- Email: [email protected]
Characterization and Quantification of Triple Helix Formation in Chromosomal DNA
doi:10.1016/j.jmb.2004.05.079
J. Mol. Biol. (2004) 341, 979–989
Characterization and Quantification of Triple Helix Formation in Chromosomal DNA Robert Besch1, Carine Giovannangeli2, Theda Schuh1 Claudia Kammerbauer1 and Klaus Degitz1* 1
Department of Dermatology Ludwig-Maximilians University, Frauenlobstr. 9-11 80337 Mu¨nchen, Germany 2 Laboratoire de Biophysique Muse´um National d’Histoire Naturelle, CNRS UMR5153/INSERM U565 Paris, France
DNA-binding molecules that recognize specific sequences offer a high potential for the understanding of chromatin structure and associated biological processes in addition to their therapeutic potential, e.g. as positioning agents for validated anticancer drugs. A prerequisite for the development of DNA-binding molecules is the availability of appropriate methods to assess their binding properties quantitatively at the desired target sequence in the human genome. We have further developed a capture assay to assess triplex-forming oligonucleotide (TFO) binding efficiency quantitatively. This assay is based on bifunctional, psoralen and biotin-conjugated, TFOs and real-time PCR analysis. We have applied this novel quantification method to address two issues that are relevant for DNA-binding molecules. First, we have compared directly the extent of TFO-binding in three experimental settings with increasing similarity to the situation in vivo, i.e. naked genomic DNA, isolated cell nuclei, or whole cells. This comparison allows us to characterize factors that influence genomic triplex formation, e.g. chromosomal DNA organization or intracellular milieu. In isolated nuclei, the binding was threefold lower compared to naked DNA, consistent with a decreased target accessibility in the nucleosomal environment. Binding was detected in whole cells, indicating that the TFO enters the nucleus and binds to its target in intact cells in vivo, but the efficiency was decreased (tenfold) compared to nuclei. Secondly, we applied the method to characterize the binding properties of two different TFOs targeting the same sequence. We found that an antiparallel-binding GT-containing TFO bound more efficiently, but with less target sequence selectivity compared to a parallel-binding CUcontaining TFO. Collectively, a sensitive method to characterize genomic triplex formation was described. This may be useful for the determination of factors driving TFO binding efficiency and, thus, may improve the usefulness of triplex-mediated gene targeting for studies of chromatin structure as well as for therapeutic antigene strategies. q 2004 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: triplex DNA; gene targeting; oligonucleotides
Introduction Triplex-forming oligonucleotides (TFOs) recognize specific sequences in double-stranded DNA by binding in the major groove of the double helix. Hoogsteen hydrogen bonds are formed between oligonucleotide bases and purine bases in the double helix, which are already engaged in Watson– Abbreviations used: TFO, triplex-forming oligonucleotide. E-mail address of the corresponding author: [email protected]
Crick hydrogen bonds. Triplex formation via Hoogsteen interactions can occur at sequences containing a stretch of pyrimidines on one DNA strand and complementary purines on the other DNA strand (oligopyrimidine$oligopurine sequences). Depending on their base composition, TFOs bind in a parallel or antiparallel orientation to the purine-containing strand.1,2 TFOs and other DNA code-reading molecules (collectively termed antigene molecules) have been used successfully, in a sequence-specific manner, to manipulate DNA-associated functions, such as transcription (initiation or elongation), replication, and repair.2–5
0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
980 In addition, the attachment of functional moieties acting on DNA (e.g. protein domains or DNAdamaging agents), led to advanced strategies in which antigene molecules direct the corresponding activities to specific sequences in the genome. Since DNA in cells is typically bound to histones and tightly packed into chromatin, the binding and activity of antigene molecules must be determined in this context. The demonstration if and how much of a triple helix complex can form at the target sequence inside cells as well as the characterization of cellular factors influencing efficient binding are critical for applications of antigene molecules, whatever their chemical nature. In order to demonstrate the binding to complementary DNA in a cellular context, a variety of methods have been developed, mainly for TFOs. TFOs covalently linked to reactive groups have been intensively used for this purpose, and target modifications reflecting triplex formation have been demonstrated (i) by detection of site-specific genomic modifications caused by triplex-forming oligonucleotides in cells6,7 and in mice,8 (ii) by detection of specific binding of triplex-forming oligonucleotides using either competitive PCR, restriction enzyme protection assay, or primer extension,9–12 or (iii) by capture of triplex structures.13 While these studies demonstrate mostly qualitative detection, a quantitative and sensitive assessment of triplex formation is desirable (i) for the selection of target sites with optimal accessibility for oligonucleotides in the chromosomal context in the cell type of interest; (ii) for the comparison of binding strength to a certain target sequence obtained with various oligonucleotides, including new oligonucleotide modifications; and (iii) for the assessment of target accessibility in different cellular contexts. These are necessary steps for the use of sequence-specific chromatin ligands, e.g. TFOs. In this study, we characterize and quantify the binding of triplex-forming oligonucleotides to their genomic target sequence. We have further developed a capture assay that is based on the use of a bifunctional, psoralen and biotin-conjugated, TFO associated with photoactivation, capture, and real-time PCR analyses: TFO binding efficiency was assessed by quantifying the amount of triplexmodified target sequences compared to the amount of unmodified targets. Here, we have used this quantitative method to address two types of questions. First, we studied how the extent of TFO binding to a determined genomic target changed in three experimental settings with increasing complexity and increasing similarity to the situation in vivo: binding efficiency of TFO to the target sequence was compared directly and quantitatively in naked genomic DNA, in isolated cell nuclei, and in whole cells. Secondly, we evaluated how the extent of TFO binding changed with the nature of the TFO, namely CU or GT-containing TFO.
Triple Helix Formation in Chromosomal DNA
Results Design of triplex-forming oligonucleotides A 5-methylcytosine/5-propynyluracil-containing oligonucleotide (TFOcu) was designed for a previously identified 16 bp oligopyrimidine$oligopurine sequence13 in the third intron of the human ICAM-1 gene14 that contains a neighboring 5 0 -TpA motif suitable for psoralen/UVA-induced photoadduct generation. TFOcu contains 5-methylcytosine (C) and 5-propynyluracil (U) nucleosides allowing formation of T$AxU and C$GxC basetriplets (with the centre dot standing for Watson– Crick interactions and the x for Hoogsteen interactions) after binding in parallel orientation with respect to the purine strand of the doublehelical target sequence (Figure 1). The 5-methylcytosine and 5-propynyluracil nucleosides were chosen because methylation of cytosine increases binding strength at physiologic pH compared to unmodified oligonucleotides, and 5-propynyluracil nucleosides increase binding strength in the presence of low intracellular concentrations of magnesium.15 TFOcu was psoralen-modified at the 5 0 end to allow photoadduct formation after UVA photoactivation and the 3 0 end was biotinmodified to allow capture with streptavidin-coated magnetic beads. A GT-containing triplex-forming oligonucleotide (TFOgt) that binds to the same target sequence in an antiparallel orientation (Figure 1C) was used and has been described elsewhere.13 For the demonstration of sequencespecific binding, two control oligonucleotides were designed for TFOcu: COcu1 (inverted sequence) and COcu2 (scrambled sequence). COgt is a scrambled control for TFOgt.13 Characterization of triplex formation and psoralen cross-linking Specificity and efficacy of triplex-induced photoadduct formation were examined with a restriction enzyme protection assay that is based on the ability of triplex structures to interfere with restriction enzyme cleavage. A plasmid containing the triplex target sequence overlapping with an EcoNI cleavage site (Figure 2A) was incubated with oligonucleotides, UVA-irradiated, purified, and digested with EcoNI and EcoRI. Fragments of 3804 bp, 453 bp and 270 bp length were generated in the absence of oligonucleotides. In contrast, a 723 bp fragment composed of the two smaller fragments was obtained with TFOcu, suggesting that the EcoNI site overlapping the triplex site was protected specifically by the targeted photoadduct (Figure 2B). The control oligonucleotide COcu1 did not prevent restriction enzyme cleavage, supporting the view that the observed inhibition was sequence-specific. At a 30-fold molar excess of TFOcu (2 mM), a 50% inhibition was observed, and almost complete inhibition was noted at a 300-fold molar excess (Figure 2B, upper panel). Less
981
Triple Helix Formation in Chromosomal DNA
formed by TFOcu was assessed in mobility-shift assays (Figure 2C). A 23 bp ICAM-1 DNA fragment containing the target sequence was incubated with TFOcu, COcu1, or COcu2, UVA-irradiated, and subjected to denaturing gel electrophoresis. With increasing UVA doses, both monoadducts (TFOcu covalently linked to one strand of the double helix) and cross-links (TFOcu linked to both strands of the double helix) were detected, whereas no specific photoadduct was detected with the control oligonucleotides COcu1 or COcu2. At a dose of 5 J/cm2 and using an oligonucleotide concentration of 10 mM, approximately 80% cross-links and 20% monoadducts were generated in a typical experiment. Collectively, these data demonstrate the ability of the oligonucleotide TFOcu to interact, and photoreact with, the 16 bp ICAM-1 target sequence in a sequence-specific manner, as demonstrated previously with TFOgt.13 Triplex formation in genomic targets
Figure 1. Target sequence and triplex-forming oligonucleotides. A, The ICAM-1 gene structure with exons (grey) and introns (white). T, target site localized in the third intron (corresponding to positions 556–571 of the ICAM-1 gene sequence (NCBI RefSeq X59288)). P, PstI restriction sites. P1–P6, position of PCR-primers. B, A depiction of the 825 bp PstI-fragment containing the target sequence. In A and B, the lengths of the PCRamplified fragments used in the study are indicated for the corresponding set of primers. C, A 23 bp gene segment including the 16 bp oligopyrimidine$oligopurine target sequence (underlined) and a 5 0 -TpA site suitable for psoralen/UVA cross-links (boxed); sequences of the triplex-forming oligonucleotide TFOcu and the control oligonucleotides COcu1 (inverted sequence) and COcu2 (scrambled sequence), as well as sequences of the triplexforming oligonucleotide TFOgt and its control COgt (scrambled sequence). Oligonucleotides consisted of the base analogues 5-methylcytosine (C) and 5-propynyluracil (U) or of guanine and thymine, respectively, and are conjugated to psoralen (Pso) and biotin (Bio) as indicated.
inhibition was observed when potassium (140 mM) was present in the incubation buffer (Figure 2B, lower panel), suggesting that triplex formation was less efficient in the presence of physiologic intracellular concentrations of potassium. The nature of triplex-induced photoproducts
With regard to the feasibility of antigene applications, we assessed TFO binding to its target locus in the natural context of genomic DNA under different experimental conditions, i.e. in naked DNA and in chromatin. TFOcu was reacted with naked genomic DNA, and triple helices were detected by a capture assay. The assay exploits the properties of bifunctional psoralen-conjugated and biotin-conjugated oligonucleotides. If a triple helix is formed in genomic DNA, the oligonucleotide can be psoralen/UVA cross-linked with the double helix. After digestion by the appropriate restriction enzyme, triple helixcontaining fragments can be captured via binding of the biotin moiety of oligonucleotides to streptavidin-coated magnetic beads and can, subsequently, be identified by PCR amplification of a portion of the captured fragment. A PCR product as evidence for triplex formation was found when TFOcu was reacted with naked genomic DNA (Figure 3A). No PCR fragment was detected when either TFO treatment or irradiation (data not shown) was omitted, which excludes nonspecific, biotin-independent binding of the targeted genomic DNA fragment to the magnetic beads. When the control oligonucleotide COcu1 was used, a very weak PCR signal was detectable. This observation can be explained with the fact that, in addition to the 5 0 -TpA motif, neighboring the target sequence, the captured 825 bp target fragment harbors 11 more 5 0 -TpA motifs, at which photoadditions might be generated by the psoralen groups independently of triplex formation. Sample aliquots not subjected to magnetic capture (“before capture” portion) were also PCR-analyzed, and PCR-signals were generated in all reactions (TFOcu, COcu1, or no oligonucleotide) indicating that the target fragment was gel-extracted and intact before magnetic separation. In a next step, the ability of TFOcu to bind to the
982
Triple Helix Formation in Chromosomal DNA
Figure 3. Triplex formation in naked genomic DNA and isolated cell nuclei. A, Naked genomic DNA was exposed to TFOcu, COcu1 or no oligonucleotide (no oligo), and magnetic capture was carried out (PCR after capture, left panel, lanes 1–3) or not (PCR before capture, right panel, lanes 4–6). B, Isolated A431 nuclei were exposed to TFOcu, COcu1, COcu2, or no oligonucleotide (no oligo), and magnetic capture was carried out (PCR after capture, left panel, lanes 1–4) or not (PCR before capture, right panel, lanes 5–8). C, Isolated A431 nuclei were exposed to TFOcu or COcu1, and magnetic capture was carried out. Coamplification of the triplex-containing target fragment (primers P1 and P2 in Figure 1) with an unrelated DNA fragment (primers P5 and P6 in Figure 1) was as described in Materials and Methods.
