Monitoring DNA triplex formation using multicolor fluorescence and application to insulin-like growth factor I promoter downregulation


  • Nadia Hégarat,

    1. Acides nucléiques: dynamique, ciblage et fonctions biologiques, INSERM U565, Paris, France
    2. Département Régulations, développement et diversité moléculaire, MNHN - CNRS UMR7196, Paris, France
    Search for more papers by this author
  • Darya Novopashina,

    1. Laboratory of RNA Chemistry, Institute of Chemical Biology and Fundamental Medicine, Siberian Division of Russian Academy of Sciences, Novosibirsk, Russia
    Search for more papers by this author
  • Alesya A. Fokina,

    1. Laboratory of RNA Chemistry, Institute of Chemical Biology and Fundamental Medicine, Siberian Division of Russian Academy of Sciences, Novosibirsk, Russia
    Search for more papers by this author
  • Alexandre S. Boutorine,

    1. Acides nucléiques: dynamique, ciblage et fonctions biologiques, INSERM U565, Paris, France
    2. Département Régulations, développement et diversité moléculaire, MNHN - CNRS UMR7196, Paris, France
    Search for more papers by this author
  • Alya G. Venyaminova,

    1. Laboratory of RNA Chemistry, Institute of Chemical Biology and Fundamental Medicine, Siberian Division of Russian Academy of Sciences, Novosibirsk, Russia
    Search for more papers by this author
  • Danièle Praseuth,

    1. Acides nucléiques: dynamique, ciblage et fonctions biologiques, INSERM U565, Paris, France
    2. Département Régulations, développement et diversité moléculaire, MNHN - CNRS UMR7196, Paris, France
    Search for more papers by this author
  • Jean-Christophe François

    Corresponding author
    1. Acides nucléiques: dynamique, ciblage et fonctions biologiques, INSERM U565, Paris, France
    2. Département Régulations, développement et diversité moléculaire, MNHN - CNRS UMR7196, Paris, France
    3. Sorbonne Universités, UPMC Univ Paris 06, UMR_S 938, CDR Saint Antoine, Paris, France
    4. Faculté de Médecine and Hôpital Saint Antoine, INSERM, UMR_S 938, CDR Saint Antoine, Paris, France
    • Correspondence

      J.-C. François, INSERM – UPMC UMRS938, CDR Saint Antoine, Faculté de Médecine Pierre et Marie Curie, Room 1107, 27 rue Chaligny F-75571 Paris 12, France

      Fax: +33 1 40 01 13 43

      Tel: +33 1 40 01 13 31


    Search for more papers by this author


Inhibition of insulin-like growth factor I (IGF–I) signaling is a promising antitumor strategy and nucleic acid-based approaches have been investigated to target genes in the pathway. Here, we sought to modulate IGF-I transcriptional activity using triple helix formation. The IGF-I P1 promoter contains a purine/pyrimidine (R/Y) sequence that is pivotal for transcription as determined by deletion analysis and can be targeted with a triplex-forming oligonucleotide (TFO). We designed modified purine- and pyrimidine-rich TFOs to bind to the R/Y sequence. To monitor TFO binding, we developed a fluorescence-based gel-retardation assay that allowed independent detection of each strand in three-stranded complexes using end-labeling with Alexa 488, cyanine (Cy)3 and Cy5 fluorochromes. We characterized TFOs for their ability to inhibit restriction enzyme activity, compete with DNA-binding proteins and inhibit IGF-I transcription in reporter assays. TFOs containing modified nucleobases, 5-methyl-2′-deoxycytidine and 5-propynyl-2′-deoxyuridine, specifically inhibited restriction enzyme cleavage and formed triplexes on the P1 promoter fragment. In cells, deletion of the R/Y-rich sequence led to 48% transcriptional inhibition of a reporter gene. Transfection with TFOs inhibited reporter gene activity to a similar extent, whereas transcription from a mutant construct with an interrupted R/Y region was unaffected, strongly suggesting the involvement of triplex formation in the inhibitory mechanisms. Our results indicate that nuclease-resistant TFOs will likely inhibit endogenous IGF-I gene function in cells.




insulin-like growth factor I




triplex-forming oligonucleotide


Insulin-like growth factor I (IGF-I) and its cognate receptor IGF-1R play key roles in the physiology of vertebrate organisms, namely in growth, development, metabolism and stress resistance. Binding of IGF-I to IGF-1R activates downstream pathways that regulate cell growth, proliferation, apoptosis and differentiation [1]. The insulin/IGF-I pathway has been linked to aging and longevity in a variety of species, from nematodes and insects to mammals. Lifespan was significantly extended in mice with liver-specific IGF-I gene inactivation [2] or with heterozygous IGF-1R knockout [3]. Because deregulation of the IGF hormone axis also stimulates tumorigenesis, several molecules aimed at inhibition of IGF-1R signaling are being developed, including receptor-blocking antibodies and specific tyrosine kinase inhibitors [1]. Various nucleic acid-based knockdown strategies that induce partial inhibition of gene expression were used to decipher IGF gene function, for instance, by blocking IGF-I or IGF-1R translation in tumor cells [4-7]. We recently demonstrated that small-interfering RNA treatment that partially downregulates IGF-1R inhibited tumor growth in vivo [7].

The discovery that double-stranded DNA can form triple helices with a third strand led to the development of antigene strategies using triplex-forming oligonucleotides (TFOs) [8]. TFOs were initially designed to interfere with the binding of transcription factors or RNA polymerase to targeted gene promoters and thereby to inhibit transcription [9]. TFOs can be transfected or produced by cells to inhibit gene expression [10-12]. Blocking of translation through triplex formation on the RNA strand was recently proposed [5]. TFOs have also been used to induce mutagenesis through error prone DNA repair pathways and recombination [13, 14]. One of the requirements for triplex formation is the recognition of a purine-rich sequence on the target strand. Upon binding, base triplets are formed by Hoogsteen or reverse Hoogsteen bonds between the third strand and the purine-rich strand in the major groove of the target. Stable triplex formation occurs essentially through DNA recognition of three-two-base motifs in the third strand, TC (pyrimidine), GA (purine) and GT motifs [9]. Extended oligopurine/oligopyrimidine (R/Y) sequences in the target DNA are therefore a prerequisite for stable triplex formation. Fortunately, such triplex-target sites are overrepresented in promoter regions [15].

