Transcription Dependence of Chromosomal Gene Targeting by Triplex-forming Oligonucleotides*

Triplex-forming oligonucleotides (TFOs) recognize and bind to specific DNA sequences and have been used to modify gene function in cells. To study factors that might influence triplex formation at chromosomal sites in mammalian cells, we developed a restriction protection assay to detect triplex-directed psoralen crosslinks in genomic DNA prepared from TFO-transfected cells. Using this assay, we detected binding of a G-rich TFO to a chromosomal site even in the absence of transcription when high concentrations of the TFO were used for transfection. However, experimental induction of transcription at the target site, via an ecdysone-responsive promoter, resulted in substantial increases (3-fold or more) in target site crosslinking, especially at low TFO concentrations. When RNA polymerase activity was inhibited, even in the ecdysone-induced cells, the level of TFO binding was significantly decreased, indicating that transcription through the target region, and not just transcription factor binding, is necessary for the enhanced chromosomal targeting by TFOs. These findings provide evidence that physiologic activity at a chromosomal target site can influence its accessibility to TFOs and suggest that gene targeting by small molecules may be most effective at highly expressed chromosomal loci.

Triplex-forming oligonucleotides (TFOs) recognize and bind to specific DNA sequences and have been used to modify gene function in cells. To study factors that might influence triplex formation at chromosomal sites in mammalian cells, we developed a restriction protection assay to detect triplex-directed psoralen crosslinks in genomic DNA prepared from TFO-transfected cells. Using this assay, we detected binding of a G-rich TFO to a chromosomal site even in the absence of transcription when high concentrations of the TFO were used for transfection. However, experimental induction of transcription at the target site, via an ecdysone-responsive promoter, resulted in substantial increases (3-fold or more) in target site crosslinking, especially at low TFO concentrations. When RNA polymerase activity was inhibited, even in the ecdysone-induced cells, the level of TFO binding was significantly decreased, indicating that transcription through the target region, and not just transcription factor binding, is necessary for the enhanced chromosomal targeting by TFOs. These findings provide evidence that physiologic activity at a chromosomal target site can influence its accessibility to TFOs and suggest that gene targeting by small molecules may be most effective at highly expressed chromosomal loci.
Triplex-forming oligonucleotides (TFOs) 1 recognize specific sequences in double-stranded DNA and bind in the major groove at homopurine stretches. Specificity arises from the formation of Hoogsteen or reverse-Hoogsteen hydrogen bonds between the bases of the pyrimidine or purine TFO and the bases of the purine strand of the duplex (1)(2)(3)(4)(5). Because of their sequence-specific DNA binding property, TFOs have potential for manipulating gene structure and function in living cells.
TFOs can inhibit transcription by interfering with regulatory protein binding or by blocking mRNA elongation by polymerases (4, 6 -14). Alternatively, TFOs can be used for targeted gene modification. They are capable of directing site-specific DNA damage by delivering a mutagen to a specific site (1,(15)(16)(17)(18). It has also been shown that TFO binding alone, without a tethered mutagen, can stimulate repair and recombination and induce site-specific genome changes in cells through a repair-dependent process (18 -30).
The ability of TFOs to induce mutagenesis is relatively efficient when they are allowed to bind to their target in plasmid DNA in vitro prior to the transfection of cells (19), but the effect of the TFO is diminished when intracellular binding is required (18). The in vivo efficacy of TFOs is potentially limited by multiple factors in the cellular environment. For example, the presence of single-strand nucleases can lead to rapid degradation of oligonucleotides if the 3Ј-end is not protected (31). The neutral pH inside cells is suboptimal for triplex formation by pyrimidine TFOs because of the requirement for protonation of cytosine at the N3 position (32), and high potassium levels inhibit purine motif triplex formation (33)(34)(35).
An additional limiting factor may be the competition with DNA-binding proteins. Because of the complex structure into which chromosomal DNA is packaged by histones and other proteins, there has been substantial debate in the field as to whether chromosomal DNA would be accessible to TFOs. In studies conducted in vitro, triplexes did not form on DNA sequences already organized into nucleosomes (36,37) except at sites located toward the ends of nucleosomal DNA fragments (38). Conversely, pre-formed triplexes blocked nucleosome formation on DNA fragments in vitro (36,37). In Xenopus oocytes, Bailey and Weeks (39) found that a specific TFO could inhibit reporter gene expression on a chromatinized plasmid only when multiple target sites were introduced into the promoter region in a pattern designed so that not all of the sites could be simultaneously bound by a histone octamer. Together, these studies suggested that nucleosome formation and TFO-mediated triplex formation may be competing processes.