Figure 2. Triplex formation in vitro. A, A plasmid containing the target sequence overlapping with an EcoNI recognition site was incubated with a 3–300-fold molar excess (0.2–20 mM) of oligonucleotides, UVAirradiated (5 J/cm2), purified, and digested with EcoNI and EcoRI (see Materials and Methods). The length of the three digestion products (270 bp, 453 bp, and 3804 bp) is indicated. Inhibition of EcoNI by triplex structures results in a longer 723 bp (Z453C270) fragment. B, Incubation of plasmid with oligonucleotides in a buffer containing 10 mM Tris–HCl (pH 7.5), 10 mM MgCl2, 1 mM spermidine (upper panel) or in a buffer containing additionally 140 mM potassium (lower panel). C, A 23 bp fragment containing the target sequence (as shown in Figure 1) was
target in its intact supranucleosomal structure was tested by exposing oligonucleotides to isolated nuclei of A431 cells. When TFOcu was reacted with nuclei, a PCR product was detected (Figure 3B). As observed for naked genomic DNA,
incubated with 10 mM oligonucleotides and irradiated with various dosages of UVA. Photoadducts (a, crosslinks; b, monoadducts) were retarded under denaturing conditions relative to unbound single-stranded target sequence (c).
Triple Helix Formation in Chromosomal DNA
no PCR product was detectable in samples without oligonucleotide, suggesting complete separation. Again, weak PCR signals were detected with the control oligonucleotides, COcu1 and COcu2 (Figure 3B), consistent with triple helix-independent background psoralen cross-links. In an attempt to confirm the assumption of random cross-links in a second and independent fashion, separated DNA was PCR-amplified using (i) primers specific for a 388 bp portion of the target fragment and, simultaneously, (ii) primers specific for an unrelated 799 bp control fragment. This control fragment is located in the first exon and intron of the ICAM-1 gene, contains 27 TpA sites, and is separated from the triple helix target fragment by multiple PstI sites (Figure 1A). Whereas the primers specific for the target fragment produced a strong signal with TFOcu and a weak signal with COcu1, the primers for the 799 bp control fragment produced a weak signal of equivalent strength with both TFOcu and COcu1 (Figure 3C). These results are consistent with the generation of cross-links due to triple helix specific binding and a background amount of triple helix independent photoadditions at 5 0 -TpA motifs randomly distributed throughout the genome. As in the experiments with naked genomic DNA, all “before capture” portions produced PCR signals indicating that the fragments were gel-extracted and intact before magnetic separation (for the 825 bp target fragment, Figure 3B, right panel; for the 799 bp control fragment, data not shown). Collectively, these data suggest that sequencespecific adducts are formed in cell nuclei by TFOcu at the 16 bp oligopyrimidine$oligopurine ICAM-1 target locus, as demonstrated with the TFOgt.13 These results indicate that the target sequence is accessible in the intact supranucleosomal structure; however, no quantitative data were obtained regarding the proportion of target sequences covered by a triplex structure. Quantification of triplex formation in naked genomic DNA In order to further characterize the binding of triplex-forming oligonucleotides to their target sequence, we quantified the amount of triplexmediated cross-links by a sequential PCR procedure combining conventional and real-time PCR (see Materials and Methods, Figure 4A). We tested the propynyl-modified oligonucleotide, TFOcu (parallel binding mode) and compared it to an unmodified GT-containing oligonucleotide, TFOgt (antiparallel binding mode). When genomic DNA (starting amount of 2 mg) was reacted with TFOcu, around 20,000 triplex-modified target fragments were detected (Figure 4B, left panel). A smaller number of target fragments (approximately 2500) was captured with the control oligonucleotide COcu1, which was about an eighth of the amount captured by the triplex-forming oligonucleotide and likely due to random photoadditions at the ICAM-1 fragment independently of the TFO, as
983 mentioned above. When no oligonucleotide was added, an even smaller amount of fragments was captured. Capture of fragments under these circumstances could be due to minimal background binding of target fragments to magnetic beads independently of biotinylation. When genomic DNA was reacted with TFOgt in the same experimental conditions, the amount of triplex-modified target fragments was higher (approximately 65,000, Figure 4B, right panel). A smaller, but still substantial amount of target fragments was captured with the control oligonucleotide COgt (approximately 22,000), which was about threefold less than that captured by TFOgt. In order to assess what percentage of target sequences contained triple helices, we compared the amount of target fragments with triple helices (captured fraction) with the amount of target fragments free from triple helices (contained in the supernatant generated during the magnetic separation). In genomic DNA reacted with TFOcu, 21.2% of targets were separated compared to 2.8% when the control oligonucleotide was used (Figure 4C, left panel). With TFOgt, 70.6% of targets were captured compared to 23.5% when the control oligonucleotide was used (Figure 4C, right panel). Collectively, these experiments quantify the specific capture of target fragments due to triple helix formation in the context of naked genomic DNA. Whereas TFOgt displayed a high binding efficiency compared to TFOcu, it showed less sequence-specific binding in our experimental conditions. Quantification of triplex formation in isolated nuclei We applied this quantification strategy to isolated A431 nuclei exposed to triplex-forming oligonucleotides, in order to examine the influence of chromosomal organization on triplex binding. Due to its more specific binding properties, we selected TFOcu for these further experiments. Similar to the results obtained with genomic DNA, exposure of nuclei to TFOcu resulted in a greater amount (around threefold) of captured target fragments compared to exposure to COcu1 or to no oligonucleotide (Figure 5A). However, the amount of fragments captured from isolated nuclei was threefold lower than the number of fragments captured from naked genomic DNA, with the percentage of target fragments carrying triple helices reduced to 7.9% (Figure 5B). The experimental conditions were otherwise identical for purified genomic DNA and nuclear DNA, and they are not likely to account for the observed difference. Consequently, this decrease may be due to: (i) the nuclear chromatin structure or the nucleosomal organization, which might be an obstacle for oligonucleotides on their way to their genomic target sequence; (ii) a lower efficacy of photoaddition at TpA sequences in the context of chromatin; or (iii) a lower concentration
984
Triple Helix Formation in Chromosomal DNA
Figure 4. Quantification of triplex formation in genomic DNA. DNA was subjected to a ten cycle conventional PCR, purified to remove beads, and subsequently quantified by real-time PCR as described in Materials and Methods. A, Various amounts of genomic DNA (2, 5, 20, 50, and 200 ng) were subjected to this sequential PCR procedure. Realtime cycle numbers are plotted against the amount of template DNA. One representative experiment is depicted. In each experiment, such a set of DNA standard amounts was analyzed along with test samples to create a standard curve for the calculation of the DNA amount in test samples. B, Number of target fragments (meanGSD of three independent experiments) captured from naked genomic DNA that had been treated with oligonucleotides as indicated. C, The ratio between triplex-bound (captured) and triplex-free (supernatant) target fragments (meanGSD of three independent experiments).
of TFO available for triplex formation due to potential trapping by nuclear proteins. Quantification of triplex formation in whole cells In a further step towards the in vivo situation, we quantified triplex formation in cultured cells. A431 cells were transfected with oligonucleotides and UVA-irradiated in order to cross-link psoralenconjugated triplex-forming oligonucleotides. Subsequently, genomic DNA was prepared and cross-linked target fragments were analyzed. When cells were transfected with TFOcu, real-time PCR analysis allowed the detection of triplex formation
of TFOcu in cells (Figure 5). The amount of triplexbearing DNA fragments captured from cells was small compared to isolated nuclei (approximately tenfold decrease) or naked genomic DNA (approximately 30-fold decrease). This could be explained by a reduced intracellular concentration of TFO due to transfection efficiency or due to an intracellular degradation of the TFO. However, the latter possibility is unlikely, since we found that the oligonucleotide was stable under cellular conditions: when TFOcu-transfected cells were cultivated for longer periods, about 50% of the initial amount of oligonucleotide could still be detected 72 hours after transfection (data not shown), and
Triple Helix Formation in Chromosomal DNA
985
Figure 6. Quantification of triplex formation in whole cells. DNA was isolated from cells and subjected to a ten cycle conventional PCR, purified, and subsequently quantified by real-time PCR as described in Materials and Methods. The number of target fragments (meanG SD of three independent experiments) captured from DNA purified from whole A431 cells that had been treated transfected with oligonucleotides is indicated. Figure 5. Quantification of triplex formation in isolated cell nuclei. DNA prepared from nuclei was subjected to a ten cycle conventional PCR, purified, and subsequently quantified by real-time PCR as described in Materials and Methods. A, Number of target fragments (meanGSD of three independent experiments) captured from DNA purified from A431 cell nuclei that had been treated with oligonucleotides as indicated. B, The ratio between triplex-bound (captured) (meanGSD of three independent experiments) and triplex-free (supernatant) genomic target fragments.
there was no degraded oligonucleotide species detectable by gel electrophoresis at all observation points. This 50% decrease may be attributed to the dilution of the oligonucleotide in a population of growing cells. Nevertheless, exposure of cells to TFOcu did result in a substantial amount of photoadducts, which was significantly higher (approximately 3.5-fold) than in samples with the control oligonucleotide COcu1 or without oligonucleotide (Figure 6). Taken together, these data indicate that the dendrimer-compacted TFOcu (i) enters intact A431 cells, (ii) has access to the nucleus, and (iii) binds to its chromosomal target sequence all under conditions of intact cells.