In this study, we investigated whether chemically synthesized TFOs can be used to silence IGF-I gene expression in tumor cells, through inhibition of transcription. Transcription from the IGF-I gene is driven by a tandem promoter and generates multiple transcripts from alternative transcription initiation sites, differential splicing and variable polyadenylation [16]. The P1 promoter is responsible for 70% of IGF-I mRNA production in liver [17]. TFOs potentially compete with RNA polymerase or transcription factors that modulate IGF-I transcription, although the regulatory proteins in these TATA-less promoters have not yet been fully characterized [16, 18-21]. One major drawback of nucleic acid-based molecules for intracellular applications is their rapid degradation. Therefore, TFOs are generally chemically modified with terminal blocking groups or moieties with unnatural backbones or sugars [9, 22]. These modifications generally increase nuclease resistance, but some decrease triplex stability [9]. We previously identified a 23-bp R/Y TFO binding sequence in the IGF-I P1 promoter that was targeted in vitro by unmodified triplex-forming oligonucleotides [23-25]. Here, we showed that nuclease-resistant pyrimidine-rich TFOs containing 5-methyl-2′-deoxycytidine and 5-propynyl-2′-deoxyuridine base modifications known to increase binding affinities [26] were able to recognize the IGF-I promoter sequence under physiological conditions. These TFOs inhibited restriction enzyme cleavage close to the R/Y sequence. We developed a new multicolor fluorescence detection system to monitor in vitro triplex formation and protein binding to IGF-I DNA. Each of the three strands was labeled with a different fluorochrome (Alexa488, Cy3 and Cy5), and thus each strand could be monitored independently. We also showed that modified pyrimidine-rich TFOs inhibited expression of an IGF-I reporter gene, consistent with the role of the R/Y sequence in transcriptional IGF-I promoter activity.


Identification of a 23-bp oligopurine–oligopyrimidine sequence in IGF-I promoter

A 23-bp R/Y sequence with similarity to a GAGA motif has been identified in the rat IGF-I P1 promoter (chr7:28,528,238-28,528,260, ENSEMBL Release 71) [27]. The region containing the 23-bp sequence is evolutionary highly conserved (Fig. S1), suggesting an important role in IGF-I regulation. We designed two TFOs targeted to the IGF-I R/Y sequence, one in the TC motif and the other in the GA motif (Fig. 1A). To achieve triplex formation within the TC motif at physiological pH, we designed pyrimidine-rich TFOs (22PM, 22PM-1, 22PM-3) that bind in a parallel orientation to the purine strand (R) of IGF-I DNA and contain 5-methyl-2′-deoxycytidine (M) and 5-propynyl-2′-deoxyuridine (P) modified residues. We also designed a purine-rich TFO (23GA) to bind in an antiparallel orientation to the purine strand of IGF-I DNA. Triplex formation with this GA motif is dependent on multivalent cations such as Mg2+, whereas triplex with TC motif is stabilized at slightly acidic pH.

Figure 1.

Triplex-forming oligonucleotides bind to an R/Y sequence in the IGF-I promoter. (A) Structure of rat IGF-I promoter region. Transcription is controlled by alternative promoters P1 and P2 located in exons 1 and 2, respectively. The first transcription initiation site (+1) and the two major start sites of P1 (+50, +149) are indicated by arrows. The 23-bp R/Y sequence in P1 was targeted by TFOs with either a pyrimidine motif (22PM) or a purine motif (23GA). TFOs bind in the major groove parallel (22PM) or antiparallel (23GA) to the purine strand of the duplex. Positions of PCR primers (fp1, rp2) generating a 1009 bp fragment are indicated. M, 5-methyl-2′-deoxycytidine; P, 5-propynyl-2′-deoxyuridine. (B) Design of restriction enzyme-dependent cleavage assay to monitor TFO binding to IGF-I R/Y sequence. A 1009 bp PCR fragment obtained from P1 contains two EarI sites at positions +657 and +754. The site at position +657 overlaps the triplex-binding site, whereas the site at +754 does not. Sequences of pyrimidine-rich oligophosphodiesters 22PM-1 and 22PM-2 (upper case) and oligophophorothioates 22PM-3 and 22PM-4 (lower case) are presented. Underlined trinucleotides were changed to CTC/GAG in the mutated PCR fragment. (C) Schematic representation of products obtained by EarI digestion of the 1009-bp PCR fragment in presence of oligonucleotides. The 357-bp fragment was not observed under the conditions used to monitor triplex-mediated inhibition of EarI cleavage. (D) Triplex formation with 22PM-1 inhibits cleavage of the IGF-I P1 fragment. Increasing concentrations of TFOs are incubated with target and EarI. Control 22PM-2 (reverse sequence of 22PM-1) has three mismatches with the target sequence. m1, 100 bp DNA ladder.

Analysis of triplex formation within the TC motif of IGF-I promoter

To monitor triplex formation on the IGF-I 23-bp R/Y sequence, we incubated a PCR fragment containing this target with TFOs and digested it with restriction nuclease EarI that recognizes 5′-CTCTTC-3′/3′-GAGAAG-5′ and cleaves both strands asymmetrically just outside the recognition sequence (Fig. 1B). The pyrimidine-rich 22PM-1 TFO inhibited EarI cleavage of the restriction site located within the 23-bp R/Y sequence, as shown by disappearance of 652- and 108-bp products of EarI digestion (Fig. 1C,D). This cleavage at position 657 was not blocked by prior incubation with the control oligonucleotide 22PM-2 (reverse sequence of 22PM-1). By contrast, cleavage at position 754 was not affected by the presence of either 22PM-1 or 22PM-2. Using gel shift, we found that the oligophosphorothioate 22PM-3 containing 5-methyl-2′-deoxycytidine and 5-propynyl-2′-deoxyuridine formed less stable triplexes than the oligophosphodiester 22PM-1. The concentration of TFO at which 50% of the triplex is formed (C50) was 1 μm for 22PM-3 and 0.3 μm for 22PM-1 (Fig S2). Like 22PM-1, 22PM-3 specifically inhibited EarI cleavage at the triplex-binding site, whereas 22PM-4 (reverse sequence of 22PM-3) did not (Fig. S3). Cleavage of a DNA fragment mutated within the triplex-binding site was not affected by 22PM-3 TFO, strongly suggesting that inhibition of EarI was due to specific triplex formation.

