ABSTRACT
We compare two techniques which enable selective, nucleotide-specific covalent modification of human genomic DNA, as assayed by quantitative ligation- mediated PCR. In the first, a purine motif triplex-forming oligonucleotide with a terminally appended chlorambucil was shown to label a target guanine residue adjacent to its binding site in 80% efficiency at 0.5 [mu]M. Efficiency was higher in the presence of the triplex-stabilizing intercalator coralyne. In the second method, an oligonucleotide targeting a site containing all four bases and bearing chlorambucil on an interior base was shown to efficiently react with a specific nucleotide in the target sequence. The targeted sequence in these cases was in the DQ[beta]1*0302 allele of the MHC II locus.
Site-directed mutagenesis by specific modification of genomic DNA offers a viable alternative to vector-based gene therapy in cases where single base changes are useful. We have been developing methods for DNA sequence alteration using oligodeoxynucleotides (ODNs) with reactive agents attached to one or more positions. These may cause mutation at the targeted site, induced by covalent modification (1 ,2 ), resulting in inactivation or modification of functional expression of the gene.
One method of DNA sequence targeting is with triplex-forming oligonucleotides (TFOs). If these are conjugated to electrophilic or photoreactive groups, they can recognize and efficiently modify bases on either one (1 ,3 -6 ) or both (7 ,8 ) strands in targeted duplex DNA. A current difficulty with TFOs is that only purines may be in the TFO binding strand of the duplex target. This severely restricts the availability of target sites.
An alternate method, useful for targeting any sequence, is use of the Escherichia coli RecA protein with single-stranded ODNs. This protein coats single-stranded DNA, and catalyzes a homology search on double-stranded DNA targets (9 ). This can occur when the single strand is a fairly short ODN. Oligonucleotides have been used with RecA protein to mask specific methylation of restriction sites, leaving those targeted sites sensitive to restriction after the other sites were modified (10 ), and to retrieve targeted plasmids from a mixture by using biotinylated ODNs, with capture of the biotinylated ODN-bound plasmids on a streptavidin column (11 ). In the latter case, supercoiled plasmids were required. These three-stranded complexes can be stable after deproteinization, however, if the single-stranded ODN bears a reactive group that covalently binds to the target's complementary strand, as we have previously shown (12 ,13 ). This strand displacement method allows modification of any target site (four letter targeting) in DNA without sequence restriction by use of such a recombinase-assisted ODN (RAO).
In this study we show that both of these two gene targeting methods give highly efficient and specific site-directed electrophilic modification of a native gene in whole genomic human DNA. The target for the crosslinking TFO is a polypurine tract in the DQ[beta]1*0302 allele of the MHC II locus, and for the RAO, a nearby site on the same gene. Quantitative ligation-mediated PCR (LMPCR) was used to monitor the extent to which the targeted nucleotide was alkylated. With either technique, efficient ODN-directed modification was shown to have occurred at a significant rate with good specificity.
Figure 1 shows the alignment of the ODNs with the targeted sequences, the sites of modification and the structure of the oligonucleotides used in this study. The target sequence for both types of ODN was the first intron of the HLA DQ[beta]1*0302 allele, an allele associated with predisposition to juvenile onset (Type I) diabetes (14 ), and the human HT-29 cells (obtained from ATCC) used here contain a single copy of this allele. We found the sequence recorded in GenBank (15 ) (accession no. K01499) to be incorrect; Figure 1 shows the correct sequence in the area we are targeting. When a TFO with a terminal alkylating group forms a triplex with double-stranded DNA, it reacts with nearby guanines at N7 in the major groove (4 ,7 ). In RAOs the reactive moiety, located on an internal base, preferentially reacts with guanines on the complementary strand of the target sequence after synaptic complex formation.
Genomic DNA from HT29 cells was prepared with a Wizard Genomic DNA Purification Kit (Promega). To 5-10 [mu]g of genomic DNA in 90 [mu]l 20 mM HEPES, pH 7.2, 140 mM KCl, 10 mM MgCl2 and 1 mM spermine was added 10 [mu]l of a 10* stock of TFO1 or controls to give a final ODN concentration of 10-6-10-9 M. After mixing and incubation 3 h at 37oC, the DNA was pelleted by addition of 10 [mu]l of 3 M NaOAc, pH 7.0, and 300 [mu]l of ice-cold 100% ethanol, chilling (-70oC), centrifugation at 12 000 r.p.m. for 15 min at 4oC, washed with EtOH, and dried.
