair of DNA-protein crosslinks. In addition, a number of naturally occurring peptides are known to efficiently conjugate with DNA, particularly through the formation of Schiff-base complexes at aldehydic DNA adducts and abasic DNA sites. Since the potential role of DNA-peptide crosslinks in promoting mutagenesis is not well elucidated, here we report on the mutagenic properties of Schiff-base-mediated DNA-peptide crosslinks in mammalian cells. Site-specific DNA-peptide crosslinks were generated by covalently trapping a lysine-tryptophan-lysine-lysine peptide to the N(6) position of deoxyadenosine (dA) or the N(2) position of deoxyguanosine (dG) via the aldehydic forms of acrolein-derived DNA adducts (gamma-hydroxypropano-dA or gamma-hydroxypropano-dG, respectively). In order to evaluate the potential of DNA-peptide crosslinks to promote mutagenesis, we inserted the modified oligodeoxynucleotides into a single-stranded pMS2 shuttle vector, replicated these vectors in simian kidney (COS-7) cells and tested the progeny DNAs for mutations. Mutagenic analyses revealed that at the site of modification, the gamma-hydroxypropano-dA-mediated crosslink induced mutations at only approximately 0.4%. In contrast, replication bypass of the gamma-hydroxypropano-dG-mediated crosslink resulted in mutations at the site of modification at an overall frequency of approximately 8.4%. Among the types of mutations observed, single base substitutions were most common, with a prevalence of G to T transversions. Interestingly, while covalent attachment of lysine-tryptophan-lysine-lysine at gamma-hydroxypropano-dG caused an increase in mutation frequencies relative to gamma-hydroxypropano-dG, similar modification of gamma-hydroxypropano-dA resulted in decreased levels of mutations. Thus, certain DNA-peptide crosslinks can be mutagenic, and their potential to cause mutations depends on the site of peptide attachment. We propose that in order to avoid error-prone replication, proteolytic degradation of proteins covalently attached to DNA and subsequent steps of DNA repair should be tightly coordinated.     

Site-specific targeting of psoralen photoadducts with a triple helix-forming oligonucleotide: characterization of psoralen monoadduct and crosslink formation.

A polypurine tract in the supF gene of bacteriophage lambda (base pairs 167-176) was selected as the target for triple helix formation and targeted mutagenesis by an oligopurine (5'-AGGAAGGGGG-3') containing a chemically linked psoralen derivative (4'-hydroxymethyl-4,5',8-trimethylpsoralen) at its 5' terminus (psoAG10). The thymines at base pairs 166 and 167, a 5'ApT site, were targeted for photomodification. Exposure of the triple helical complex to long wavelength ultraviolet radiation led to the covalent binding of psoAG10 to the targeted region in the supF gene and to the induction of site-specific mutations. We report here experiments to characterize the photomodification of the targeted region of the supF gene in the context of triple helix formation. An electrophoretic mobility-shift assay showed that, at low radiation doses, monoadducts at base pair 166 were the major photoadducts. At higher doses the monoadducts were converted to crosslinks between base pairs 166 and 167. HPLC analysis of enzymatically hydrolyzed photoreaction mixtures was used to confirm the electrophoresis results. A strong strand preference for specific photoadduct formation was also detected.     

Structure of the Escherichia coli 16 S ribosomal RNA. Psoralen crosslinks and N-acetyl-N'-(p-glyoxylylbenzoyl)cystamine crosslinks detected by electron microscopy

Escherichia coli 16 S ribosomal RNA in reconstitution buffer has been photochemically crosslinked with aminomethyltrimethylpsoralen and chemically crosslinked with N-acetyl-N'-(p-glyoxylylbenzoyl)cystamine. The positions of crosslinking have been detected by viewing the molecules in the electron microscope. DNA restriction fragments that contain psoralen mono-adducts were hybridized and crosslinked to the samples so that the orientations of the crosslinked molecules were seen directly. A two-dimensional histogram method has been used to classify the different types of looped crosslinked molecules. These methods allow the identification of 13 distinct types of loops in the photochemically crosslinked molecules and 31 distinct types of loops in the chemically crosslinked molecules. The psoralen experiments are a reinvestigation of some of our earlier results. Some of the crosslinks were previously reported in the incorrect orientation; with the corrected orientation, seven of the psoralen crosslinks can now be correlated with complementarities in the proposed secondary-structure models. However, there are still six other psoralen crosslinks that indicate additional contacts not found in the current models. The chemical crosslinks indicate pairs of single-stranded regions that must be close in the folded molecule. Many of these crosslinks occur between regions that are distant in the secondary structure; these crosslinks indicate part of the three-dimensional form of the folded molecule.     

IIIegitimate integration of non-replicative vectors in the genome of Rhodococcus fascians upon electro-transformation as an insertional mutagenesis system

Electrotransformation of Rhodococcus fascians by non-replicating plasmids containing a suitable resistance marker resulted in stable transformants by integration of these constructs at various sites in the genome, thereby generating different mutations. Tagged genes could be isolated in Escherichia coli owing to the presence of a CoIE1 replicon and an ampicillin resistance gene in the inserted sequences. Southern analysis and nucleotide sequencing revealed that recombination can occur at defined locations in the plasmid, while no site preference for target sequences could be detected. Low homology between the recombining sequences indicates illegitimate recombination. The specificity of the plasmid sites could be explained by assuming a linear recombination intermediate, generated by cleavage of the transformed plasmid.     