Use of fluorescent probes to monitor triplex formation

As an alternative to EMSA, we developed a fluorescence-based methodology to detect triplex formation by three strands, each conjugated to a different fluorochrome. As a double-stranded target we chose the [+37, +79] region of the IGF-I P1 promoter and labeled R and Y strands at their 5′-ends with Cy3 and Cy5 dyes, respectively (Fig. 2A). The third strand TFO was linked to Alexa488. Detection of the three strands in the polyacrylamide gel was performed using a fluorescence scanner. A combination of filter sets and laser sources was used to minimize spectral cross-talk between fluorochromes A488 and Cy3 (Fig. 2B). It should be noted that when large amounts of A488 and Cy5-conjugates were analyzed, cross-talk was detected in the Cy3 channel (an example is given in Fig S4). We efficiently reduced the impact of this cross-talk through two steps. First, cross-talk was substantially diminished by loading no more than 1 pmol of fluorochrome-labeled oligonucleotide per well and by limiting photomultiplier tube settings to 400 V. Second, we used fluorsep software to measure bleed-through for each channel and fluorochrome, and automatically eliminated cross-talk based on these results. This procedure was used in all experiments.

Figure 2.

Use of three-colored fluorescent probes to monitor triple helix formation. (A) Fluorescent oligonucleotides. The 42-bp duplex fragment corresponding to the rat IGF-1 P1 sequence was fluorescently labeled by conjugation of Cy3 to the 5′-end of the purine-rich strand (Cy3–42R) or by conjugation of Cy5 to the 5′-end of the pyrimidine-rich strand (Cy5–42Y). Purine- and pyrimidine-rich TFOs were 3′-end-labeled with Alexa488 (23GA–A488, 22PM–A488). (B) Normalized emission (solid lines) and absorption (dashed lines) spectra of Alexa488 (dark blue, 488 nm), Cy3 (green, 532 nm) and Cy5 (red, 633 nm) oligonucleotide conjugates. Excitation wavelengths are indicated by arrows; ranges of emission filters are highlighted. (C) Chemical structures of the fluorescence-labeled oligonucleotides. Dyes are connected to oligonucleotides through trimethylene (Cy3, Cy5) or hexamethylene (A488) linkers. (D) Gel-retardation assay to analyze triple helix formation. Increasing concentrations of 23GA–A488 TFO were incubated with doubly labeled fluorescent duplex, Cy3–42R/Cy5–42Y (0.1 μm) in triplex-binding buffer (pH 7.2) at 37 °C. Gels were scanned for Cy5, Cy3 and Alexa488 fluorescence. Merged channels show single-, double- and triple-stranded structures: 23GA–A488 (blue), Cy3–42R/Cy5–42Y duplex (yellow) and Cy3–42R/Cy5–42Y/23GA–A488 triplex (white). (E) Standard gel-retardation assay using radiolabeled duplex 42R*/42Y and 23GA–A488 TFO (42R* indicates 5′-end-labeled strand).

Dyes were coupled to oligonucleotides through a C3 linker for cyanine-conjugates and a C6 linker for A488 (Fig. 2C) to avoid interference with triplex formation. On the basis of previous studies, we expected that cyanine fluorophores attached through a C3 linker at the 5′-terminus would be stacked at the end of the double helix like an additional base pair, minimizing interaction with the Alexa conjugate [28]. As shown in Fig. 2D, the addition of 23GA–A488 TFO to Cy3–42R/Cy5–42Y target resulted in a mobility shift from duplex to triplex that was dependent on TFO concentration. The three channel settings were then used to follow triplex formation. With Cy5 channel, only fluorescence of Cy5–42Y strand present in duplex and triplex species was observed. Similarly, the fluorescence of Cy3–42R was detected in duplex and triplex bands in Cy3 channel. The detection of third strand 23GA–A488 fluorescence in the A488 channel showed a strong increase in band intensity at the expected location of the triplex due to increasing concentrations of A488 conjugate. Within the merged image, the Cy3–42R/Cy5–42Y duplex was yellow, and the triplex structure Cy3–42R/Cy5–42Y/23GA–A488 was white, whereas the free 23GA–A488 was blue. C50 values determined from Cy3 or Cy5 channel intensities were similar (0.43 μm based on Cy3 channel data, 0.41 μm for Cy5), confirming that triplex formation could be monitored by analysis of either duplex strand (Fig. 2D). A similar C50 was obtained when triplex formation with 23GA–A488 was monitored using a radiolabeled duplex target (Fig. 2E). Note that for radiolabeled EMSA, we used 10 nm duplex with 10–500-fold excess TFOs, compared with 100 nm fluorescent duplex with 1–50-fold excess of TFO (Fig. 2D,E). This duplex concentration was found to be optimal and allowed appropriate estimation of the dissociation constant. To find out whether results from all three channels could be used to evaluate C50, we fitted data with the assumption that one oligonucleotide-A488 is bound per Cy3–42R/Cy5–42Y duplex (Doc. S1). Dissociation constants were determined using an oligonucleotide ligand concentration in excess of the duplex target (Fig. S5). Dissociation constants obtained for pyrimidine-rich 22PM–A488 were found similar in all three channels (Fig. S5A). By contrast, dissociation constants for 23GA–A488 from Alexa488 and Cy5 data differed significantly, suggesting that the proposed model did not fit triplex formation with 23GA–A488 (Fig S5B). Indeed, C50 for 23GA–A488 using A488 data could not be determined using equilibrium defined as D + O ⇌ T (Doc. S1). This is likely to be due to trapping of TFO in homoduplex structures, a feature that we reported previously for GA-rich oligonucleotides [25]. Indeed, 23GA–A488 oligonucleotide is able to form GA-homoduplexes just like its unconjugated counterpart 23GA (Fig. S6). It should be noted that both GA-oligonucleotides formed homoduplexes at 0.1 μm. We hypothesized that this GA-homodimerization competed with triplex formation and consequently impacted on dissociation constant determination (Fig. S5).