Despite these studies, several lines of evidence support the ability of TFOs to target chromosomal sites. TFOs have been shown to induce site-specific chromosomal mutagenesis both in yeast (24) and mammalian cells (22,23) and even in mouse tissues following systemic administration of TFOs in mice (25). Other studies have reported direct demonstrations of triplex formation on chromosomal targets using a number of methods for physical detection, including restriction protection, ligationmediated PCR, and primer extension (40 -43). However, some questions have been raised about possible artifacts in certain detection assays (44). In particular, Becker and Maher (44) showed that unbound oligonucleotides present in cells could persist following cell lysis and thereby influence the apparent detection of in vivo formed triplexes in a ligation-mediated PCR assay (44).
We hypothesized that the apparent discrepancies between the in vitro and in vivo studies of triplex formation on chromatinized targets may reflect the dynamic nature of chromatin structure in vivo. To test this hypothesis, we set out to examine the effect of transcription on triplex formation in vivo based on the concept that transcription is a key factor in altering chro-matin structure in living cells. Here, we report the development of an assay to detect triplex formation at a chromosomal site in mammalian cells. The assay depends on resistance to restriction enzyme cleavage at the target site conferred by triple helix-directed psoralen crosslinks. Using this assay, we tested the effect of transcription on TFO binding at a chromosomal site where transcription can be specifically regulated by hormone treatment to activate a specialized transcription factor. We show here that transcriptional activity at a chromosomal site can substantially influence the accessibility of that site to binding by a TFO, especially at low TFO concentrations. In addition, even in hormone-treated cells when RNA polymerase activity is poisoned by ␣-amanitin treatment, TFO targeting is reduced, further demonstrating a direct effect of transcription, and not just transcription factor binding, on the targeting of chromosomal sites by TFOs.

EXPERIMENTAL PROCEDURES
Oligonucleotides-Psoralen-conjugated oligonucleotides were synthesized by either Midland Certified Reagent Co. (Midland, TX) or Oligos, Etc. (Wilsonville, OR) and were gel or high pressure liquid chromatography (HPLC) purified as described previously (22). These oligonucleotides were synthesized with a propylamine group (Glen Research) on the 3Ј-end to prevent degradation by nucleases (45). Oligonucleotides for cloning were synthesized at the Keck Facility (Yale University). Their sequences are the following: MM1 (5Ј-CTA GGA TCC TTC CCC CCC CTC CTC CCC CTC CCC CTC-3Ј) and MM2 (5Ј-AGC TGA GGG GGA GGG GGA GGA GGG GGG GGA AGG ATC-3Ј).
Plasmid Constructs-Plasmid pMM2 was derived from pIND/LacZ (Invitrogen) and was constructed to contain a G:C bp-rich site amenable to triplex formation in the purine motif. The polypurine duplex target site was created by annealing oligonucleotides MM1 and MM2. These oligonucleotides were engineered to contain a BamHI restriction site overlapping the polypurine sequence at one end. pMM2 was generated by cloning the synthetic duplex into the HindIII site within the inducible expression cassette in pIND/LacZ.
Cell Lines-Cells in which transcription at the TFO binding site could be manipulated were generated by transfecting ECR-293 cells (Invitrogen), which stably express the hormone-responsive ecdysone receptor heterodimer, with ScaI-linearized pMM2 plasmid DNA using GenePORTER reagent (Gene Therapy Systems, San Diego, CA). Stable clones with chromosomally integrated pMM2 DNA were selected by colony formation upon 2 weeks of growth in selective medium consisting of high glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen), 2 mM L-glutamine, 1% penicillinstreptomycin, 400 g/ml zeocin, and 400 g/ml G418. One resulting cell line, EC293-6, was chosen for further study. In this cell line, stable chromosomal integration of the pMM2 DNA was confirmed by genomic Southern analysis. The target sequence remained stably integrated in the genome of EC293-6 cells in culture for several months as determined by genomic Southern analysis (data not shown).