Discussion An important step in the development of sequence-specific DNA ligands is the quantitative assessment of target accessibility in a defined cellular context. To that end, we have developed a quantitative assay based on the use of bifunctional
psoralen-conjugated and biotin-conjugated TFOs associated with photoactivation, magnetic capture, and a PCR procedure. Oligonucleotides were biotin-conjugated to allow magnetic capture of the modified DNA fragment and psoralen-conjugated to allow UVA-induced photoaddition after triplex formation at the target site (Figure 2).13 This covalent cross-linking proved technically useful for the demonstration of triplex formation in various experimental settings, including whole cells. Our approach permits the PCR quantification of triplex-modified target sequences (capture fraction), and of unmodified target sequences (supernatant fraction) allowing the determination of the percentage of triplex-modified targets. Such a direct and sensitive quantitative assay has not been described to our knowledge. Triplex formation and accessibility of target sequences in chromatin had been quantified in a limited number of studies in isolated nuclei9,11 or in whole cells.10,12 However, no quantitative comparison is available concerning triplex formation to a given DNA sequence as a function of DNA environments (namely, naked DNA, isolated nuclei, and whole cells), or of the nature of TFO (CU or GT-containing TFO). Here, we have used our quantitative capture assay to address these issues. In our system, the amount of triplex-bound target sequences was decreased threefold in nuclei compared to naked genomic DNA. Since experimental conditions were kept identical for the two different sources of DNA, this likely reflects the association of the double helix with nucleosomes, which could obscure the oligonucleotide target site and/or the TpA site required for photoadduct formation and subsequent detection in the capture assay. Indeed, it
986 was shown in vitro with histone-associated short DNA fragments that the accessibility of a target sequence varies, depending on its position relative to the histone.16 In addition, the accessibility of a certain target sequence might change, depending on the transcriptional status of the targeted gene,17 and may depend on the cell-cycle, since increased triplex formation during S phase has been demonstrated.18 In our experiments, cells were not synchronized, which reflects an in vivo situation, and the level of ICAM-1 transcription is low:19 these conditions might have caused low target accessibility. Finally, oligonucleotide trapping by nuclear proteins may lead to a decrease in effective oligonucleotide concentration and to a subsequent decrease in oligonucleotide binding to DNA compared to naked DNA. In whole cells, the amount of triplex-carrying target fragments was decreased even further (tenfold compared to isolated nuclei), indicating that there are additional factors beside the nuclear structure that are inhibitory to triplex formation. Several reasons may explain this difference between the level of triplex formation in isolated nuclei and in whole cells: (i) intracellular concentrations of oligonucleotide may be reduced, depending on transfection efficacy and oligonucleotide entrapment in endosomes prior to reaching the nuclear target; (ii) intranuclear ion concentrations in cells, e.g. magnesium or potassium, may be different from the one with the isolated nuclei and less favorable for triplex formation. In this regard, it is noteworthy that the presence of a physiologic concentration of potassium reduced triplex formation in restriction enzyme protection assays (Figure 2). Besides providing a quantitative assessment of triplex formation in different target structures, the assay proved helpful in the comparative evaluation of different triplex-forming oligonucleotides. Here, we characterized triplex formation at the genomic ICAM-1 target site using oligonucleotides that bind to the target sequence in parallel (CU-containing oligonucleotide) or antiparallel (GT-containing oligonucleotide) orientation. In the case of TFOcu, the cytosine nucleotides were 5-methylated and the thymine nucleotides were replaced with 5-propynyluracil, because these modifications improve binding characteristics to double-stranded DNA,15 as well as resistance towards cellular nucleases.20 Indeed, in our system, the modified oligonucleotide was found to be stable after transfection into the keratinocyte cell line A431, since no degradation was noticed. Furthermore, propynyl-modified oligonucleotides were found to penetrate stripped skin, which is an interesting feature with regard to the development of a topical antigene strategy when targeting the ICAM-1 gene.21 The specificity and efficacy of TFO binding was evaluated in restriction enzyme protection and mobility-shift assays, and in capture assays in which we demonstrated TFO target binding in the context of genomic DNA from various sources. When naked genomic DNA was used as a target,
Triple Helix Formation in Chromosomal DNA
the GT-containing oligonucleotide appeared to bind more strongly compared to the CU-containing oligonucleotide (Figure 4B), an observation consistent with the data obtained with plasmid targets (Figure 2B).13 On the other hand, the TFO/CO ratio is higher with the CU-containing oligonucleotide, suggesting that the GT-containing oligonucleotide binds less sequence-specifically compared to the CU-containing oligonucleotide in our experimental conditions (Figure 4B). This is consistent with semiquantitative data for nuclear DNA comparing those two oligonucleotides: the TFO/CO ratio appeared to be higher with the CU-containing oligonucleotide (Figure 3B) than with the GT-containing oligonucleotide.13 Quantitative analysis of chromosomal binding of TFOs is desirable in the development of an antigene strategy to address several issues: (i) it allows assessment of the accessibility of target sequences in a gene of interest and selection of optimal sites for targeting with triplex-forming oligonucleotides; (ii) it permits evaluation of binding characteristics like target affinity and target selectivity of triplexforming oligonucleotides in the genomic context, especially for newly developed oligonucleotide analogues or modifications; (iii) and it is a way to quantitatively investigate the accessibility of a DNA sequence in various cellular situations, e.g. during inflammatory processes, or during the cell-cycle, which may have an influence on the transcriptional status of a targeted gene(s). Indeed, it appears that the chromatin structure is dynamic enough to allow TFO binding. Antigene molecules as sequencespecific chromatin ligands can, therefore, be useful probes of chromatin structure in vivo and of associated regulatory pathways. They may be used to control these processes in a sequencespecific manner. However, some sites may be more accessible to antigene molecules than others. A better understanding of the factors that influence complex formation in the nuclear context will be critical for the improvement of gene-targeting strategies. In this context, the development of methods to quantitatively determine triplex formation in chromatin, such as that described here, will be useful.
Materials and Methods Oligonucleotides The triplex-forming oligonucleotide TFOcu and its control oligonucleotides COcu1 (inverted sequence) and COcu2 (scrambled sequence), and the triplex-forming oligonucleotide TFOgt and its control oligonucleotide COgt were obtained from Eurogentec (Seraing, Belgium) or Qbiogene (Heidelberg, Germany); the sequences are displayed in Figure 1. The 5 0 or 3 0 end was modified with 4,5 0 ,8-trimethylpsoralen via a six carbon linker and the other end was conjugated to biotin. Oligonucleotides for PCR were obtained from Metabion (Mu¨nchen, Germany).
987
Triple Helix Formation in Chromosomal DNA
UVA treatment
Mobility-shift assays
UVA was applied with a PUVA 200 light arch (Waldmann, Villingen-Schwenningen, Germany) with F8T5 PUVA bulbs, whose emission spectrum is mainly between 315 nm and 365 nm. Irradiation was done at room temperature through window glass in order to eliminate traces of UVB. The irradiation dose delivered to the samples was determined by a UV-meter (Waldmann) and the fluence was 2 mW/cm2 at a 19 cm source to target distance.
Triplex formation and photoadduct formation in vitro was additionally analyzed in mobility-shift assays. A 23 bp double-stranded DNA fragment representing the ICAM-1 target locus containing the 16 bp oligopyrimidine$oligopurine sequence (Figure 1C) was produced by annealing equimolar amounts of complementary singlestranded oligodeoxynucleotides. Both 3 0 -ends were labelled with digoxigenin (DIG Gel Shift Kit, Roche). Labelled target DNA (30 fmol) was incubated with 10 mM oligonucleotides in 10 ml of triplex-forming buffer. After incubation for 90 minutes at 37 8C, samples were either not irradiated or irradiated with various doses of UVA. TFOs bound to the target duplex during irradiation were linked covalently to the duplex at the 5 0 TpA psoralen acceptor site via psoralen/UVA photoreactions, and the resulting covalent complexes were assessed as mobility shifts by denaturing gel electrophoresis (16% polyacrylamide gel containing 7 M urea; 10% (v/v) glycerine in 44.5 mM Tris (pH 8.3), 44.5 mM boric acid, 1 mM EDTA).
Cell culture The human squameous cell carcinoma-derived cell line A431 was from American Type Culture Collection, Rockville, MD. Cells were cultured in DMEM supplemented with 2 mM L-glutamine, 100 units/ml of penicillin, 100 mg/ml of streptomycin, 1 mg/ml of amphotericin B (all from Life Technologies, Karlsruhe, Germany), and 10% (v/v) heat-inactivated fetal calf serum (ccpro, Neustadt, Germany), at 37 8C and 5% (v/v) CO2. Cells were kept from confluency. UVA irradiation of cells was done in PBS at room temperature.
Intracellular stability of oligonucleotides A431 cells were transfected with 3 mM TFOcu and Superfecte (Qiagen, Hilden, Germany) as described.13 DNA was prepared 24 hours, 48 hours, and 72 hours after transfection by cell lysis, extraction with phenol/chloroform, and precipitation in ethanol. Purified DNA was electrophoresed on a denaturing 16% (w/v) polyacrylamide gel, blotted to nylon membranes, and detected via its biotin moiety with the Biotin Luminescent Detection Kit (Roche Molecular Biochemicals, Mannheim, Germany).
Restriction enzyme protection assay Specific triplex formation was evaluated in vitro with a restriction enzyme protection assay. The 16 bp ICAM-1 target sequence was included in a 28 bp sequence to create an EcoNI recognition site that overlaps with the 16 bp oligopyrimidine$oligopurine ICAM-1 target sequence (Figure 2A). This sequence was inserted into a 4.5 kb plasmid that had been constructed from the vectors pCAT-Basice and pSV beta Gale (Promega, Mannheim, Germany). Plasmid DNA (2 mg; 0.66 pmol) was incubated with increasing amounts of oligonucleotides for 90 minutes at 37 8C in 10 ml of a triplex-forming buffer (10 mM Tris–HCl (pH 7.5), 10 mM MgCl2, 1 mM spermidine). Reactions were UVA-irradiated (5 J/cm2) in order to covalently link TFOs bound to the target duplex at the 5 0 TpA psoralen acceptor site via psoralen/UVA photoreactions. DNA was then purified and, subsequently, digested with EcoNI and EcoRI (New England Biolabs, Beverly, USA) at 37 8C overnight. Photoadduct formation was necessary to retain triple helix integrity during DNA purification, and purification had to be performed, because the triplex-forming buffer interferes with EcoNI enzyme activity. The extent of inhibition of EcoNI cleavage by triplex-mediated photoadducts was assessed by gel electrophoresis and quantified densitometrically (GelDoc system, BioRad, Mu¨nchen, Germany).