We then studied the binding of pyrimidine- and purine-rich A488 conjugates and unconjugated TFOs to double- and mono-labeled fluorescent duplexes, Cy3–42R/Cy5–42Y, Cy3–42R/42Y, 42R/Cy5–42Y (Fig. 3 and Table 1). Multiple colored bands were obtained depending on the structures formed (Fig. 3E). Triplex structures formed by purine TFO 23GA–A488 were more stable under physiological conditions (pH 7.2 and 37°C) than triplexes formed with DNA pyrimidine TFO 22TC–A488 with unmodified nucleobases (compare Fig. 3A–D for TC-TFO and Fig. 3F–I for GA-TFO). We also showed that the presence of the Alexa dye at the 3′-end decreased TFO affinity for Cy3–42R/Cy5–42Y and 42R/Cy5–42Y duplex targets to some extent (Table 1; Cy5 strand used to monitor triplex formation). This effect was more pronounced with purine-rich TFO (23GA) than with pyrimidine-rich TFO containing 5-methyl-2′-deoxycytidine and 5-propynyl-2′-deoxyuridine (22PM).

Table 1. Stabilities of triplexes determined by fluorescence-based gel electrophoresis assay. Gels were analyzed by monitoring Cy5 fluorescence, except for the 42R/42Y duplex which was 5′ end-radiolabeled and analyzed in storage phosphor mode. 23GA and 22PM oligophosphodiesters carried an amino C6 linker at their 3′-end.
Target duplexTFOC50m)
Figure 3.

Triple helix formation with fluorescently labeled oligonucleotides. (A–D, F–I) Gel-retardation analyses of duplexes Cy3–42R/Cy5–42Y, Cy3–42R/42Y and 42R/Cy5–42Y (labeled Cy3–R/Cy5–Y, Cy3–R/Y and R/Cy5–Y, respectively) incubated with (A–D) 5 μm pyrimidine-rich TFO (22TC–A488, labeled TC) or (F–I) 0.5 μm purine-rich TFO (23GA–A488, labeled GA). Gel analysis was performed as described in the legend to Fig. 2 (A, G, Cy5; B, H, Cy3; C, I, Alexa488). Merged images (D, F) are presented showing multicolored single-, double- and triple-stranded structures. O, T and D represent unbound TFO, DNA triplex and duplex structures, respectively. (E) Schema showing all combinations of TFO–A488 (blue) with fluorescent duplexes, Cy3–42R/Cy5–42Y (yellow), Cy3–42R/42Y (green) and 42R/Cy5–42Y) (red). In triplex structures, the color of duplexes is added to TFO blue resulting in white (Cy3–42R/Cy5–42Y/TFO–A488), cyan (Cy3–42R/42Y/TFO–A488) or magenta (42R/Cy5–42Y/TFO–A488).

Use of fluorescent probes to monitor protein binding

We then investigated whether the fluorescence-based method could be used to detect protein–DNA interactions. For that purpose, we studied protein binding to double-labeled duplex Cy3–42R/Cy5–42Y using several different cell extracts. Extracts and duplex were incubated under various conditions and complexes were analyzed by nondenaturing PAGE. Figure 4 shows a representative experiment performed with double-labeled duplex incubated with HeLa nuclear extracts in a triplex-binding buffer [29]. Several proteins were bound to the [+37, +79] fragment of the IGF-I promoter (Fig. 4, lane 2), in accord with studies showing that fragments around the R/Y sequence are recognized by rat liver and human nuclear proteins [30, 31]. Although proteins were not definitively identified, we were able to draw some conclusions from our data. Two types of shifted bands were observed: those corresponding to proteins that recognize both strands of target duplex (Fig. 4, lane 2, yellow bands), and those corresponding to proteins recognizing only the purine-rich strand Cy3 single-strand of the target (green bands). The latter likely unwound the duplex and are proteins with helicase-like activity.

Figure 4.

Binding of nuclear proteins to the IGF-I promoter revealed by fluorescent probes. Nuclear protein extracts (NE) were coincubated with fluorescent-labeled TFO 23GA–A488, double-labeled DNA duplex (R/Y for Cy3–42R/Cy5–42Y) or triple-labeled DNA triplex (Cy3–42R/Cy5–42Y/23GA–A488). Gels were scanned as described in the legend to Fig. 2. Merged images show multicolored shifted bands in the presence of protein extracts depending on which strands were recognized by proteins. The presence (+) or absence (−) of duplex (10 nm), TFO (2 μm) or HeLa NE (2 μg·μL−1) in the reaction is indicated. HeLa proteins bound to fluorescent probes are represented schematically as hexagons. W indicates locations of sample wells. Note that in the presence of proteins, unbound 23GA–A488 was partially cleaved into smaller products (lower band in lanes 4 and 6).

We then analyzed the binding of proteins to 23GA–A488 (Fig. 4, lane 6). We observed several bands and the strongest likely corresponded to binding by single-strand DNA-binding proteins that are abundant in nuclear extract. We next examined protein binding to the triplex structure formed by 23GA–A488 and Cy3–42R/Cy5–42Y (Fig. 4, lane 4). The yellow bands in Fig. 4 (lane 4) corresponded to proteins that recognize both strands of the double-stranded target. White bands indicated that all three strands comigrated within the protein. Although all three strands are present in these complexes, this does not necessarily reflect binding of proteins to the triplex structure, because proteins may recognize the double-stranded regions flanking the sequence bound to TFO (represented schematically in Fig. 4). Some proteins that are present in the band corresponding to triplex may be bound to single-stranded 23GA–A488 (based on comparison with migration of the blue band in lane 6). The comparison between proteins recognizing duplex and triplex probes demonstrated that binding of certain proteins was affected by addition of the third strand, suggesting that triplex formation blocked their binding to the IGF-I duplex. The green band recognizing purine-rich Cy3–42R (Fig. 4, lane 2) disappeared in lane 4, suggesting either competition with the purine-rich strand of 23GA–A488 or inhibition of the unwinding activity of helicases due to triplex formation [32]. In addition, certain proteins bound to bands containing all three strands (Cy3–42R/Cy5–42Y/23GA–A488) were displaced by an excess of unlabeled duplex 42R/42Y, but not by unlabeled 23GA, indicating that they were specific for the double-stranded target (Fig. S7).