Induction of Transcription and Expression Analysis-Transcription at the target locus was specifically induced by treatment of the cells with the ecdysone analog, ponasterone A. Ponasterone A was dissolved in 95% ethanol at a concentration of 1 mM and added to the growth media of EC293-6 cells at a final concentration of 5 M. Cells were incubated at 37°C for 20 h, extracts were prepared, and induced ␤-galactosidase activity was detected using the Galacto-Star chemiluminescent assay (TROPIX, Inc., Bedford, MA). A luminometer was used to measure light signal output. For direct visualization following induction with 5 M ponasterone for 20 h at 37°C, cells were fixed with 2% formaldehyde and 0.2% glutaraldehyde in PBS at 4°C for 5 min, washed three times with PBS, and stained with 20 mM potassium ferricyanide, 20 mM potassium ferrocyanide, 2 mM MgCl 2 , and 0.5 mg/ml X-gal in PBS. After incubation for 2 h at 37°C, the cells were visualized by light microscopy.
Inhibition of Transcription-To inhibit transcription, ␣-amanitin was dissolved in water at a concentration of 1 mg/ml and added to the growth media of EC293-6 cells at a final concentration of 10 g/ml. The cells were incubated at 37°C for 24 h. If cells were exposed to ␣-amanitin and ponasterone, ponasterone was added 12 h after ␣-amanitin addition.
Oligonucleotide Transfection-Cells were transfected by permeabilization with digitonin as described by Giovannangeli et al. (41). Oligonucleotides, either pso-AG30 or pso-SCR30, were added to the digito-nin-permeabilized cells at final concentrations of 0 -20 M. Following a 1.5 h incubation of the cells at 37°C in suspension, the cells were irradiated with long wavelength UV light (UVA; using a broad band UVA light source (320 -400 nm, centered at 365 nm; Southern New England Ultraviolet, Branford, CT) as described previously in Ref. 46) for psoralen photoactivation to crosslink bound psoralen-TFOs to the chromosomal DNA. The irradiation took place over 6 min, corresponding to a total dose of 1.8 J/cm 2 . UVA doses were determined by radiometry (International Light, Newburyport, MA), as described previously (46).
Cell Lysis and Preparation of Genomic DNA-Immediately following UVA irradiation, cells were washed with PBS and resuspended at a concentration of 5 ϫ 10 6 cells/ml in lysis buffer (50 mM Tris-Cl, pH 7.5, 20 mM EDTA, 100 mM NaCl, 0.1% SDS), heated to 60°C for 15 min, and treated with proteinase K at a concentration of 100 g/ml overnight at 37°C. Lysates were extracted once with phenol (equilibrated with Tris) and twice with chloroform/isoamyl alcohol (24:1). The DNA was ethanol precipitated and resuspended in 10 mM Tris-Cl, pH 8.0, 1 mM EDTA, pH 8.0, and 100 mM KCl. Once in solution, genomic DNA samples were incubated at 60°C for 2 h to disrupt non-covalent triplexes. Unbound and un-crosslinked oligonucleotides were removed by filtration through a Centricon-100 (Millipore, Bedford, MA).
Restriction Protection and Southern Analysis-Following lysis and specialized sample preparation as above, genomic DNA from EC293-6 cells was digested with BglII, EcoRI, and BamHI and analyzed by Southern blotting. The 3.1-kb HindIII/EcoRI fragment of pIND/LacZ (Invitrogen) was used as a probe. Band intensities were quantified by a PhosphorImager (Amersham Biosciences).

RESULTS
Experimental Design-To study the effect of transcription on triplex formation, we created a cell line in which we could manipulate transcription at the TFO target site in genomic DNA. Human 293 cells, previously engineered to express the ecdysone receptor heterodimer, were stably transfected with a linearized vector (pMM2), containing a target site for TFO binding downstream of an ecdysone-inducible promoter and upstream of a lacZ reporter gene (Fig. 1A), and stable clones with the vector sequence integrated within a chromosomal locus were selected by long term growth in G418. One cell line, EC293-6, was selected for further study, and chromosomal integration of the vector DNA was confirmed by genomic Southern analysis (data not shown). In the EC293-6 cells, it was also confirmed that transcription across the TFO target site could be specifically induced by treatment of the cells with ponasterone, an ecdysone analog. Ponasterone binds to the ecdysone receptor heterodimer in the cells, and this causes the receptor to bind to the ecdysone response element situated upstream of the P hsp minimal promoter, the TFO target site, and the ␤-galactosidase coding region (in that order), thereby activating transcription downstream. We measured ␤-galactosidase activity in these cells before and after induction with ponasterone. There was very little ␤-galactosidase expression in uninduced cells, but ␤-galactosidase expression was induced ϳ250-fold above background in cells exposed to 5 M ponasterone for 20 h (Fig. 1B). Direct staining revealed that expression was induced in almost all of the cells (Fig. 1C). Hence, transcription across the TFO target site could be specifically induced by ponasterone treatment of the EC293-6 cells.