Triplex formation in naked genomic DNA Genomic DNA isolated from A431 cells was digested with PstI (Roche) resulting in an 825 bp ICAM-1 fragment containing the triplex target sequence (Figure 1B). After restriction, 2 mg of genomic DNA was incubated with various oligonucleotides (10 mM) in 10 ml of triplexforming buffer for 90 minutes at 37 8C. UVA irradiation (5 J/cm2) was performed to allow photoadduct formation. Oligonucleotides bound to the ICAM-1 target sequence were separated from unbound oligonucleotides by agarose gel electrophoresis, and fragments sized between approximately 600 bp and 1000 bp (including the 825 bp target fragment) were extracted from gels. This separation step was necessary prior to the magnetic capture, because unbound oligonucleotides would compete with target-bound oligonucleotides for streptavidin on magnetic beads. Aliquots of each sample were: (i) saved (termed “before capture” portion) for control of successful gel extraction of the 825 bp ICAM-1 target fragment; and (ii) subjected to magnetic separation (termed “after capture” portion). Triplex formation in isolated nuclei Nuclei were prepared by incubation of 3!106 A431 cells in a buffer containing 0.32 M sucrose, 10 mM Tris– HCl (pH 7.5), 5 mM MgCl2, 0.5% (v/v) Triton X-100, for ten minutes on ice. Nuclei were then washed in the same buffer, transferred to triplex-forming buffer, incubated with various oligonucleotides (10 mM) for 90 minutes at 37 8C in a 100 ml reaction volume with gentle resuspension to avoid sedimentation of nuclei, and UVAirradiated (5 J/cm2). The integrity of nuclei was checked microscopically. DNA was prepared using the DNeasy kit (Qiagen), digested with PstI, and gel-isolated as described above. Aliquots of each sample were either saved as “before capture” portions or subjected to magnetic separation (“after capture” portions). Triplex formation in cells Cells were transfected with oligonucleotides using Superfecte (Qiagen) as described previously13 with modifications: Oligonucleotides (in 1950 ml of serum-free DMEM without antibiotic and antimycotic supplementation) were mixed with Superfecte (50 ml). The final
988
Triple Helix Formation in Chromosomal DNA
oligonucleotide concentration was 10 mM. After 15 minutes incubation at room temperature, the transfection mixture was added to HBSS-washed cells at 70% confluency (approximately 900,000 cells per 6 cm well). Cells were incubated for five hours. For cross-link formation, cells were washed, transferred to PBS, and irradiated with UVA (5 J/cm2) at room temperature. After irradiation, cells were trypsinized, harvested, and genomic DNA was prepared utilizing anion-exchange-resin columns (Blood & Cell Culture Mini Kit, Qiagen). DNA (2 mg) was digested with PstI and gel-extracted. Aliquots of each sample were subjected to magnetic separation. Magnetic capture Triple helix structures from naked genomic DNA, isolated nuclei, or cells were separated by capture of oligonucleotides via their biotin moieties using streptavidin-coated magnetic beads.13 To reduce unspecific DNA binding, magnetic beads (Dynabeads, Dynal, Hamburg, Germany) were preincubated with salmon testes DNA (5 mg/ml; Sigma-Aldrich, Deisenhofen, Germany) in 2 M LiCl for one hour at 37 8C. Then 10 mg of magnetic beads was incubated with 2 mg of genomic DNA from various sources in a 50 ml reaction volume containing 2 M LiCl and 2 mg/ml of salmon testes DNA for 30 minutes at 37 8C. Separation of beads was carried out with a magnetic separator (MPC-E, Dynal) for two minutes. Beads were washed in order to remove nonbiotinylated DNA fragments (three washes with 300 ml of redistilled water for five minutes at 60 8C). Subsequently, beads were resuspended in 50 ml of redistilled water, and aliquots were subjected to PCR. For the analysis of target fragments that do not carry triplex-forming oligonucleotides and, therefore, do not carry biotin and cannot be captured, the supernatant remaining after magnetic separation and the subsequent washes were collected, purified (PCR Purification Kit, Qiagen), and subjected to PCR. In experiments involving quantification of target fragments in supernatants, salmon testes DNA was omitted during the magnetic separation procedure (see above) due to interference with fluorescence detection of real-time PCR. PCR reactions
fluorescence detection of real-time PCR products, the DNA was preamplified in a conventional PCR followed by a purification step to remove beads. More precisely, aliquots of resuspended beads harboring captured DNA were subjected to a ten cycle PCR with 250 nM primers P3 and P4 P3 : 5 0 K CCAACCTCACCGTGGTGCTGCT K 3 0 P4 : 5 0 K CCCACCTTCTCCCTGCTGGCTT K 3 0 which amplify a 242 bp stretch within the 825 bp ICAM-1 fragment. PCR conditions were 1 mM MgCl2, 1.5 units Taq polymerase (peqlab), 60 seconds at 94 8C, 60 seconds at 63 8C, 90 seconds at 72 8C. PCR amplificates were purified (PCR Purification Kit, Qiagen) in order to remove beads. Aliquots were then analyzed by real-time PCR in a LightCyclere (Roche) apparatus using the FastStart DNA Master SYBR Green I Kit (Roche) for fluorescence detection of PCR products. PCR conditions were: 500 nM primers P3 and P4, 2 mM MgCl2, 35 cycles with 15 seconds at 95 8C, ten seconds at 63 8C, 15 seconds at 72 8C, 15 seconds at 84 8C. The last temperature-step was implemented to reduce fluorescence produced by primer dimers. Specific amplification of the 242 bp fragment was controlled by melting curve analysis22 and gel electrophoresis. The amount of DNA was calculated by comparison with standards (2–200 ng of PstI-digested genomic A431 DNA) that were subjected to PCR analysis in the same way as the test reactions, i.e. conventional PCR, purification and real-time PCR. Real-time cycle numbers (defined as the crossing point calculated as the second derivative maximum by the Light cycler software) were found to be proportional to the amount of template DNA (Figure 4A). The number of target fragments in each sample was calculated from the DNA amounts determined by PCR assuming that 1 ng of human genomic DNA contains 280 haploid genomes and, therefore, 280 target fragments. The number of target fragments was measured both in the captured fraction and in the supernatant, which together represent the total amount of DNA subjected to capture. The sum of target fragments in the captured portion and in the supernatant was determined for each sample and used for normalization among all samples allowing the comparison of genomic, nuclear, and cellular DNA sources.
PCR-amplification of captured DNA was performed with 250 nM primers P1 and P2: P1 : 5 0 GAACTGGCACCCCTCCCCTCTT 3 0 P2 : 5 0 K CCGGGGCCACACCCATCTCAAA K 3 0 specific for a 388 bp portion of the 825 bp PstI target fragment that contains the triplex target-site (for primer locations see Figure 1). Reaction conditions were: 1.0 mM MgCl2, 1.5 units of Taq polymerase (peqlab, Erlangen, Germany), 32 cycles with 30 seconds at 94 8C, 30 seconds at 63 8C, 30 seconds at 72 8C. In some experiments, an unrelated 799 bp fragment of the ICAM-1 gene was coamplified in the same reaction tube as a control using 250 nM primers P5 and P6 P5 : 5 0 K CCAGTCGACGCTGAGCTCCTCT K 3 0 P6 : 5 0 K CCTCTTCCCTAGCTGGGCTGGA K 3 0 For a quantitative and more sensitive assessment of the percentage of captured DNA, a real-time PCR procedure was developed. Since magnetic beads interfere with the
References 1. Thuong, N. T. & He´le`ne, C. (1993). Sequence-specific recognition and modification of double-helical DNA by oligonucleotides. Angew. Chem. Int. Ed. Engl. 32, 666–690. 2. Faria, M. & Giovannangeli, C. (2001). Triplex-forming molecules: from concepts to applications. J. Gene. Med. 3, 299–310. 3. Uil, T. G., Haisma, H. J. & Rots, M. G. (2003). Therapeutic modulation of endogenous gene function by agents with designed DNA-sequence specificities. Nucl. Acids Res. 31, 6064–6078. 4. Ebbinghaus, S. W. (2003). Site-selectiveDNA binding drugs. Chem. Biol. 10, 895–897. 5. Dudouet, B., Burnett, R., Dickinson, L. A., Wood, M. R., Melander, C. Belitsky, J. M. et al. (2003). Accessibility of nuclear chromatin by DNA binding polyamides. Chem. Biol. 10, 859–867. 6. Vasquez, K. M., Wang, G., Havre, P. A. & Glazer, P. M.
Triple Helix Formation in Chromosomal DNA
7.
8. 9.
10.
11.
12.
13.
14.
(1999). Chromosomal mutations induced by triplexforming oligonucleotides in mammalian cells. Nucl. Acids Res. 27, 1176–1181. Vasquez, K. M., Dagle, J. M., Weeks, D. L. & Glazer, P. M. (2001). Chromosome targeting at short polypurine sites by cationic triplex-forming oligonucleotides. J. Biol. Chem. 276, 38536–38541. Vasquez, K. M., Narayanan, L. & Glazer, P. M. (2000). Specific mutations induced by triplex-forming oligonucleotides in mice. Science, 290, 530–533. Giovannangeli, C., Diviacco, S., Labrousse, V., Gryaznov, S., Charneau, P. & Helene, C. (1997). Accessibility of nuclear DNA to triplex-forming oligonucleotides: the integrated HIV-1 provirus as a target. Proc. Natl Acad. Sci. USA, 94, 79–84. Belousov, E. S., Afonina, I. A., Kutyavin, I. V., Gall, A. A., Reed, M. W. Gamper, H. B. et al. (1998). Triplex targeting of a native gene in permeabilized intact cells: covalent modification of the gene for the chemokine receptor CCR5. Nucl. Acids Res. 26, 1324–1328. Faria, M., Wood, C. D., Perrouault, L., Nelson, J. S., Winter, A. White, M. R. et al. (2000). Targeted inhibition of transcription elongation in cells mediated by triplex-forming oligonucleotides. Proc. Natl Acad. Sci. USA, 97, 3862–3867. Oh, D. H. & Hanawalt, P. C. (1999). Triple helixforming oligonucleotides target psoralen adducts to specific chromosomal sequences in human cells. Nucl. Acids Res. 27, 4734–4742. Besch, R., Giovannangeli, C., Kammerbauer, C. & Degitz, K. (2002). Specific inhibition of ICAM-1 expression mediated by gene targeting with triplexforming oligonucleotides. J. Biol. Chem. 277, 32473–32479. Degitz, K., Li, L. J. & Caughman, S. W. (1991). Cloning and characterization of the 5 0 -transcriptional regulatory region of the human intercellular adhesion molecule 1 gene. J. Biol. Chem. 266, 14024–14030.
989 15. Lacroix, L., Lacoste, J., Reddoch, J. F., Mergny, J. L., Levy, D. D. Seidman, M. M. et al. (1999). Triplex formation by oligonucleotides containing 5-(1-propynyl)-2 0 -deoxyuridine: decreased magnesium dependence and improved intracellular gene targeting. Biochemistry, 38, 1893–1901. 16. Brown, P. M. & Fox, K. R. (1998). DNA triple-helix formation on nucleosome-bound poly(dA)*poly(dT) tracts. Biochem. J. 333, 259–267. 17. Macris, M. A. & Glazer, P. M. (2003). Transcription dependence of chromosomal gene targeting by triplex-forming oligonucleotides. J. Biol. Chem. 278, 3357–3362. 18. Majumdar, A., Puri, N., Cuenoud, B., Natt, F., Martin, P. Khorlin, A. et al. (2003). Cell cycle modulation of gene targeting by a triple helix forming oligonucleotide. J. Biol. Chem. 278, 11072–11077. 19. Caughman, S. W., Li, L. J. & Degitz, K. (1990). Characterization and functional analysis of interferon-gamma-induced intercellular adhesion molecule-1 expression in human keratinocytes and A-431 cells. J. Invest. Dermatol. 94, 22s–26s. 20. Wagner, R. W., Matteucci, M. D., Lewis, J. G., Gutierrez, A. J., Moulds, C. & Froehler, B. C. (1993). Antisense gene inhibition by oligonucleotides containing C-5 propyne pyrimidines. Science, 260, 1510–1513. 21. White, P. J., Fogarty, R. D., Liepe, I. J., Delaney, P. M., Werther, G. A. & Wraight, C. J. (1999). Live confocal microscopy of oligonucleotide uptake by keratinocytes in human skin grafts on nude mice. J. Invest. Dermatol. 112, 887–892. 22. Ririe, K. M., Rasmussen, R. P. & Wittwer, C. T. (1997). Product differentiation by analysis of DNA melting curves during the polymerase chain reaction. Anal. Biochem. 245, 154–160.