Role of oligopurine–oligopyrimidine sequence in the IGF-I promoter

We showed that the [+37, +79] region of the IGF-I promoter bound several nuclear proteins that might be transcription factors. We then investigated the role of the R/Y sequence in promoter activity using deletion mutagenesis coupled with transient transfection assays. Previous work demonstrated that long R/Y stretches positively or negatively impact promoter strength [33-35]. As shown in Fig. 5A, deletion of the 23-bp R/Y sequence in the rat IGF-I P1 promoter decreased transcription by 48%. A 45% inhibition of human IGF-I transcription was previously observed after deletion of a larger region corresponding to the [+13, +57] region in the rat IGF-I promoter [30].

Figure 5.

Essential role of 23-bp R/Y sequence in the rat IGF-I promoter 1. (A) Deletion analysis of the IGF-I P1 promoter. Transcriptional activity of mutated IGF-I promoters pIGF–1711∆Pu4, pIGF–1711∆Pu23 and pIGF–1711mut are compared with activity from wild-type pIGF–1711/luc. Quantitative analysis from five independent experiments with mean ± SEM. ***< 0.001; ns, not significant. (B) Effect of TFOs on wild-type and mutated rat IGF-I promoter activity. pIGF–1711/luc or pIGF–1711mut were cotransfected with TFOs (22PM-1 and 22PM-3) or control oligonucleotide (CONT). Quantitative analysis of three independent experiments with mean ± SEM. ***< 0.001; ns, not significant.

We next investigated whether TFOs designed to bind to the IGF-I R/Y sequence were able to inhibit IGF-I transcription in cells. For that purpose, we cotransfected a reporter plasmid pIGF–1711/luc containing the luciferase gene under the control of the rat IGF-I P1 promoter with TFOs or controls. The pyrimidine-rich triplex-forming oligophosphorothioate, 22PM-3, inhibited IGF-I transcription by 51%, whereas control oligonucleotides showed no inhibition (Fig. 5B). To demonstrate triplex-mediated inhibition of IGF-I transcription, we mutated pIGF–1711/luc to obtain control plasmid pIGF–1711/lucMut carrying an interrupted R/Y unable to bind TFOs. Because it was previously found that nuclear proteins specifically recognized the region on the 5′ side of this R/Y sequence [30, 31], the replacement of a trinucleotide in this sequence should not affect promoter transcriptional activity, as shown in Fig. 5A. The absence of inhibition after cotransfection of control plasmid pIGF–1711/lucMut with 22PM-3 strongly suggested that inhibition observed with the rat IGF-I promoter was due to triplex formation [36, 37]. Interestingly, triple helix formation on the IGF-I gene reduced transcription to 51%, similar to results obtained by deleting the triplex-binding site. This inhibition was obtained exclusively with nuclease-resistant TFOs, because phosphodiester 22PM-1 did not significantly decrease IGF-I transcriptional activity.


Because single-stranded oligonucleotides can recognize double-stranded DNA sequences [9], it was speculated that the triple helical structure might interfere with gene transcription or replication by competing for binding with regulatory proteins. Already widely used as tools in molecular biology, TFOs have potential as therapeutic agents in a so-called antigene strategy [9]. This antigene approach shares features with other nucleic acid-based approaches to gene silencing like antisense oligonucleotides and small-interfering RNAs. Oligonucleotides bind to genes or mRNAs with high affinity and specificity but suffer from nuclease sensitivity and limited cell delivery [38]. Because the triple helical approach requires purine-rich target sequences, it is possible to increase the triplex formation repertoire using chemically modified nucleotide components [39]. Current efforts are focused on increasing triplex stability and TFO nuclease resistance by designing new compounds with modified nucleobases and sugars [9, 40, 41].

Here, we showed that a nuclease-resistant TFO bound to an IGF-I promoter fragment, interfered with protein binding and inhibited IGF-I transcription. Inhibition required triplex formation as suggested by the absence of an inhibitory effect with mutated target and control oligonucleotides. It has been described that full phosphorothioate oligonucleotides formed triplexes with lower binding affinities compared with phosphodiester oligonucleotides, depending on the motifs [36]. We replaced thymine and cytosine of phosphorothioate TFO with 5-propynyl-2′-deoxyuridine and 5-methyl-2′-deoxycytidine to compensate for the loss of binding affinity observed with pyrimidine-rich oligophosphorothioates, because these modified nucleobases stabilize base triplets [26, 42]. Although the stability of the triplex structure with phosphorothioate 22PM-3 TFO was lower than that of phosphodiester 22PM-1, it was sufficient to inhibit restriction enzyme activity and block IGF-I transcription (Fig. S2).

We also developed a new method to check triplex formation in vitro based on differential fluorescence detection of TFO and purine- and pyrimidine-rich target strands after gel electrophoresis analysis. Oligonucleotides were conjugated to Alexa488, Cy3 or Cy5 fluorochromes, respectively, and were detected simultaneously. We were able to determine the concentration at which 50% of the triplex was formed (C50) by monitoring either Cy3 or Cy5 channels, corresponding to the fluorescence of 42R- and 42Y-labeled target strands. Comparison with results from standard gel-shift assays using radiolabeled targets suggested that dyes at both ends of the duplex did not significantly alter binding of TFO–A488 (Table 1). Therefore, our assay format made it possible to monitor triplex formation by measuring fluorescence in Cy3 or Cy5 channels. The same is possible in A488 channel with the label used for TFO, provided that no competing TFO complexes are formed (Figs S5 and S6). Our method was also validated with other triplex motifs (Fig. 3 and Table 1).