In previous work, we had found that the 30-nucleotide, Grich TFO, AG30, could mediate site-directed mutagenesis and recombination in the supFG1 reporter gene in chromosomal DNA in mammalian cells and mice (22,25,30). We therefore chose the 30-bp supFG1 polypurine site as a suitable target sequence to test the influence of transcription on TFO binding. As explained above, the inducible expression cassette in the EC293-6 cells was designed to contain this 30-bp target site at the position indicated in Fig. 1A so that transcription induced by ponasterone would occur through the 30-bp site. In addition, the site was designed to overlap a BamHI restriction site at one end. In this arrangement, triple helix formation by the 5Ј-psoralen-conjugated TFO, pso-AG30, would be expected to position the psoralen for photoreaction at the 5Ј ApT 3Ј-sequence within the BamHI recognition site (Fig. 1D). Because triplexdirected psoralen adducts have been shown to block restriction enzyme cleavage if the adduct is formed within the recognition site (41,47), we used this substrate as the basis for an assay to directly measure intracellular triplex formation. We expected that triplex-directed psoralen adducts could be measured by the extent of BamHI resistance at the target site in a quantitative genomic Southern blot analysis.
However, we were concerned that unbound or un-crosslinked TFOs present in cell lysates might bind to the genomic DNA in vitro after lysis and so influence the apparent detection of in vivo formed triplexes in the restriction protection assay. We therefore developed a genomic DNA purification procedure to eliminate unbound oligonucleotides and disrupt non-covalent triplexes, thereby allowing persistence only of covalent psoralen third strand adducts in the genomic DNA. To achieve this, the genomic DNA samples were incubated in a buffer containing 10 mM Tris, 1 mM EDTA, and 100 mM KCl at 60°C for 2 h, and unbound and un-crosslinked oligonucleotides were removed via size filtration prior to restriction digestion. Under these conditions, triplex formation is inhibited because EDTA chelates Mg 2ϩ , which is important for stabilizing triple helices via charge neutralization, and the high concentration of K ϩ favors the self-aggregation of the G-rich TFO. In addition, non-covalent triplexes formed by AG30 melt at 60°C while duplexes remain intact, and the size filtration separates the free oligonucleotides from the genomic DNA. This approach was tested and validated via a series of in vitro experiments on defined DNA samples (data not shown).
Transcription Dependence of TFO Binding at a Chromosomal Site-To study the influence of transcription on TFO binding at the chromosomal site in EC293-6 cells, the cells were either treated or not with ponasterone to induce transcription at the TFO target site. They were subsequently transfected with selected oligonucleotides at a concentration of 20 M via permeabilization with digitonin to achieve high levels of intracellular delivery. The oligonucleotides included either pso-AG30 or a control oligonucleotide, pso-SCR30, which has the same base composition as pso-AG30 but possesses a scrambled sequence creating 12 mismatches so that it is not capable of binding to the target site as a third strand (Fig. 1D and Ref. 22). One and a half hours after transfection, the cells were irradiated with long wavelength UV light. A total UVA dose of 1.8 J/cm 2 was given over a period of 6 min to fix the bound TFO to its binding site via photoactivation of the psoralen moiety.
Immediately following UVA irradiation, the cells were lysed for preparation of genomic DNA using the protocol described above. The genomic DNA was digested with BglII, BamHI, and EcoRI, and the samples were analyzed by Southern blotting using a probe spanning the lacZ gene (Fig. 2). The appearance of a 3.6-kb fragment indicates inhibition of BamHI cleavage due to TFO-directed psoralen crosslinks at the overlapping site.

FIG. 1. Experimental design to study the influence of transcription on chromosome targeting by TFOs.