To 10 [mu]l of an ice cold 1 [mu]M chlorambucil-ODN (RAO1 or RAO2) solution in 10 mM Tris-HCl, pH 7.5, and 1 mM EDTA was added 10 [mu]l of 10* RecA buffer [100 mM Tris-acetate, pH 7.5, 500 mM NaOAc, 120 mM Mg(OAc)2, 10 mM dithiothreitol (DTT) and 50% glycerol], along with 10 [mu]l of 10 mM ATP[gamma]S and water to give a final volume of 87 [mu]l, which was kept ice cold. After addition of 3 [mu]l of a cold solution of RecA protein (New England BioLabs #249L) solution (2 mg/ml), 5-10 [mu]g of genomic DNA was added and the mixture was incubated for 6 h at 37oC. After addition of 82 [mu]l of 10 mM Tris-HCl, pH 7.5, containing 1 mM EDTA, 10 [mu]l of 10% SDS and 8 [mu]l of Proteinase K (5 mg/ml), and incubation for 30 min at 37oC, the reaction mixture was extracted with an equal volume of phenol-chloroform. The aqueous phase was extracted three times with ether and the DNA precipitated as above.
Most steps of this technique were performed as described (20 ,21 ), although certain modifications (22 ) were made for this work. Since we required a method to quantitate the amount of site-specific alkylation and cleavage, we generated an internal control site by cleavage of the DNA with a restriction enzyme after the chlorambucil-ODN reaction. Selection of a restriction endonuclease was based on the enzyme having a recognition site 50-200 nt upstream (5') of the alkylated base, and no recognition sites between that base and the downstream (3') sequence complementary to the first primer. A BamHI site for each strand was identified as shown in the schematic in Figure 1 . Initial experiments confirmed completion of digestion at these sites, using Southern blot hybridization and a PCR-generated probe, and also confirmed the lack of dependence of the ratio of target site cleavage to restriction site cleavage on amount of treated DNA (0.5-5 [mu]g tested) used in the LMPCR protocol.
The second modification to the protocol was heating the DNA to 95oC in the buffer used for first strand synthesis, pH 8.9, for 10 min prior to annealing and extending the first primer. This is slightly longer than the standard 3 min of the protocol (21 ), and causes quantitative depurination and cleavage of the DNA at the site of alkylation with the generation of the 5'-phosphate required for the ligation step. All chlorambucil alkylation sites in this work were designed to be on N-7 of guanines.
For each experiment an aliquot of the genomic DNA was treated with dimethylsulfate and amplified along with the rest of the samples to provide a G-ladder (20 ). The DNA samples were restricted with BamHI to completion by incubating 3 h under optimal conditions with a 3-fold excess of restriction enzyme. The volume was adjusted to 100 [mu]l with water and the DNA was precipitated as described above. The pellet was resuspended in 10 mM Tris-HCl, pH 7.5, 1 mM EDTA, to give a DNA concentration of ~0.5 [mu]g/[mu]l. To 5 [mu]l of this chilled solution in a PCR tube was added 25 [mu]l of the primer 1 solution (21 ). First strand synthesis and ligation of the universal linker (see above) was performed as described by Mueller and co-workers (21 ), except that the 95oC heating step was extended to 10 min. Nested PCR was then performed as described by Pfeifer and Riggs (20 ).
Phosphorimaging was used to analyze the polyacrylamide gel electropherograms of the LMPCR results, and the ratio of intensities of the bands due to the restriction and alkylation sites gave the efficiency of targeted alkylation.
Figure 2 shows the results of incubation of TFO1 with isolated genomic DNA from human HT29 cells in physiological buffer, as analyzed by LMPCR. The BamHI (internal standard) site was located 36 bases 5' from the target site. Since we showed by Southern blot that 100% of the DNA was cleaved by BamHI, the efficiency of cleavage at the targeted site is given by the ratio of the density in the cleavage site to the sum of the density in the cleavage and the restriction site, assuming negligible background. Table 1 gives this quantitation of specific targeting.