The approach has some drawbacks. The fluorochrome triad A488, Cy3 and Cy5 was suboptimal, because it was necessary to use fluorsep software to correct for bleed-through from each channel. We found that a high concentration of Cy5-labeled oligonucleotide emitted at shorter wavelength that interfered with A488 and Cy3 channels, likely due to an hypsochromic effect of H-aggregates or dimers [43] (Fig. S4). Cross-talk due to TFO–A488, which emitted in Cy3 channel due to its high concentration, was eliminated by fluorsep treatment (data not shown). Formation of a more stable triplex structure using next generation TFOs would mitigate this issue. Designing a covalent intermolecular triplex would be advantageous because a TFO excess would no longer be necessary. However, we preferred to focus on reversible intermolecular triplexes for the detection of sequence-specific proteins and to use TFO to displace these proteins under physiological conditions. We are convinced that an appropriate combination of three fluorochromes with nonoverlapping excitation and emission spectra will be useful, provided that dye linkers do not interfere with triplex formation. A combination of fluorochromes (Cy3, Cy5, Cy7) with fully separated emission peaks as recently described in a single-molecule three-color assay could be useful [44]. However, such an approach requires the availability of corresponding fluorescence imager configuration. In our assay, we chose C6- and C3-linkers for TFOs and 42-mer duplex strands, respectively. With this configuration, the strands were visualized independently in three-stranded complexes, demonstrating that the three-color approach in gel electrophoresis can be useful for the quantification of triplex formation. It should be mentioned that in order to obtain useful signal-to-noise ratios in all three fluorescence channels, we performed fluorescent EMSA with 100 nm fluorescent duplexes, which is a high concentration compared with radionuclide-based EMSA. Although this was taken into account in the dissociation constant determination, we still need to be cautious about the accuracy of these estimated dissociation constants (Kd). Fitted values generally depend on experimental conditions and the binding model, but conditions with fluorescent target concentrations > 100 nM can be used to evaluate dissociation constants [45]. Here, we undertook quantification to compare data from the third color channel monitoring TFO with data from two other channels monitoring the duplex, resulting in the estimated dissociation constant. Resonance energy transfer (FRET) of fluorescent probes was also used to monitor triplex formation in solution [45-48]. Because A488/Cy3, A488/Cy5 and Cy3/Cy5 pairs have been used in FRET assays [49-51], it would be interesting to investigate whether any FRET occurs in the gel-retardation assay. This information could eventually be used to evaluate triplex formation. We think that, in our experiment, no FRET occurred between Cy3 and Cy5 dyes due to the distance between the fluorochromes in the 42-bp duplex. However, we cannot exclude the possibility of FRET between A488 and Cy3– or Cy5–oligonucleotide conjugates when the triplex was formed. Note that Förster radius values for A488/Cy5, Cy3/Cy5 and A488/Cy3 pairs are 4.9, 5.3 and 6.7 nm, respectively [49-51].

Collectively, our data show that triplex formation can be monitored using multicolor fluorescence of three strands. Each strand can be visualized individually in the absence or presence of proteins and quantification is conceivable depending on the experimental conditions and provided stable triplex is formed. Nevertheless, quantification of triplex formation using third strand color is restricted by features that depend on the specific DNA triplex formed. In the case of IGF-I constructs, GA–TFO homodimerization interferes with triplex formation, as previously described by us and others [25, 52]. Triplex formation is inherently limited by binding affinity, obliging to use third-strand DNA in excess to achieve efficient binding.

We applied our methodology to detect nucleic-acid-binding proteins recognizing the [+37, +79] IGF-I P1 promoter fragment. Using the double-labeled duplex Cy3–42R/Cy5–42Y in triplex-binding conditions, we observed several shifted bands, confirming the profile obtained with radiolabeled duplex [30]. Further studies are needed to identify proteins that bind to this region [20]. The fluorescent labeling of both strands of the target duplex facilitated the detection of proteins that were bound to only one strand (e.g. green bands) or to both strands (yellow bands). The presence of a third fluorescent strand, the TFO, impaired some protein binding to the IGF-I promoter fragment. Competition with specific duplex target strongly decreased nuclear protein binding to three-stranded structures (Fig. S5). Addition of 23-GA did not inhibit protein binding to the three-stranded target. Together, these results suggest that the observed shifted bands correspond to proteins bound to the double strand fragment. Our data do not exclude that proteins may bind directly to triple helical structures [8]. Because several endogenous proteins recognize triple helical structures, it would be interesting to apply our method to monitor the binding of such proteins in vitro on triple helical structures with each of the three strands labeled by fluorochromes. Our fluorescent gel-shift assay could be used to confirm binding of known triplex-binding proteins (for review see [8]), but also allow the identification of new proteins that recognized particular triplex structures, e.g. purine- or pyrimidine-rich RNA third-strand bound to DNA duplex [10-12]. Isolation of these proteins using our method would be facilitated through the use of covalent triple helical structure. Further characterization of proteins bound to these DNA substrates is the next important step and will necessitate use of complementary approaches such as supershift assay, southwestern blots or affinity purification [53].

Oligopurine–oligopyrimidine sequences are observed in known promoter and regulatory regions. Nevertheless, their role in regulating transcriptional activity was not systematically explored. In the case of the rat IGF-I promoter, the 23-bp R/Y sequence played a role in transcription, likely through binding of transcription factor in its vicinity, because deletion led to 48% transcription inhibition. This conserved sequence certainly represents an important feature of the IGF-I promoter. The presence of alternating d(GA)6 repeats and the 23GA sequence both appear to be important. Indeed a long d(GA/CT)n dinucleotide sequence may adopt an intramolecular triple helical structure (H-DNA) that potentially has a role in transcriptional regulation [54]. We demonstrated that the 23-bp R/Y sequence found in IGF-I promoter region bound TFOs in vitro. We observed specific inhibition of transcription from the IGF-I promoter by TFO cotransfected into hepatocarcinoma cells. The 51% decrease observed with nuclease-resistant oligophosphorothioate TFO in transient transfection assays showed that triplex formation interfered with IGF-I transcription. Because deletion of the R/Y sequence did not completely abolish transcriptional activity, it was not unexpected that the TFO did not completely inhibit IGF-I promoter-mediated transcription. To potentiate the activity of the TFO, we might induce irreversible modifications on the IGF-I gene using reagents that are linked to TFO that would cleave, cross-link or photodamage the DNA to induce mutagenic lesions [9, 14, 55]. For instance, covalent attachment of camptothecin derivatives to the phosphodiester 22PM-1 strongly increase its inhibitory effect on IGF-I expression due to recruitment of topoisomerase I that induced cleavage at triplex-binding site [56].