A, ecdysone-inducible transcription at the third strand target site. A human 293 cell sub-clone (EC293-6) expressing the ecdysone receptor heterodimer was engineered to contain an expression cassette incorporating a third strand binding site downstream of an ecdysone-inducible promoter and upstream of a lacZ reporter gene within a chromosomal locus, as indicated. E/GRE, ecdysone/glucocorticoid-responsive element. B, ponasterone-induced expression of the lacZ reporter gene. EC293-6 cells were grown in the absence or presence of 5 M ponasterone, an ecdysone analog, for 20 h, and ␤-galactosidase activity was measured in cell lysates using a chemiluminescent substrate. The bars indicate light signal output. RLU, relative light unit. C, direct staining of ponasterone-induced EC293-6 cells for ␤-galactosidase activity by histochemical methods. D, sequence of the triplex target site within the lacZ reporter construct and corresponding oligonucleotides. Pso-AG30 binds to the G:C bp-rich target site in an anti-parallel orientation relative to the purine strand of the duplex. Pso-SCR30 is a control oligonucleotide with the same base composition as pso-AG30 but with an altered sequence that creates 12 mismatches. All TFOs were conjugated to psoralen at the 5Ј-end and synthesized with a 3Ј-propylamine group. The third strand binding site overlaps a BamHI restriction site in the duplex. Restriction sites and the lengths of the fragments that would be obtained after BglII, BamHI, and EcoRI digestion of the DNA are indicated.
In the absence of crosslinks at that site, 3.1-and 0.5-kb fragments are produced (Fig. 1D). A 3.6-kb protected fragment is seen in samples from cells incubated with pso-AG30 and UVA irradiated ( Fig. 2A, lanes 3 and 8), suggesting that intracellular triplex formation can be detected at the target chromosomal site. Only the specific TFO, pso-AG30, and not the control oligonucleotide, pso-SCR30, produced site-specific adducts to inhibit BamHI cleavage at the target site (compare Fig. 2A,  lanes 3 and 8 with lanes 5 and 10). The inhibition of BamHI cleavage must be due to target site crosslinking in cells rather than binding of pso-AG30 in vitro after lysis of the cells, because the 3.6-kb fragment is absent in samples from cells transfected with pso-AG30 but not UVA irradiated ( Fig. 2A,  lanes 2 and 7).
This experiment was repeated three times, and a representative gel is shown. The intensities of bands corresponding to the protected and unprotected fragments were quantified for each sample using a PhosphorImager. The ratio of radioactive signal detected in the protected band to total signal for both the 3.6-and 3.1-kb bands provides an estimate of the extent of target site crosslinking, and standard errors were calculated for the percentage crosslinking values. Substantial binding (about 10%) was detected in the absence of induced transcription, so even minimally or untranscribed regions can be accessible to triplex formation, at least at high concentrations of transfected TFOs (20 M). Importantly, however, this analysis revealed a 3-fold increase in target site crosslinking by pso-AG30 when transcription was specifically induced at the target site (Fig. 2B). These results demonstrate that transcriptional activity at a chromosomal site substantially influences the accessibility of that site to oligonucleotide-mediated triplex formation.
Dose Dependence of Chromosomal Gene Targeting by TFOs-The experiment was repeated using varying concentrations of pso-AG30 and quantified as above to determine the dose dependence of TFO binding at the target site with or without induction of transcription. As the concentration of pso-AG30 used to transfect EC293-6 cells was increased from 0.2 to 20 M, we detected an increase in the abundance of the 3.6-kb protected fragment relative to the 3.1-kb unprotected fragment in both ponasterone-induced and -uninduced samples (Fig. 3). These data provide further evidence that inhibition of BamHI cleavage is in fact TFO dependent. These results are also quantitatively consistent with the above experiments, as we saw a ϳ3-fold increase in target site crosslinking in DNA samples from cells in which transcription was induced at the target site relative to DNA from uninduced cells when we used a 20 M dose of pso-AG30 (Fig. 3). However, when cells were transfected with pso-AG30 at a concentration of 10 M, the level of targeting was almost 10-fold higher in the ponasterone-induced cells than in the uninduced cells. At a 2 M concentration of pso-AG30, we still detected substantial TFO-mediated crosslinking in DNA from cells in which transcription was induced at the target site (about 10%), but target site crosslinking was less than 1% in DNA from uninduced cells (Fig. 3). These results indicate that at lower concentrations of oligonucleotide, transcription-modulated accessibility of a chromosomal site to TFO binding is of increased importance.