New mammalian cellular systems to study mutations introduced at the break site by non-homologous end-joining

The non-homologous end-joining (NHEJ) pathway is a mechanism to repair DNA double strand breaks, which can introduce mutations at repair sites. We constructed new cellular systems to specifically analyze sequence modifications occurring at the repair site. In particular, we looked for the presence of telomeric repeats at the repair junctions, since our previous work indicated that telomeric sequences could be inserted at break sites in germ-line cells during primate evolution. To induce specific DNA breaks, we used the I-SceI system of Saccharomyces cerevisiae or digestion with restriction enzymes. We isolated human and hamster cell lines containing the I-SceI target site integrated in a single chromosomal locus and we exposed the cells to a continuous expression of the I-SceI endonuclease gene. Additionally, we isolated human cell lines that expressed constitutively the I-SceI endonuclease and we introduced the target site on an episomal plasmid stably transfected into the cells. These strategies allowed us to recover repair junctions in which the I-SceI target site was modified at high frequency (100% in hamster cells and about 70% in human cells). Finally, we analyzed junctions produced on an episomal plasmid linearized by restriction enzymes. In all the systems studied, sequence analysis of individual repair junctions showed that deletions were the most frequent modifications, being present in more than 80% of the junctions. On the episomal plasmids, the average deletion length was greater than at intrachromosomal sites. Insertions of nucleotides or deletions associated with insertions were rare events. Junction organization suggested different mechanisms of formation. To check for the insertion of telomeric sequences, we screened plasmid libraries representing about 3.5 x 10(5) junctions with a telomeric repeat probe. No positive clones were detected, suggesting that the addition of telomeric sequences during double strand break repair in somatic cells in culture is either a very rare event or does not occur at all.     

Repairing the Sickle Cell mutation. III. Effect of irradiation wavelength on the specificity and type of photoproduct formed by a 3'-terminal psoralen on a third strand directed to the mutant base pair

Using a psoralen delivery system mediated by a DNA third strand that binds selectively to linear target duplexes immediately downstream from the Sickle Cell β-globin gene mutation and the comparable wild-type β-globin gene sequence, the kinetics of formation and yield of psoralen monoadducts and crosslinks with pyrimidine residues at and near the mutant base pair site and its wild-type counterpart were determined. By exploiting irradiation specificities at 300, 365 and 419 nm, it was possible to evaluate the orientation equilibrium of 3′-linked intercalated psoralen and to develop conditions that lead to preferential formation of each type of photoproduct in both the mutant and wild-type sequences. This makes possible the preparation of each type of photoproduct for use as a substrate for DNA repair. In this way, the base pair change(s) that each generates can be established.     

Site-specific gene modification by PNAs conjugated to psoralen

DNA-binding molecules, including triplex-forming oligonucleotides (TFOs) and peptide nucleic acids (PNAs), can be utilized to introduce site-specific mutations or to promote recombination at selected genomic sites. To further evaluate the utility of PNAs for site-specific gene modification, we tested dimeric bis-PNAs conjugated to psoralen. These PNAs are designed to form a triplex-invasion complex within the supF reporter gene in an episomal shuttle vector and to direct site-specific photoadduct formation by the conjugated psoralen. The psoralen-bis-PNA conjugate was found to direct photoadduct formation to the intended 5'-TpA base step next to the PNA-binding site, and the photoadduct formation efficiency displayed both concentration and UVA irradiation dependence. The effect of PNA-targeted photoadducts in a mammalian system was tested by SV40-based shuttle vector assay. After in vitro binding, we found that photoadducts directed by PNAs conjugated to psoralen-induced mutations at frequencies in the range of 0.46%, 6.5-fold above the background. In a protocol for intracellular gene targeting in the episomal shuttle vector, the psoralen-PNA-induced mutation frequency was 0.13%, 3.5-fold higher than the background. Most of the induced mutations were deletions and single-base-pair substitutions at or adjacent to the targeted PNA-binding and photoadduct-formation sites. When the results are taken together, they demonstrate the ability of bis-PNAs conjugated with psoralen to mediate site-specific gene modification, and they further support the development of PNAs as tools for gene-targeting applications.     

Characteristics of triplex-directed photoadduct formation by psoralen-linked oligodeoxynucleotides.

A triplex-forming oligopyrimidine has been attached at its 5'-end to a photoreactive psoralen derivative and used to target a sequence which forms part of the coding region of the human aromatase gene. The 20 base pair sequence is not a perfect triplex target since it contains three pyrimidine interruptions within the purine-rich strand. Despite this, we have detected triplex-directed photoadduct formation at pH 7.0 between the psoralen-linked oligonucleotide and a 30mer duplex representing the aromatase target. Photoadduct formation was found to be sensitive to pH, temperature, cation concentration and the base composition of the third strand. By varying the base sequence of the target duplex around the psoralen intercalation site, we have characterised the site and mode of psoralen intercalation. The attached psoralen has been found to intercalate at the triplex-duplex junction with a strong preference for one orientation. We have shown that the psoralen will bind at the junction even when there is a preferred TpA step at an adjacent site. We have also compared the binding affinity and photoreactivity of oligodeoxyribonucleotides linked to two different psoralen derivatives and found differences in the rate of crosslinking and the extent of crosslink formation. Finally, we have examined oligodeoxyribonucleotides which are attached to psoralen by polymethylene linkers of different lengths.