Two versatile techniques by which naked genomic DNA may be covalently modified with high efficiency and at specifically targeted nucleotides have been described here. Previous demonstrations of efficient targeting at this level of complexity have been limited to sites engineered in chromosomal DNA (8 ) and to integrated viral sequence in cells which were photochemically modified with lower efficiency (29 ). We now show the ability to target a site in a native gene with a triplex forming ODN with an appended electrophilic group with physiological salt concentrations. We further show the ability to target any desired sequence, by design, using reactive oligonucleotides and RecA protein catalysis.
Use of TFOs continues to be limited by their ability to `read' only two letters of the genetic code. Although work in other laboratories has attempted to expand this code by incorporating into a TFO the ability to recognize pyrimidine interruptions in the polypurine strand (30 -32 ) or to recognize two short purine tracts on alternate strands of the DNA (33 ), a general solution to recognition of DNA by triplex forming oligonucleotides is not likely to be available in the near term. Given these limitations, however, one can recognize and label targeted sites in genomic DNA at low concentrations in very high efficiency under physiological conditions. The addition of a triplex-specific intercalator that stabilizes the triplex (23 ,34 ,35 ) gave an increase in this efficiency and demonstrates the applicability of these agents to complex nucleic acid systems. With a view to cell culture work with these ODNs, we have found that the concentration of the intercalator we used, 8 [mu]M coralyne, is non-toxic to cells in culture (data not shown).
Ultimately, the targeting of user-designed sites in DNA should have no limitations regarding sequence. The method shown here is an effective method for accomplishing this: the use of RecA protein to catalyze the homology search and strand exchange by ODNs, enabling the covalent crosslinking to the targeted site. Our previous demonstration (13 ) of efficiency and selectivity of this method in plasmid DNA is now shown applicable to whole genomic DNA. We had shown that the length of the oligonucleotide needed to be at least 30 bases (13 ). In the more complex system used here, efficiency may benefit by using an ODN a few more bases longer, such as the 50mer RAO1 in Figure 3 .
The utility of agents that specifically label DNA is broad. These reactive ODNs could see immediate use in the isolation of rare sequences in complex dsDNA. For example, using chlorambucil-bearing ODNs with a biotin tag, it would be straightforward to affinity-capture complementary sequences of dsDNA which are hybridized and crosslinked to the ODN. Release of the captured DNA could be effected by depurination and strand scission of the adducted guanine adduct using the mild, non-denaturing conditions previously described by Povsic et al. (8 ). Another application of these ODNs might be their use as artificial restriction enzymes. Double-stranded cleavage of DNA would require two sites of crosslinking from the ODN to the target site on DNA, which we have shown with TFOs (7 ) and RAOs (M. Podyminogin, unpublished results).
The potential of this approach for modification of genes in cells for therapeutic purposes is most interesting. Covalent modification of targeted sites in cells would be expected to inhibit transcription by physical blockage (36 ). More importantly, permanent alteration in gene function would be achieved by the site-directed mutagenesis these agents would induce (1 ). Shuttle vector experiments have shown that targeted modification of the supF gene by photoreactive TFOs elicits mutations when processed by mammalian cells, although the efficiency reported to date is low (37 ,38 ). Modification of both strands of the target site has the potential to produce much higher rates of mutation, as we have found in preliminary studies. This powerful ability to directly change the code of a gene may be applied to correction of genetic defects due to single point mutations or to inactivation of gene function. The latter might find use, for instance, in inactivation of genes responsible for certain autoimmune diseases like type I diabetes (14 ). Ultimately, the genetic modification principles described herein are preliminary steps to a novel type of gene therapy based on synthetic oligonucleotides that modify the function of endogenous genes.
We wish to thank Dr Vladimir Gorn for synthesis of oligonucleotides, Deborah Lucas for conjugation reactions, and Drs Igor Kutyavin, David Brown and Joe Hedgpeth for helpful discussions.
*To whom correspondence should be addressed. Tel: +1 425 485 8566; Fax: +1 425 486 8336; Email: rmeyer@epochpharm.com
REFERENCES