Requirement of Transcription through the Target Region for Enhanced TFO Binding-To further study the influence of transcription on TFO binding, we conducted experiments to  5, 9, and 10). In some cases as indicated, the cells were irradiated 1.5 h later with long wavelength UV light. A total UVA dose of 1.8 J/cm 2 was given over a period of 6 min to crosslink the bound TFO to its target site (lanes 3, 5, 8, and 10). Purified genomic DNA was digested with BglII, BamHI, and EcoRI and analyzed by Southern blotting. The probe detects a 3.1-kb fragment in unprotected, digested DNA. If the DNA is protected from BamHI cleavage by TFO-directed psoralen adducts at the third strand binding site, a longer fragment of about 3.6-kb results (lanes 3 and 8). B, quantification of restriction protection. The intensities of bands corresponding to the protected and unprotected fragments were quantified for each sample using a PhosphorImager. The bars indicate the percentage of target site crosslinking as measured by the ratio of radioactive signal detected in the protected band to the total signal for both bands. The experiment was repeated three times, and the error bars give standard error calculations for the percentage crosslinking values.

FIG. 3. Dose dependence of chromosomal gene targeting by
TFOs. EC293-6 cells grown in either the absence or presence of ponasterone were transfected with concentrations of pso-AG30 ranging from 0.2 to 20 M. The cells were irradiated 1.5 h later with long wavelength UV light. A total UVA dose of 1.8 J/cm 2 was given over a period of 6 min to crosslink the bound TFO to its target site. Purified genomic DNA was digested with BglII, BamHI, and EcoRI and analyzed by Southern blotting. The intensities of bands corresponding to the protected and unprotected fragments were quantified for each sample using a Phos-phorImager. The percent target site crosslinking values were determined by the ratio of radioactive signal detected in the protected band to the total signal for both bands. The experiment was repeated three times, and the error bars give standard error calculations for the percentage crosslinking values. determine the relative contributions to chromosomal accessibility of transcription through the target site versus transcription factor binding nearby the site alone. To do this, we used ␣-amanitin, an RNA polymerase inhibitor, to block transcription through the target region even when the ecdysone receptor was activated by ponasterone and capable of binding to the response element. We first evaluated the effect of ␣-amanitin on expression at the target locus. EC293-6 cells were treated with ponasterone, ␣-amanitin, or both, and ␤-galactosidase activity was measured in cell extracts. ␤-galactosidase activity in the untreated EC293-6 cells (Fig. 4A) was similar to the background level (4.9 ϫ 10 4 relative light units by this scale) in the parental cell line ECR-293, which does not contain the ␤-galactosidase expression cassette. As above, ponasterone treatment stimulated high levels of lacZ expression in the EC293-6 cells. However, when ␣-amanitin was present in either ponasterone-treated or -untreated cells, lacZ expression was prevented and in the range of background (Fig. 4A).
In parallel samples exposed to the same combinations of ponasterone and ␣-amanitin, EC293-6 cells were transfected with pso-AG30 at a concentration of 20 M and UVA irradiated 1.5 h later. The genomic DNA was purified, and target site crosslinking was analyzed by Southern blotting and quantified as above (Fig. 4B). We found that ␣-amanitin treatment of the cells led to substantially reduced TFO binding even when pon-asterone was added, indicating that transcription through the target region (and not just hormone-induced transcription factor binding) is important for the enhanced chromosomal site targeting by TFOs (Fig. 4B). However, we still detected some (ϳ10%) target site crosslinking in samples from cells treated with both ␣-amanitin and ponasterone. In comparison, less crosslinking was seen in cells treated with ␣-amanitin in the absence of ponasterone (Fig. 4B). Therefore, it is possible that transcription factor binding, by itself, may alter local chromatin structure to some degree to facilitate TFO binding at nearby sites. In the comparison of samples from cells not treated with ponasterone, ␣-amanitin by itself was found to reduce even the baseline level of TFO binding (Fig. 4B). This reduction in TFO binding may reflect effects on chromatin accessibility that are exerted at a distance because of transcription suppression at flanking loci. DISCUSSION We have developed a restriction protection assay to physically detect triple helix-directed psoralen crosslinks at a chromosomal site in mammalian cells. We found that such an assay can be prone to artifacts due to the persistence of uncrosslinked oligonucleotide in cell lysates (data not shown), but the artifacts were eliminated by modifying the DNA purification procedure. Using the modified protocol, we provided direct evidence that the TFO, pso-AG30, could bind to a chromosomal target site in human cells even in the absence of transcription, with the degree of binding dependent on TFO dose. When transcription was specifically induced at the target site, however, we measured substantial increases in target site binding over a range of TFO concentrations. However, the differential effect of transcription on binding was greatest at the lower concentrations of the TFO.