Edited by J. O. Thomas (Received 15 April 2004; accepted 28 May 2004)
J. Mol. Biol. (2004) 341, 979–989
Characterization and Quantification of Triple Helix Formation in Chromosomal DNA Robert Besch1, Carine Giovannangeli2, Theda Schuh1 Claudia Kammerbauer1 and Klaus Degitz1* 1
Department of Dermatology Ludwig-Maximilians University, Frauenlobstr. 9-11 80337 Mu¨nchen, Germany 2 Laboratoire de Biophysique Muse´um National d’Histoire Naturelle, CNRS UMR5153/INSERM U565 Paris, France
DNA-binding molecules that recognize specific sequences offer a high potential for the understanding of chromatin structure and associated biological processes in addition to their therapeutic potential, e.g. as positioning agents for validated anticancer drugs. A prerequisite for the development of DNA-binding molecules is the availability of appropriate methods to assess their binding properties quantitatively at the desired target sequence in the human genome. We have further developed a capture assay to assess triplex-forming oligonucleotide (TFO) binding efficiency quantitatively. This assay is based on bifunctional, psoralen and biotin-conjugated, TFOs and real-time PCR analysis. We have applied this novel quantification method to address two issues that are relevant for DNA-binding molecules. First, we have compared directly the extent of TFO-binding in three experimental settings with increasing similarity to the situation in vivo, i.e. naked genomic DNA, isolated cell nuclei, or whole cells. This comparison allows us to characterize factors that influence genomic triplex formation, e.g. chromosomal DNA organization or intracellular milieu. In isolated nuclei, the binding was threefold lower compared to naked DNA, consistent with a decreased target accessibility in the nucleosomal environment. Binding was detected in whole cells, indicating that the TFO enters the nucleus and binds to its target in intact cells in vivo, but the efficiency was decreased (tenfold) compared to nuclei. Secondly, we applied the method to characterize the binding properties of two different TFOs targeting the same sequence. We found that an antiparallel-binding GT-containing TFO bound more efficiently, but with less target sequence selectivity compared to a parallel-binding CUcontaining TFO. Collectively, a sensitive method to characterize genomic triplex formation was described. This may be useful for the determination of factors driving TFO binding efficiency and, thus, may improve the usefulness of triplex-mediated gene targeting for studies of chromatin structure as well as for therapeutic antigene strategies. q 2004 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: triplex DNA; gene targeting; oligonucleotides
Introduction Triplex-forming oligonucleotides (TFOs) recognize specific sequences in double-stranded DNA by binding in the major groove of the double helix. Hoogsteen hydrogen bonds are formed between oligonucleotide bases and purine bases in the double helix, which are already engaged in Watson– Abbreviations used: TFO, triplex-forming oligonucleotide. E-mail address of the corresponding author: [email protected]
Crick hydrogen bonds. Triplex formation via Hoogsteen interactions can occur at sequences containing a stretch of pyrimidines on one DNA strand and complementary purines on the other DNA strand (oligopyrimidine$oligopurine sequences). Depending on their base composition, TFOs bind in a parallel or antiparallel orientation to the purine-containing strand.1,2 TFOs and other DNA code-reading molecules (collectively termed antigene molecules) have been used successfully, in a sequence-specific manner, to manipulate DNA-associated functions, such as transcription (initiation or elongation), replication, and repair.2–5
0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
980 In addition, the attachment of functional moieties acting on DNA (e.g. protein domains or DNAdamaging agents), led to advanced strategies in which antigene molecules direct the corresponding activities to specific sequences in the genome. Since DNA in cells is typically bound to histones and tightly packed into chromatin, the binding and activity of antigene molecules must be determined in this context. The demonstration if and how much of a triple helix complex can form at the target sequence inside cells as well as the characterization of cellular factors influencing efficient binding are critical for applications of antigene molecules, whatever their chemical nature. In order to demonstrate the binding to complementary DNA in a cellular context, a variety of methods have been developed, mainly for TFOs. TFOs covalently linked to reactive groups have been intensively used for this purpose, and target modifications reflecting triplex formation have been demonstrated (i) by detection of site-specific genomic modifications caused by triplex-forming oligonucleotides in cells6,7 and in mice,8 (ii) by detection of specific binding of triplex-forming oligonucleotides using either competitive PCR, restriction enzyme protection assay, or primer extension,9–12 or (iii) by capture of triplex structures.13 While these studies demonstrate mostly qualitative detection, a quantitative and sensitive assessment of triplex formation is desirable (i) for the selection of target sites with optimal accessibility for oligonucleotides in the chromosomal context in the cell type of interest; (ii) for the comparison of binding strength to a certain target sequence obtained with various oligonucleotides, including new oligonucleotide modifications; and (iii) for the assessment of target accessibility in different cellular contexts. These are necessary steps for the use of sequence-specific chromatin ligands, e.g. TFOs. In this study, we characterize and quantify the binding of triplex-forming oligonucleotides to their genomic target sequence. We have further developed a capture assay that is based on the use of a bifunctional, psoralen and biotin-conjugated, TFO associated with photoactivation, capture, and real-time PCR analyses: TFO binding efficiency was assessed by quantifying the amount of triplexmodified target sequences compared to the amount of unmodified targets. Here, we have used this quantitative method to address two types of questions. First, we studied how the extent of TFO binding to a determined genomic target changed in three experimental settings with increasing complexity and increasing similarity to the situation in vivo: binding efficiency of TFO to the target sequence was compared directly and quantitatively in naked genomic DNA, in isolated cell nuclei, and in whole cells. Secondly, we evaluated how the extent of TFO binding changed with the nature of the TFO, namely CU or GT-containing TFO.
Triple Helix Formation in Chromosomal DNA
Results Design of triplex-forming oligonucleotides A 5-methylcytosine/5-propynyluracil-containing oligonucleotide (TFOcu) was designed for a previously identified 16 bp oligopyrimidine$oligopurine sequence13 in the third intron of the human ICAM-1 gene14 that contains a neighboring 5 0 -TpA motif suitable for psoralen/UVA-induced photoadduct generation. TFOcu contains 5-methylcytosine (C) and 5-propynyluracil (U) nucleosides allowing formation of T$AxU and C$GxC basetriplets (with the centre dot standing for Watson– Crick interactions and the x for Hoogsteen interactions) after binding in parallel orientation with respect to the purine strand of the doublehelical target sequence (Figure 1). The 5-methylcytosine and 5-propynyluracil nucleosides were chosen because methylation of cytosine increases binding strength at physiologic pH compared to unmodified oligonucleotides, and 5-propynyluracil nucleosides increase binding strength in the presence of low intracellular concentrations of magnesium.15 TFOcu was psoralen-modified at the 5 0 end to allow photoadduct formation after UVA photoactivation and the 3 0 end was biotinmodified to allow capture with streptavidin-coated magnetic beads. A GT-containing triplex-forming oligonucleotide (TFOgt) that binds to the same target sequence in an antiparallel orientation (Figure 1C) was used and has been described elsewhere.13 For the demonstration of sequencespecific binding, two control oligonucleotides were designed for TFOcu: COcu1 (inverted sequence) and COcu2 (scrambled sequence). COgt is a scrambled control for TFOgt.13 Characterization of triplex formation and psoralen cross-linking Specificity and efficacy of triplex-induced photoadduct formation were examined with a restriction enzyme protection assay that is based on the ability of triplex structures to interfere with restriction enzyme cleavage. A plasmid containing the triplex target sequence overlapping with an EcoNI cleavage site (Figure 2A) was incubated with oligonucleotides, UVA-irradiated, purified, and digested with EcoNI and EcoRI. Fragments of 3804 bp, 453 bp and 270 bp length were generated in the absence of oligonucleotides. In contrast, a 723 bp fragment composed of the two smaller fragments was obtained with TFOcu, suggesting that the EcoNI site overlapping the triplex site was protected specifically by the targeted photoadduct (Figure 2B). The control oligonucleotide COcu1 did not prevent restriction enzyme cleavage, supporting the view that the observed inhibition was sequence-specific. At a 30-fold molar excess of TFOcu (2 mM), a 50% inhibition was observed, and almost complete inhibition was noted at a 300-fold molar excess (Figure 2B, upper panel). Less
981
Triple Helix Formation in Chromosomal DNA
formed by TFOcu was assessed in mobility-shift assays (Figure 2C). A 23 bp ICAM-1 DNA fragment containing the target sequence was incubated with TFOcu, COcu1, or COcu2, UVA-irradiated, and subjected to denaturing gel electrophoresis. With increasing UVA doses, both monoadducts (TFOcu covalently linked to one strand of the double helix) and cross-links (TFOcu linked to both strands of the double helix) were detected, whereas no specific photoadduct was detected with the control oligonucleotides COcu1 or COcu2. At a dose of 5 J/cm2 and using an oligonucleotide concentration of 10 mM, approximately 80% cross-links and 20% monoadducts were generated in a typical experiment. Collectively, these data demonstrate the ability of the oligonucleotide TFOcu to interact, and photoreact with, the 16 bp ICAM-1 target sequence in a sequence-specific manner, as demonstrated previously with TFOgt.13 Triplex formation in genomic targets
Figure 1. Target sequence and triplex-forming oligonucleotides. A, The ICAM-1 gene structure with exons (grey) and introns (white). T, target site localized in the third intron (corresponding to positions 556–571 of the ICAM-1 gene sequence (NCBI RefSeq X59288)). P, PstI restriction sites. P1–P6, position of PCR-primers. B, A depiction of the 825 bp PstI-fragment containing the target sequence. In A and B, the lengths of the PCRamplified fragments used in the study are indicated for the corresponding set of primers. C, A 23 bp gene segment including the 16 bp oligopyrimidine$oligopurine target sequence (underlined) and a 5 0 -TpA site suitable for psoralen/UVA cross-links (boxed); sequences of the triplex-forming oligonucleotide TFOcu and the control oligonucleotides COcu1 (inverted sequence) and COcu2 (scrambled sequence), as well as sequences of the triplexforming oligonucleotide TFOgt and its control COgt (scrambled sequence). Oligonucleotides consisted of the base analogues 5-methylcytosine (C) and 5-propynyluracil (U) or of guanine and thymine, respectively, and are conjugated to psoralen (Pso) and biotin (Bio) as indicated.
inhibition was observed when potassium (140 mM) was present in the incubation buffer (Figure 2B, lower panel), suggesting that triplex formation was less efficient in the presence of physiologic intracellular concentrations of potassium. The nature of triplex-induced photoproducts
With regard to the feasibility of antigene applications, we assessed TFO binding to its target locus in the natural context of genomic DNA under different experimental conditions, i.e. in naked DNA and in chromatin. TFOcu was reacted with naked genomic DNA, and triple helices were detected by a capture assay. The assay exploits the properties of bifunctional psoralen-conjugated and biotin-conjugated oligonucleotides. If a triple helix is formed in genomic DNA, the oligonucleotide can be psoralen/UVA cross-linked with the double helix. After digestion by the appropriate restriction enzyme, triple helixcontaining fragments can be captured via binding of the biotin moiety of oligonucleotides to streptavidin-coated magnetic beads and can, subsequently, be identified by PCR amplification of a portion of the captured fragment. A PCR product as evidence for triplex formation was found when TFOcu was reacted with naked genomic DNA (Figure 3A). No PCR fragment was detected when either TFO treatment or irradiation (data not shown) was omitted, which excludes nonspecific, biotin-independent binding of the targeted genomic DNA fragment to the magnetic beads. When the control oligonucleotide COcu1 was used, a very weak PCR signal was detectable. This observation can be explained with the fact that, in addition to the 5 0 -TpA motif, neighboring the target sequence, the captured 825 bp target fragment harbors 11 more 5 0 -TpA motifs, at which photoadditions might be generated by the psoralen groups independently of triplex formation. Sample aliquots not subjected to magnetic capture (“before capture” portion) were also PCR-analyzed, and PCR-signals were generated in all reactions (TFOcu, COcu1, or no oligonucleotide) indicating that the target fragment was gel-extracted and intact before magnetic separation. In a next step, the ability of TFOcu to bind to the
982
Triple Helix Formation in Chromosomal DNA
Figure 3. Triplex formation in naked genomic DNA and isolated cell nuclei. A, Naked genomic DNA was exposed to TFOcu, COcu1 or no oligonucleotide (no oligo), and magnetic capture was carried out (PCR after capture, left panel, lanes 1–3) or not (PCR before capture, right panel, lanes 4–6). B, Isolated A431 nuclei were exposed to TFOcu, COcu1, COcu2, or no oligonucleotide (no oligo), and magnetic capture was carried out (PCR after capture, left panel, lanes 1–4) or not (PCR before capture, right panel, lanes 5–8). C, Isolated A431 nuclei were exposed to TFOcu or COcu1, and magnetic capture was carried out. Coamplification of the triplex-containing target fragment (primers P1 and P2 in Figure 1) with an unrelated DNA fragment (primers P5 and P6 in Figure 1) was as described in Materials and Methods.