We showed that modified TFOs bind to the 23-bp R/Y sequence in the IGF-I promoter in vitro and in transfected cells. TFO binding competed with restriction enzyme activity in the vicinity of the triplex. We developed a novel gel-shift assay where each strand is labeled with a different fluorochrome that proved useful for characterizing specific DNA-binding proteins recognizing single-, double- or triple-stranded structures. In transfected cells, nuclease-resistant TFOs targeting the P1 region of the promoter resulted in similar transcriptional inhibition to deletion of the polypurine region. Although the inhibition achieved with TFOs was moderate, it could be useful in modulating IGF-I gene expression to produce interesting knockdown phenotypes such as increased longevity or reduction in tumorigenicity [2, 3, 7].

Experimental procedures

Oligonucleotide synthesis and labeling

Unmodified and amino-modified oligodeoxynucleotides were synthesized by Eurogentec S.A. (Seraing, Belgium). 23GA–A488, 22TC–A488 and 22PM–A488 (analogs to 22TC–A488, but with modified nucleobases 5-methyl-2′-deoxycytidine and 5-propynyl-2′-deoxyuridine) were obtained by conjugating 23GA-NH2, 22TC-NH2 and 22PM-NH2 bearing 3′ amino C6 linker to Alexa488 carboxylic acid succinimidyl ester using the Alexa Fluor Oligonucleotide Amine labeling kit (Molecular Probes, Life Technologies, Grand Island, NY, USA). Briefly, 400 μg oligo-NH2 were incubated overnight at 37°C with 600 μg Alexa488 carboxylic acid succinimidyl ester in Hepes buffer pH 9. Purification of Alexa488 conjugates was performed through HPLC using linear acetonitrile/water step gradients (5–50%). Oligonucleotides linked to A488 were then separated from nonconjugated A488 by gel filtration on BioSpin 6 Columns (Bio-Rad, Hercules, CA, USA). Purity was evaluated by UV-shadowing and fluorescence analysis after PAGE (8% acrylamide, 8 m urea). Cy3 and Cy5 oligonucleotide conjugates were obtained from Eurogentec. Purity was evaluated by gel electrophoresis after 3′-end radiolabeling using [32P]ddATP[αP] and terminal deoxyribonucleotidyl transferase (New England Biolabs, Beverly, MA, USA). Concentrations of unmodified and fluorescence-labeled oligonucleotide conjugates were determined using molar absorption coefficients calculated as described previously [57] (Cy3, E260 = 4930 cm−1·m−1; Cy5, E260 = 10000 cm−1·m−1). The extinction coefficient for Alexa488 is 71000 cm−1·m−1 at 492 nm, with a correction factor of 0.30 for absorbance readings at 260 nm, as specified in the manufacturer's protocol.

Plasmids and DNA fragments

pIGF–1711/luc contained the promoter (from −1711 to +1) and the 5′-UTR (position +1 to +328) of rat IGF-I upstream of the coding sequence of firefly luciferase [17]. Mutations in the R/Y sequence of IGF-I P1 were introduced by site-directed mutagenesis using the Transformer™ kit (Clontech, Palo Alto, CA, USA) based on the unique restriction site elimination method using two primers, 5′-GCCTTCTCGAGCTCTCCCTCTTC-3′ and 5′-CGTAAG AGCTCGGGCCCTCCCGGGTTATGTTAGC-3′. Underlined sequences were inserted in pIGF–1711/luc to give pIGF–1711Mut construct. A unique XhoI site was introduced in the middle of R/Y sequence (Fig. 1). pIGF–1711∆Pu4 was obtained after treatment of XhoI-digested pIGF–1711Mut by mung bean nuclease (New England Biolabs). In pIGF–1711∆Pu23, the R/Y sequence was deleted by inverse PCR on XhoI linearized pIGF–1711Mut using two primers equipped with BglII sites (underlined) for circularization after PCR 5′-GAAGATCTTCGAGTAAGGACTTTTTTGGGC-3′, 5′-GAAGATCTTCGAATGTTCCCCCAGCTGTTTCC-3′. In all final constructs, regions of the IGF-I P1 and the luciferase gene were sequenced. The 1009 bp fragment for restriction enzyme protection assay was prepared by PCR with Pfu DNA polymerase (Promega, Madison, WI, USA) using pIGF–1711/luc or pIGF–1711Mut as templates and two primers (Fig. 1A): fp1, 5′-AATGGGAAATAGTGTGTGCC-3′ and rp2, 5′-CTGTTGGTAAAATGGAAGACGCCAA-3′. The pGL2basic (Promega) contained the firefly luciferase gene (luc) without promoter or enhancer.

Restriction enzyme protection assay

The specificity of triplex formation was determined in vitro using a restriction enzyme protection assay. The 1009 bp PCR fragment (250 ng) corresponding to the IGF-I P1 region of pIGF–1711/luc (−608, +400) was incubated with TFOs in 10 μL of triplex-forming buffer (10 mm bis–Tris–propane–HCl (pH 7), 10 mm MgCl2, 100 mm KCl, 1 mm dithiothreitol, 1 mg·mL−1 BSA) and digested with the restriction enzyme EarI (5 U; New England Biolabs) at 37 °C for 30 min. Cleavage reactions were stopped by addition of 3 μL EDTA (0.25 m). Samples were subsequently incubated with 1 μL proteinase K (10 mg·mL−1) and 1 μL SDS 1% for 10 min at 37 °C and electrophoresed on 1.5% agarose gel in TBE 0.5× (45 mm Tris pH 8, 45 mm borate, 0.5 mm EDTA). Gels were quantified using imagej software (NIH, Bethesda, MD, USA) after densitometry on G:BOX Chemi XT4 (Syngene, Cambridge, UK). The relative abundance of 652 bp fragment indicated the extent of cleavage of EarI site at position 657 overlapping the triplex-binding site (Fig. 1B,C). The specificity of triplex-mediated inhibition of EarI was demonstrated by the absence of cleavage inhibition of the second restriction site at position 754 that led to 249 and 108-bp DNA fragments both in presence and absence of TFO (data not shown).