By using ␣-amanitin to inhibit RNA polymerase II in the cells, we further demonstrated that transcription through the target region, in particular, plays a major role in the enhanced chromosomal targeting by TFOs. However, transcription factor binding appeared to moderately influence accessibility, perhaps via localized effects on chromatin structure. In addition, a modest effect on accessibility due to overall cellular transcriptional activity could be inferred from the effect of ␣-amanitin on the cells not treated with ponasterone. This effect may reflect the ability of transcription at nearby sites to alter chromatin structure or DNA topology at the target site, although altered expression of a putative chromatin accessibility factor could be another explanation.
DNA within the eukaryotic nucleus is organized and packaged into chromatin by association with histone proteins. Nucleosomes are the fundamental units of chromatin and consist of 146 base pairs of DNA wound around an octamer of histone proteins, which includes two molecules each of histones H2A, H2B, H3, and H4. Because there are 10 bases per turn of the DNA helix, at least some portion of the 30-base pair AG30 target sequence would be on the inner face of the helix, which is in contact with the histone octamer, possibly rendering it inaccessible to TFO binding. Surprisingly, we detected 10% target site crosslinking by pso-AG30 at the chromosomal locus even in the absence of transcription, so chromatin structure in the basal state is not an absolute barrier to triplex formation in vivo if a sufficient concentration of oligonucleotide is present in the cell.
However, chromatin structure is dynamic in vivo due to cellular processes such as transcription, replication, repair, and recombination. We found that the accessibility of a chromosomal site to TFOs can increase at least 3-fold when transcription is specifically induced at that site. The increase in binding at the transcribed site was even greater when the TFO FIG. 4. Inhibition of RNA polymerase activity substantially diminishes chromosomal TFO binding. A, lacZ expression is inhibited in the presence of the RNA polymerase inhibitor, ␣-amanitin. EC293-6 cells were treated with either ␣-amanitin, ponasterone, both, or neither, and ␤-galactosidase activity was measured in cell lysates using a chemiluminescent substrate. The bars indicate light signal output. B, influence of ␣-amanitin on TFO target site binding. EC293-6 cells grown in the absence or presence of ponasterone (P) and with or without ␣-amanitin (A), as indicated, were transfected with pso-AG30 at a concentration of 20 M followed 1.5 h later by UVA irradiation to crosslink the bound TFO to its target site. Purified genomic DNA was digested with BglII, BamHI, and EcoRI and analyzed by Southern blotting. The bars represent percent target site crosslinking as measured by the ratio of radioactive signal detected in the protected band to the total signal for both bands. concentration was low. Because intracellular delivery is a major limiting factor in the use of oligonucleotides for genetic manipulation (48), we elected to use a specialized method of transfection (digitonin permeabilization) to maximize cellular uptake of oligonucleotides for our experiments. Because other methods of transfection typically yield lower intracellular oligonucleotide concentrations, transcription may be an especially important determinant of TFO targeting efficiency when such alternative methods are used.
Some previous studies were interpreted to suggest that there is no difference in chromosomal binding of TFOs based on transcription (41,43). However, these studies tested the effects of agents thought to have global effects on transcription and did not directly measure transcriptional activity at the specific target loci. Because the regions examined in the previous studies were downstream of active promoters, it is possible that basal transcriptional activity was already relatively high at these other target sites.
In conclusion, our results demonstrate that at sufficient TFO concentrations, triplex formation is possible in mammalian cells even at non-transcribed chromosomal loci. Therefore, even unexpressed or minimally expressed genes or intergenic regions may be targets for gene modification by TFOs. However, transcription at a chromosomal site does substantially increase its accessibility to TFOs, especially at lower TFO doses, and so TFO-mediated gene targeting may be most effective when transcriptional activity at the target region is high.