Figure 2. Triplex formation in vitro. A, A plasmid containing the target sequence overlapping with an EcoNI recognition site was incubated with a 3–300-fold molar excess (0.2–20 mM) of oligonucleotides, UVAirradiated (5 J/cm2), purified, and digested with EcoNI and EcoRI (see Materials and Methods). The length of the three digestion products (270 bp, 453 bp, and 3804 bp) is indicated. Inhibition of EcoNI by triplex structures results in a longer 723 bp (Z453C270) fragment. B, Incubation of plasmid with oligonucleotides in a buffer containing 10 mM Tris–HCl (pH 7.5), 10 mM MgCl2, 1 mM spermidine (upper panel) or in a buffer containing additionally 140 mM potassium (lower panel). C, A 23 bp fragment containing the target sequence (as shown in Figure 1) was
target in its intact supranucleosomal structure was tested by exposing oligonucleotides to isolated nuclei of A431 cells. When TFOcu was reacted with nuclei, a PCR product was detected (Figure 3B). As observed for naked genomic DNA,
incubated with 10 mM oligonucleotides and irradiated with various dosages of UVA. Photoadducts (a, crosslinks; b, monoadducts) were retarded under denaturing conditions relative to unbound single-stranded target sequence (c).
Triple Helix Formation in Chromosomal DNA
no PCR product was detectable in samples without oligonucleotide, suggesting complete separation. Again, weak PCR signals were detected with the control oligonucleotides, COcu1 and COcu2 (Figure 3B), consistent with triple helix-independent background psoralen cross-links. In an attempt to confirm the assumption of random cross-links in a second and independent fashion, separated DNA was PCR-amplified using (i) primers specific for a 388 bp portion of the target fragment and, simultaneously, (ii) primers specific for an unrelated 799 bp control fragment. This control fragment is located in the first exon and intron of the ICAM-1 gene, contains 27 TpA sites, and is separated from the triple helix target fragment by multiple PstI sites (Figure 1A). Whereas the primers specific for the target fragment produced a strong signal with TFOcu and a weak signal with COcu1, the primers for the 799 bp control fragment produced a weak signal of equivalent strength with both TFOcu and COcu1 (Figure 3C). These results are consistent with the generation of cross-links due to triple helix specific binding and a background amount of triple helix independent photoadditions at 5 0 -TpA motifs randomly distributed throughout the genome. As in the experiments with naked genomic DNA, all “before capture” portions produced PCR signals indicating that the fragments were gel-extracted and intact before magnetic separation (for the 825 bp target fragment, Figure 3B, right panel; for the 799 bp control fragment, data not shown). Collectively, these data suggest that sequencespecific adducts are formed in cell nuclei by TFOcu at the 16 bp oligopyrimidine$oligopurine ICAM-1 target locus, as demonstrated with the TFOgt.13 These results indicate that the target sequence is accessible in the intact supranucleosomal structure; however, no quantitative data were obtained regarding the proportion of target sequences covered by a triplex structure. Quantification of triplex formation in naked genomic DNA In order to further characterize the binding of triplex-forming oligonucleotides to their target sequence, we quantified the amount of triplexmediated cross-links by a sequential PCR procedure combining conventional and real-time PCR (see Materials and Methods, Figure 4A). We tested the propynyl-modified oligonucleotide, TFOcu (parallel binding mode) and compared it to an unmodified GT-containing oligonucleotide, TFOgt (antiparallel binding mode). When genomic DNA (starting amount of 2 mg) was reacted with TFOcu, around 20,000 triplex-modified target fragments were detected (Figure 4B, left panel). A smaller number of target fragments (approximately 2500) was captured with the control oligonucleotide COcu1, which was about an eighth of the amount captured by the triplex-forming oligonucleotide and likely due to random photoadditions at the ICAM-1 fragment independently of the TFO, as
983 mentioned above. When no oligonucleotide was added, an even smaller amount of fragments was captured. Capture of fragments under these circumstances could be due to minimal background binding of target fragments to magnetic beads independently of biotinylation. When genomic DNA was reacted with TFOgt in the same experimental conditions, the amount of triplex-modified target fragments was higher (approximately 65,000, Figure 4B, right panel). A smaller, but still substantial amount of target fragments was captured with the control oligonucleotide COgt (approximately 22,000), which was about threefold less than that captured by TFOgt. In order to assess what percentage of target sequences contained triple helices, we compared the amount of target fragments with triple helices (captured fraction) with the amount of target fragments free from triple helices (contained in the supernatant generated during the magnetic separation). In genomic DNA reacted with TFOcu, 21.2% of targets were separated compared to 2.8% when the control oligonucleotide was used (Figure 4C, left panel). With TFOgt, 70.6% of targets were captured compared to 23.5% when the control oligonucleotide was used (Figure 4C, right panel). Collectively, these experiments quantify the specific capture of target fragments due to triple helix formation in the context of naked genomic DNA. Whereas TFOgt displayed a high binding efficiency compared to TFOcu, it showed less sequence-specific binding in our experimental conditions. Quantification of triplex formation in isolated nuclei We applied this quantification strategy to isolated A431 nuclei exposed to triplex-forming oligonucleotides, in order to examine the influence of chromosomal organization on triplex binding. Due to its more specific binding properties, we selected TFOcu for these further experiments. Similar to the results obtained with genomic DNA, exposure of nuclei to TFOcu resulted in a greater amount (around threefold) of captured target fragments compared to exposure to COcu1 or to no oligonucleotide (Figure 5A). However, the amount of fragments captured from isolated nuclei was threefold lower than the number of fragments captured from naked genomic DNA, with the percentage of target fragments carrying triple helices reduced to 7.9% (Figure 5B). The experimental conditions were otherwise identical for purified genomic DNA and nuclear DNA, and they are not likely to account for the observed difference. Consequently, this decrease may be due to: (i) the nuclear chromatin structure or the nucleosomal organization, which might be an obstacle for oligonucleotides on their way to their genomic target sequence; (ii) a lower efficacy of photoaddition at TpA sequences in the context of chromatin; or (iii) a lower concentration
984
Triple Helix Formation in Chromosomal DNA
Figure 4. Quantification of triplex formation in genomic DNA. DNA was subjected to a ten cycle conventional PCR, purified to remove beads, and subsequently quantified by real-time PCR as described in Materials and Methods. A, Various amounts of genomic DNA (2, 5, 20, 50, and 200 ng) were subjected to this sequential PCR procedure. Realtime cycle numbers are plotted against the amount of template DNA. One representative experiment is depicted. In each experiment, such a set of DNA standard amounts was analyzed along with test samples to create a standard curve for the calculation of the DNA amount in test samples. B, Number of target fragments (meanGSD of three independent experiments) captured from naked genomic DNA that had been treated with oligonucleotides as indicated. C, The ratio between triplex-bound (captured) and triplex-free (supernatant) target fragments (meanGSD of three independent experiments).
of TFO available for triplex formation due to potential trapping by nuclear proteins. Quantification of triplex formation in whole cells In a further step towards the in vivo situation, we quantified triplex formation in cultured cells. A431 cells were transfected with oligonucleotides and UVA-irradiated in order to cross-link psoralenconjugated triplex-forming oligonucleotides. Subsequently, genomic DNA was prepared and cross-linked target fragments were analyzed. When cells were transfected with TFOcu, real-time PCR analysis allowed the detection of triplex formation
of TFOcu in cells (Figure 5). The amount of triplexbearing DNA fragments captured from cells was small compared to isolated nuclei (approximately tenfold decrease) or naked genomic DNA (approximately 30-fold decrease). This could be explained by a reduced intracellular concentration of TFO due to transfection efficiency or due to an intracellular degradation of the TFO. However, the latter possibility is unlikely, since we found that the oligonucleotide was stable under cellular conditions: when TFOcu-transfected cells were cultivated for longer periods, about 50% of the initial amount of oligonucleotide could still be detected 72 hours after transfection (data not shown), and
Triple Helix Formation in Chromosomal DNA
985
Figure 6. Quantification of triplex formation in whole cells. DNA was isolated from cells and subjected to a ten cycle conventional PCR, purified, and subsequently quantified by real-time PCR as described in Materials and Methods. The number of target fragments (meanG SD of three independent experiments) captured from DNA purified from whole A431 cells that had been treated transfected with oligonucleotides is indicated. Figure 5. Quantification of triplex formation in isolated cell nuclei. DNA prepared from nuclei was subjected to a ten cycle conventional PCR, purified, and subsequently quantified by real-time PCR as described in Materials and Methods. A, Number of target fragments (meanGSD of three independent experiments) captured from DNA purified from A431 cell nuclei that had been treated with oligonucleotides as indicated. B, The ratio between triplex-bound (captured) (meanGSD of three independent experiments) and triplex-free (supernatant) genomic target fragments.
there was no degraded oligonucleotide species detectable by gel electrophoresis at all observation points. This 50% decrease may be attributed to the dilution of the oligonucleotide in a population of growing cells. Nevertheless, exposure of cells to TFOcu did result in a substantial amount of photoadducts, which was significantly higher (approximately 3.5-fold) than in samples with the control oligonucleotide COcu1 or without oligonucleotide (Figure 6). Taken together, these data indicate that the dendrimer-compacted TFOcu (i) enters intact A431 cells, (ii) has access to the nucleus, and (iii) binds to its chromosomal target sequence all under conditions of intact cells.