Electrophoretic mobility shift assays

To monitor triplex formation by gel-retardation assays, 42 nucleotide duplexes were incubated at 37 °C for 16 h with increasing concentrations of TFOs in a buffer containing 20 mm Hepes pH 7.2, 10 mm MgCl2, 100 mm NaCl, 0.1 μg·μL−1 tRNA and 10% sucrose. Nondenaturing 10% PAGE (19 : 1) was performed at 37 °C in 10 mm MgCl2 and 50 mm Hepes, pH 7.2. Triplex formation was analyzed using radiolabeled duplexes (10 nm) that were 5′-end-labeled using [32P]ATP[γP] (GE Healthcare Life Sciences, Piscataway, NJ, USA) and T4 polynucleotide kinase (New England Biolabs). Fluorescence-labeled duplexes (100 nm) were also used to evaluate triplex formation in similar conditions as used for radiolabeled duplexes. In this case, duplex formation was obtained mixing equal amount of 42-nucleotide cyanine–oligonucleotide conjugates (500 pmol) in a 20 mm Hepes buffer (pH 7.2) containing 50 mm NaCl. Samples were heated for 5 min to 90 °C and cooled gradually to room temperature overnight before subsequent incubation with TFOs. After electrophoresis, gels were scanned for radioactivity or for fluorescence using a Typhoon™ 9410 imager (GE Healthcare, Chalfont ST Giles, UK). All gel-shift quantification was performed using imagej software. As an estimate of apparent dissociation constant, we used the concentration at which 50% of the triplex was formed (C50), which was the oligonucleotide concentration that resulted in a shift of 50% of the targets from double-stranded to triple-stranded state, based on gel-retardation assays.

To analyze the effect of TFOs on proteins bound to IGF-I P1 fragment, fluorescent duplexes (100 nm) were first incubated for 2 h at 37 °C with TFO–Alexa488 (2 μm) in a 20 mm Hepes buffer pH 7.2 containing 10 mm MgCl2, 10% sucrose and 0.1 μg·μL−1 tRNA. After incubation, three-stranded probes were added to HeLa protein nuclear extracts (2 μg·μL−1) for 20 min at 23 °C in 10 μL containing 25 mm Hepes buffer pH 7.9, 50 mm KCl, 1 mm MgCl2, 10% glycerol, 0.5 mm dithiothreitol and 0.2 μg·μL−1 poly(dI-C) [29]. Products were resolved by nondenaturing 5% PAGE using 0.5× TBE as the running buffer. Fluorescent gel-shift assays with nuclear extract were repeated several times, and the pattern shown in Fig. 4 was reproduced three times.

Fluorophore selection and fluorescence imaging

Fluorochromes Alexa488, Cy3 and Cy5 were chosen for use with a Typhoon 9410 Phosphorimager/Fluorescence Imager equipped with three fluorescence light sources and three emission filters. To evaluate overlap between emission and excitation spectra of oligonucleotide conjugates, their fluorescence emission and excitation spectra were recorded on a Spex Fluoromax3 instrument (Jobin-Yvon Horiba, Chilly-Mazarin, France) at 25°C. Fluorescence emission and absorption spectra were collected using a bandwidth of 5 nm and 0.2 × 1 cm2 quartz cuvettes containing 600 μL solution. Images from gels were taken using a +3 mm focal plane. For fluorescence detection, we used a blue light argon ion laser (488 nm) with a short pass filter (526SP) for Alexa488; a green solid-state doubled frequency SYAG light laser (532 mm) and a narrow band-pass emission filter (580BP30) for Cy3; a red helium/neon light laser (633 mm) and a narrow band-pass emission filter (670BP30) for Cy5. We used photomultiplier tube settings of 400 V for scanning as this provides acceptable background. Under this condition, the Cy3 channel detected in gel ~ 1 pmol of Cy5–conjugate and A488–conjugate (Fig. S4). Moreover, fluorescence of > 10 pmol of Cy5–oligonucleotide was also visualized in the A488 channel and > 1 pmol in the Cy3 channels, whereas 100 pmol of Cy3– or 10 pmol of A488–conjugates were undetectable in the Cy5 channel. The corresponding 16-bit images obtained after recording the fluorescence of one gel into three different files were separated using fluorsep 2.2 software (GE Healthcare). Little cross-talk was observed between Cy3, Cy5 and A488 channels after fluorsep separation if 1 pmol oligonucleotide conjugate was used (Fig. S4). The resulting images were colored and merged using imagej. Blue, green and red colors were attributed to A488, Cy3 and Cy5 channels, respectively. The colors of composite image were then adjusted to optimize brightness and contrast.

Cell culture, transfection and luciferase assays

The LFCL2A rat hepatocarcinoma cell line was maintained as described [6]. These cells were co-transfected with plasmids [3.3 μg·mL−1 of Photinus pyralis luciferase (F-luc) containing constructs and 0.06 μg·mL−1 pRL-Tk that express Renilla reniformis luciferase under the control of thymidine kinase promoter] and oligonucleotides (0.07 μg·mL−1, 10 nm) with GS3815 cytofectin (10 μg·mL−1; Glen Research, Sterling, VA, USA). Luciferase assays measuring F-luc and R-luc activities were performed 24 h after transfection using the dual-luciferase assay kits (Promega) [6]. Relative luciferase activity was calculated by dividing F-luc by R-luc activities allowing normalization of transfection efficiency. All transfection assays were performed four or more times, and each consisted of independent quadruplicates. All transfection data are presented as means ± SEM and were analyzed using ANOVA followed by Dunnett post hoc test. As reference for this test, we used results from pIGF–1711/luc transfected cells (Fig. 5A) or from cells transfected with pIGF–1711/luc and control oligonucleotide CONT (dA15-NH2) (Fig. 5B). Values of P < 0.05 were considered significant.


We thank Dr Peter Rotwein for providing pIGF–1711/luc, Dr Christiane Frayssinet for the rat hepatocarcinoma LFCL2A cells, Drs Marina B. Gottikh, Laurent Lacroix and Martin Holzenberger for useful discussions and Sophie Polo, Dorothée Raoux and Mathieu Bernardelli for technical assistance. This study was supported by INTAS (Grant 03-51-5281) and Hubert Curien partnership 18908VK to J-CF. NH was supported by Ministère délégué à l'enseignement supérieur et à la recherche and Ligue Nationale Contre le Cancer.