Discussion An important step in the development of sequence-specific DNA ligands is the quantitative assessment of target accessibility in a defined cellular context. To that end, we have developed a quantitative assay based on the use of bifunctional
psoralen-conjugated and biotin-conjugated TFOs associated with photoactivation, magnetic capture, and a PCR procedure. Oligonucleotides were biotin-conjugated to allow magnetic capture of the modified DNA fragment and psoralen-conjugated to allow UVA-induced photoaddition after triplex formation at the target site (Figure 2).13 This covalent cross-linking proved technically useful for the demonstration of triplex formation in various experimental settings, including whole cells. Our approach permits the PCR quantification of triplex-modified target sequences (capture fraction), and of unmodified target sequences (supernatant fraction) allowing the determination of the percentage of triplex-modified targets. Such a direct and sensitive quantitative assay has not been described to our knowledge. Triplex formation and accessibility of target sequences in chromatin had been quantified in a limited number of studies in isolated nuclei9,11 or in whole cells.10,12 However, no quantitative comparison is available concerning triplex formation to a given DNA sequence as a function of DNA environments (namely, naked DNA, isolated nuclei, and whole cells), or of the nature of TFO (CU or GT-containing TFO). Here, we have used our quantitative capture assay to address these issues. In our system, the amount of triplex-bound target sequences was decreased threefold in nuclei compared to naked genomic DNA. Since experimental conditions were kept identical for the two different sources of DNA, this likely reflects the association of the double helix with nucleosomes, which could obscure the oligonucleotide target site and/or the TpA site required for photoadduct formation and subsequent detection in the capture assay. Indeed, it
986 was shown in vitro with histone-associated short DNA fragments that the accessibility of a target sequence varies, depending on its position relative to the histone.16 In addition, the accessibility of a certain target sequence might change, depending on the transcriptional status of the targeted gene,17 and may depend on the cell-cycle, since increased triplex formation during S phase has been demonstrated.18 In our experiments, cells were not synchronized, which reflects an in vivo situation, and the level of ICAM-1 transcription is low:19 these conditions might have caused low target accessibility. Finally, oligonucleotide trapping by nuclear proteins may lead to a decrease in effective oligonucleotide concentration and to a subsequent decrease in oligonucleotide binding to DNA compared to naked DNA. In whole cells, the amount of triplex-carrying target fragments was decreased even further (tenfold compared to isolated nuclei), indicating that there are additional factors beside the nuclear structure that are inhibitory to triplex formation. Several reasons may explain this difference between the level of triplex formation in isolated nuclei and in whole cells: (i) intracellular concentrations of oligonucleotide may be reduced, depending on transfection efficacy and oligonucleotide entrapment in endosomes prior to reaching the nuclear target; (ii) intranuclear ion concentrations in cells, e.g. magnesium or potassium, may be different from the one with the isolated nuclei and less favorable for triplex formation. In this regard, it is noteworthy that the presence of a physiologic concentration of potassium reduced triplex formation in restriction enzyme protection assays (Figure 2). Besides providing a quantitative assessment of triplex formation in different target structures, the assay proved helpful in the comparative evaluation of different triplex-forming oligonucleotides. Here, we characterized triplex formation at the genomic ICAM-1 target site using oligonucleotides that bind to the target sequence in parallel (CU-containing oligonucleotide) or antiparallel (GT-containing oligonucleotide) orientation. In the case of TFOcu, the cytosine nucleotides were 5-methylated and the thymine nucleotides were replaced with 5-propynyluracil, because these modifications improve binding characteristics to double-stranded DNA,15 as well as resistance towards cellular nucleases.20 Indeed, in our system, the modified oligonucleotide was found to be stable after transfection into the keratinocyte cell line A431, since no degradation was noticed. Furthermore, propynyl-modified oligonucleotides were found to penetrate stripped skin, which is an interesting feature with regard to the development of a topical antigene strategy when targeting the ICAM-1 gene.21 The specificity and efficacy of TFO binding was evaluated in restriction enzyme protection and mobility-shift assays, and in capture assays in which we demonstrated TFO target binding in the context of genomic DNA from various sources. When naked genomic DNA was used as a target,
Triple Helix Formation in Chromosomal DNA
the GT-containing oligonucleotide appeared to bind more strongly compared to the CU-containing oligonucleotide (Figure 4B), an observation consistent with the data obtained with plasmid targets (Figure 2B).13 On the other hand, the TFO/CO ratio is higher with the CU-containing oligonucleotide, suggesting that the GT-containing oligonucleotide binds less sequence-specifically compared to the CU-containing oligonucleotide in our experimental conditions (Figure 4B). This is consistent with semiquantitative data for nuclear DNA comparing those two oligonucleotides: the TFO/CO ratio appeared to be higher with the CU-containing oligonucleotide (Figure 3B) than with the GT-containing oligonucleotide.13 Quantitative analysis of chromosomal binding of TFOs is desirable in the development of an antigene strategy to address several issues: (i) it allows assessment of the accessibility of target sequences in a gene of interest and selection of optimal sites for targeting with triplex-forming oligonucleotides; (ii) it permits evaluation of binding characteristics like target affinity and target selectivity of triplexforming oligonucleotides in the genomic context, especially for newly developed oligonucleotide analogues or modifications; (iii) and it is a way to quantitatively investigate the accessibility of a DNA sequence in various cellular situations, e.g. during inflammatory processes, or during the cell-cycle, which may have an influence on the transcriptional status of a targeted gene(s). Indeed, it appears that the chromatin structure is dynamic enough to allow TFO binding. Antigene molecules as sequencespecific chromatin ligands can, therefore, be useful probes of chromatin structure in vivo and of associated regulatory pathways. They may be used to control these processes in a sequencespecific manner. However, some sites may be more accessible to antigene molecules than others. A better understanding of the factors that influence complex formation in the nuclear context will be critical for the improvement of gene-targeting strategies. In this context, the development of methods to quantitatively determine triplex formation in chromatin, such as that described here, will be useful.
Materials and Methods Oligonucleotides The triplex-forming oligonucleotide TFOcu and its control oligonucleotides COcu1 (inverted sequence) and COcu2 (scrambled sequence), and the triplex-forming oligonucleotide TFOgt and its control oligonucleotide COgt were obtained from Eurogentec (Seraing, Belgium) or Qbiogene (Heidelberg, Germany); the sequences are displayed in Figure 1. The 5 0 or 3 0 end was modified with 4,5 0 ,8-trimethylpsoralen via a six carbon linker and the other end was conjugated to biotin. Oligonucleotides for PCR were obtained from Metabion (Mu¨nchen, Germany).
987
Triple Helix Formation in Chromosomal DNA
UVA treatment
Mobility-shift assays
UVA was applied with a PUVA 200 light arch (Waldmann, Villingen-Schwenningen, Germany) with F8T5 PUVA bulbs, whose emission spectrum is mainly between 315 nm and 365 nm. Irradiation was done at room temperature through window glass in order to eliminate traces of UVB. The irradiation dose delivered to the samples was determined by a UV-meter (Waldmann) and the fluence was 2 mW/cm2 at a 19 cm source to target distance.
Triplex formation and photoadduct formation in vitro was additionally analyzed in mobility-shift assays. A 23 bp double-stranded DNA fragment representing the ICAM-1 target locus containing the 16 bp oligopyrimidine$oligopurine sequence (Figure 1C) was produced by annealing equimolar amounts of complementary singlestranded oligodeoxynucleotides. Both 3 0 -ends were labelled with digoxigenin (DIG Gel Shift Kit, Roche). Labelled target DNA (30 fmol) was incubated with 10 mM oligonucleotides in 10 ml of triplex-forming buffer. After incubation for 90 minutes at 37 8C, samples were either not irradiated or irradiated with various doses of UVA. TFOs bound to the target duplex during irradiation were linked covalently to the duplex at the 5 0 TpA psoralen acceptor site via psoralen/UVA photoreactions, and the resulting covalent complexes were assessed as mobility shifts by denaturing gel electrophoresis (16% polyacrylamide gel containing 7 M urea; 10% (v/v) glycerine in 44.5 mM Tris (pH 8.3), 44.5 mM boric acid, 1 mM EDTA).
Cell culture The human squameous cell carcinoma-derived cell line A431 was from American Type Culture Collection, Rockville, MD. Cells were cultured in DMEM supplemented with 2 mM L-glutamine, 100 units/ml of penicillin, 100 mg/ml of streptomycin, 1 mg/ml of amphotericin B (all from Life Technologies, Karlsruhe, Germany), and 10% (v/v) heat-inactivated fetal calf serum (ccpro, Neustadt, Germany), at 37 8C and 5% (v/v) CO2. Cells were kept from confluency. UVA irradiation of cells was done in PBS at room temperature.
Intracellular stability of oligonucleotides A431 cells were transfected with 3 mM TFOcu and Superfecte (Qiagen, Hilden, Germany) as described.13 DNA was prepared 24 hours, 48 hours, and 72 hours after transfection by cell lysis, extraction with phenol/chloroform, and precipitation in ethanol. Purified DNA was electrophoresed on a denaturing 16% (w/v) polyacrylamide gel, blotted to nylon membranes, and detected via its biotin moiety with the Biotin Luminescent Detection Kit (Roche Molecular Biochemicals, Mannheim, Germany).
Restriction enzyme protection assay Specific triplex formation was evaluated in vitro with a restriction enzyme protection assay. The 16 bp ICAM-1 target sequence was included in a 28 bp sequence to create an EcoNI recognition site that overlaps with the 16 bp oligopyrimidine$oligopurine ICAM-1 target sequence (Figure 2A). This sequence was inserted into a 4.5 kb plasmid that had been constructed from the vectors pCAT-Basice and pSV beta Gale (Promega, Mannheim, Germany). Plasmid DNA (2 mg; 0.66 pmol) was incubated with increasing amounts of oligonucleotides for 90 minutes at 37 8C in 10 ml of a triplex-forming buffer (10 mM Tris–HCl (pH 7.5), 10 mM MgCl2, 1 mM spermidine). Reactions were UVA-irradiated (5 J/cm2) in order to covalently link TFOs bound to the target duplex at the 5 0 TpA psoralen acceptor site via psoralen/UVA photoreactions. DNA was then purified and, subsequently, digested with EcoNI and EcoRI (New England Biolabs, Beverly, USA) at 37 8C overnight. Photoadduct formation was necessary to retain triple helix integrity during DNA purification, and purification had to be performed, because the triplex-forming buffer interferes with EcoNI enzyme activity. The extent of inhibition of EcoNI cleavage by triplex-mediated photoadducts was assessed by gel electrophoresis and quantified densitometrically (GelDoc system, BioRad, Mu¨nchen, Germany).
Triplex formation in naked genomic DNA Genomic DNA isolated from A431 cells was digested with PstI (Roche) resulting in an 825 bp ICAM-1 fragment containing the triplex target sequence (Figure 1B). After restriction, 2 mg of genomic DNA was incubated with various oligonucleotides (10 mM) in 10 ml of triplexforming buffer for 90 minutes at 37 8C. UVA irradiation (5 J/cm2) was performed to allow photoadduct formation. Oligonucleotides bound to the ICAM-1 target sequence were separated from unbound oligonucleotides by agarose gel electrophoresis, and fragments sized between approximately 600 bp and 1000 bp (including the 825 bp target fragment) were extracted from gels. This separation step was necessary prior to the magnetic capture, because unbound oligonucleotides would compete with target-bound oligonucleotides for streptavidin on magnetic beads. Aliquots of each sample were: (i) saved (termed “before capture” portion) for control of successful gel extraction of the 825 bp ICAM-1 target fragment; and (ii) subjected to magnetic separation (termed “after capture” portion). Triplex formation in isolated nuclei Nuclei were prepared by incubation of 3!106 A431 cells in a buffer containing 0.32 M sucrose, 10 mM Tris– HCl (pH 7.5), 5 mM MgCl2, 0.5% (v/v) Triton X-100, for ten minutes on ice. Nuclei were then washed in the same buffer, transferred to triplex-forming buffer, incubated with various oligonucleotides (10 mM) for 90 minutes at 37 8C in a 100 ml reaction volume with gentle resuspension to avoid sedimentation of nuclei, and UVAirradiated (5 J/cm2). The integrity of nuclei was checked microscopically. DNA was prepared using the DNeasy kit (Qiagen), digested with PstI, and gel-isolated as described above. Aliquots of each sample were either saved as “before capture” portions or subjected to magnetic separation (“after capture” portions). Triplex formation in cells Cells were transfected with oligonucleotides using Superfecte (Qiagen) as described previously13 with modifications: Oligonucleotides (in 1950 ml of serum-free DMEM without antibiotic and antimycotic supplementation) were mixed with Superfecte (50 ml). The final
988
Triple Helix Formation in Chromosomal DNA
oligonucleotide concentration was 10 mM. After 15 minutes incubation at room temperature, the transfection mixture was added to HBSS-washed cells at 70% confluency (approximately 900,000 cells per 6 cm well). Cells were incubated for five hours. For cross-link formation, cells were washed, transferred to PBS, and irradiated with UVA (5 J/cm2) at room temperature. After irradiation, cells were trypsinized, harvested, and genomic DNA was prepared utilizing anion-exchange-resin columns (Blood & Cell Culture Mini Kit, Qiagen). DNA (2 mg) was digested with PstI and gel-extracted. Aliquots of each sample were subjected to magnetic separation. Magnetic capture Triple helix structures from naked genomic DNA, isolated nuclei, or cells were separated by capture of oligonucleotides via their biotin moieties using streptavidin-coated magnetic beads.13 To reduce unspecific DNA binding, magnetic beads (Dynabeads, Dynal, Hamburg, Germany) were preincubated with salmon testes DNA (5 mg/ml; Sigma-Aldrich, Deisenhofen, Germany) in 2 M LiCl for one hour at 37 8C. Then 10 mg of magnetic beads was incubated with 2 mg of genomic DNA from various sources in a 50 ml reaction volume containing 2 M LiCl and 2 mg/ml of salmon testes DNA for 30 minutes at 37 8C. Separation of beads was carried out with a magnetic separator (MPC-E, Dynal) for two minutes. Beads were washed in order to remove nonbiotinylated DNA fragments (three washes with 300 ml of redistilled water for five minutes at 60 8C). Subsequently, beads were resuspended in 50 ml of redistilled water, and aliquots were subjected to PCR. For the analysis of target fragments that do not carry triplex-forming oligonucleotides and, therefore, do not carry biotin and cannot be captured, the supernatant remaining after magnetic separation and the subsequent washes were collected, purified (PCR Purification Kit, Qiagen), and subjected to PCR. In experiments involving quantification of target fragments in supernatants, salmon testes DNA was omitted during the magnetic separation procedure (see above) due to interference with fluorescence detection of real-time PCR. PCR reactions
fluorescence detection of real-time PCR products, the DNA was preamplified in a conventional PCR followed by a purification step to remove beads. More precisely, aliquots of resuspended beads harboring captured DNA were subjected to a ten cycle PCR with 250 nM primers P3 and P4 P3 : 5 0 K CCAACCTCACCGTGGTGCTGCT K 3 0 P4 : 5 0 K CCCACCTTCTCCCTGCTGGCTT K 3 0 which amplify a 242 bp stretch within the 825 bp ICAM-1 fragment. PCR conditions were 1 mM MgCl2, 1.5 units Taq polymerase (peqlab), 60 seconds at 94 8C, 60 seconds at 63 8C, 90 seconds at 72 8C. PCR amplificates were purified (PCR Purification Kit, Qiagen) in order to remove beads. Aliquots were then analyzed by real-time PCR in a LightCyclere (Roche) apparatus using the FastStart DNA Master SYBR Green I Kit (Roche) for fluorescence detection of PCR products. PCR conditions were: 500 nM primers P3 and P4, 2 mM MgCl2, 35 cycles with 15 seconds at 95 8C, ten seconds at 63 8C, 15 seconds at 72 8C, 15 seconds at 84 8C. The last temperature-step was implemented to reduce fluorescence produced by primer dimers. Specific amplification of the 242 bp fragment was controlled by melting curve analysis22 and gel electrophoresis. The amount of DNA was calculated by comparison with standards (2–200 ng of PstI-digested genomic A431 DNA) that were subjected to PCR analysis in the same way as the test reactions, i.e. conventional PCR, purification and real-time PCR. Real-time cycle numbers (defined as the crossing point calculated as the second derivative maximum by the Light cycler software) were found to be proportional to the amount of template DNA (Figure 4A). The number of target fragments in each sample was calculated from the DNA amounts determined by PCR assuming that 1 ng of human genomic DNA contains 280 haploid genomes and, therefore, 280 target fragments. The number of target fragments was measured both in the captured fraction and in the supernatant, which together represent the total amount of DNA subjected to capture. The sum of target fragments in the captured portion and in the supernatant was determined for each sample and used for normalization among all samples allowing the comparison of genomic, nuclear, and cellular DNA sources.
PCR-amplification of captured DNA was performed with 250 nM primers P1 and P2: P1 : 5 0 GAACTGGCACCCCTCCCCTCTT 3 0 P2 : 5 0 K CCGGGGCCACACCCATCTCAAA K 3 0 specific for a 388 bp portion of the 825 bp PstI target fragment that contains the triplex target-site (for primer locations see Figure 1). Reaction conditions were: 1.0 mM MgCl2, 1.5 units of Taq polymerase (peqlab, Erlangen, Germany), 32 cycles with 30 seconds at 94 8C, 30 seconds at 63 8C, 30 seconds at 72 8C. In some experiments, an unrelated 799 bp fragment of the ICAM-1 gene was coamplified in the same reaction tube as a control using 250 nM primers P5 and P6 P5 : 5 0 K CCAGTCGACGCTGAGCTCCTCT K 3 0 P6 : 5 0 K CCTCTTCCCTAGCTGGGCTGGA K 3 0 For a quantitative and more sensitive assessment of the percentage of captured DNA, a real-time PCR procedure was developed. Since magnetic beads interfere with the
References 1. Thuong, N. T. & He´le`ne, C. (1993). Sequence-specific recognition and modification of double-helical DNA by oligonucleotides. Angew. Chem. Int. Ed. Engl. 32, 666–690. 2. Faria, M. & Giovannangeli, C. (2001). Triplex-forming molecules: from concepts to applications. J. Gene. Med. 3, 299–310. 3. Uil, T. G., Haisma, H. J. & Rots, M. G. (2003). Therapeutic modulation of endogenous gene function by agents with designed DNA-sequence specificities. Nucl. Acids Res. 31, 6064–6078. 4. Ebbinghaus, S. W. (2003). Site-selectiveDNA binding drugs. Chem. Biol. 10, 895–897. 5. Dudouet, B., Burnett, R., Dickinson, L. A., Wood, M. R., Melander, C. Belitsky, J. M. et al. (2003). Accessibility of nuclear chromatin by DNA binding polyamides. Chem. Biol. 10, 859–867. 6. Vasquez, K. M., Wang, G., Havre, P. A. & Glazer, P. M.
Triple Helix Formation in Chromosomal DNA
7.
8. 9.
10.
11.
12.
13.
14.
(1999). Chromosomal mutations induced by triplexforming oligonucleotides in mammalian cells. Nucl. Acids Res. 27, 1176–1181. Vasquez, K. M., Dagle, J. M., Weeks, D. L. & Glazer, P. M. (2001). Chromosome targeting at short polypurine sites by cationic triplex-forming oligonucleotides. J. Biol. Chem. 276, 38536–38541. Vasquez, K. M., Narayanan, L. & Glazer, P. M. (2000). Specific mutations induced by triplex-forming oligonucleotides in mice. Science, 290, 530–533. Giovannangeli, C., Diviacco, S., Labrousse, V., Gryaznov, S., Charneau, P. & Helene, C. (1997). Accessibility of nuclear DNA to triplex-forming oligonucleotides: the integrated HIV-1 provirus as a target. Proc. Natl Acad. Sci. USA, 94, 79–84. Belousov, E. S., Afonina, I. A., Kutyavin, I. V., Gall, A. A., Reed, M. W. Gamper, H. B. et al. (1998). Triplex targeting of a native gene in permeabilized intact cells: covalent modification of the gene for the chemokine receptor CCR5. Nucl. Acids Res. 26, 1324–1328. Faria, M., Wood, C. D., Perrouault, L., Nelson, J. S., Winter, A. White, M. R. et al. (2000). Targeted inhibition of transcription elongation in cells mediated by triplex-forming oligonucleotides. Proc. Natl Acad. Sci. USA, 97, 3862–3867. Oh, D. H. & Hanawalt, P. C. (1999). Triple helixforming oligonucleotides target psoralen adducts to specific chromosomal sequences in human cells. Nucl. Acids Res. 27, 4734–4742. Besch, R., Giovannangeli, C., Kammerbauer, C. & Degitz, K. (2002). Specific inhibition of ICAM-1 expression mediated by gene targeting with triplexforming oligonucleotides. J. Biol. Chem. 277, 32473–32479. Degitz, K., Li, L. J. & Caughman, S. W. (1991). Cloning and characterization of the 5 0 -transcriptional regulatory region of the human intercellular adhesion molecule 1 gene. J. Biol. Chem. 266, 14024–14030.
989 15. Lacroix, L., Lacoste, J., Reddoch, J. F., Mergny, J. L., Levy, D. D. Seidman, M. M. et al. (1999). Triplex formation by oligonucleotides containing 5-(1-propynyl)-2 0 -deoxyuridine: decreased magnesium dependence and improved intracellular gene targeting. Biochemistry, 38, 1893–1901. 16. Brown, P. M. & Fox, K. R. (1998). DNA triple-helix formation on nucleosome-bound poly(dA)*poly(dT) tracts. Biochem. J. 333, 259–267. 17. Macris, M. A. & Glazer, P. M. (2003). Transcription dependence of chromosomal gene targeting by triplex-forming oligonucleotides. J. Biol. Chem. 278, 3357–3362. 18. Majumdar, A., Puri, N., Cuenoud, B., Natt, F., Martin, P. Khorlin, A. et al. (2003). Cell cycle modulation of gene targeting by a triple helix forming oligonucleotide. J. Biol. Chem. 278, 11072–11077. 19. Caughman, S. W., Li, L. J. & Degitz, K. (1990). Characterization and functional analysis of interferon-gamma-induced intercellular adhesion molecule-1 expression in human keratinocytes and A-431 cells. J. Invest. Dermatol. 94, 22s–26s. 20. Wagner, R. W., Matteucci, M. D., Lewis, J. G., Gutierrez, A. J., Moulds, C. & Froehler, B. C. (1993). Antisense gene inhibition by oligonucleotides containing C-5 propyne pyrimidines. Science, 260, 1510–1513. 21. White, P. J., Fogarty, R. D., Liepe, I. J., Delaney, P. M., Werther, G. A. & Wraight, C. J. (1999). Live confocal microscopy of oligonucleotide uptake by keratinocytes in human skin grafts on nude mice. J. Invest. Dermatol. 112, 887–892. 22. Ririe, K. M., Rasmussen, R. P. & Wittwer, C. T. (1997). Product differentiation by analysis of DNA melting curves during the polymerase chain reaction. Anal. Biochem. 245, 154–160.
Edited by J. O. Thomas (Received 15 April 2004; accepted 28 May 2004)