Exploring cellular activity of locked nucleic acid-modified triplex-forming oligonucleotides and defining its molecular basis.

Triplex-forming oligonucleotides (TFOs), as DNA-binding molecules that recognize specific sequences, offer unique potential for the understanding of processes occurring on DNA and associated functions. They are also powerful DNA recognition elements for the positioning of ubiquitous molecules acting on DNA, such as anticancer drugs. A prerequisite for further development of DNA code-reading molecules including TFOs is their ability to form a complex in a cellular context: their binding affinities must be comparable to those of DNA-associated proteins. To reach this goal, chemically modified TFOs must be developed. In this work, we present triplex-forming properties (kinetics and thermodynamics) and cellular activity of G-containing locked nucleic acid-modified TFOs (TFO/LNAs). In conditions simulating physiological ones, these TFO/LNAs strongly enhanced triplex stability compared with the non-modified TFO or with the pyrimidine TFO/LNA directed against the same oligopyrimidine.oligopurine sequence, mainly by decreasing the dissociation rate constant and conferring an entropic gain. We provide evidence of their biological activity by a triplex-based mechanism, in vitro and in a cellular context, under conditions in which the parent phosphodiester oligonucleotide did not exhibit any inhibitory effect.

The design of synthetic non-protein molecules able to control DNA-associated biological functions through their interaction with a specific DNA sequence represents a very attractive approach. Among DNA code-reading molecules, triplex-forming oligonucleotides (TFOs) 1 are able to bind to the major groove of oligopyrimidine⅐oligopurine regions in doublestranded DNA. A series of results have validated triplex-based approaches at the molecular and cellular levels: triplexes have been shown to interfere with transcription (initiation and elongation), replication, repair, and recombination (1,2). However, the intracellular efficiency of TFOs still has to be improved.
One possible approach consists of increasing the stability of the non-covalent triple helices under physiological conditions to reach binding affinities comparable to those of DNA-associated regulatory proteins. To this end, a variety of chemically modified nucleic acids have been developed. Among them are locked nucleic acids (LNAs) that contain LNA nucleotide monomers, i.e. ribonucleotides with a 2Ј-O,4Ј-C-methylene linkage that effects conformational fixation of the furanose ring in a C3Јendo conformation (3,4). LNA-containing oligonucleotides have been recently shown to enhance triplex stability and to alleviate in part the sequence constraints imposed by the triple helical recognition motifs (5,6). Only a few hybridization properties of LNA-modified TFOs (TFO/LNAs) have been reported so far, and they all concern (T,C)-containing TFO/LNAs. It has been shown that fully modified TFO/LNAs failed to bind to double-stranded DNA (7,8), likely due to conformational restraint of TFO/LNA. However, alternating DNA and LNA nucleotides in TFO sequences is appropriate for efficient triplex formation. A study on a 15-mer pyrimidine TFO/LNA provided evidence that LNA-induced triplex stabilization is associated with a slower dissociation rate constant and a less unfavorable entropic contribution compared with the non-modified TFO (9). In previous works, TFO/LNAs have been used only for in vitro assays, including plasmid functionalization (10). However, LNAmodified oligonucleotides are appealing molecules: they have been successfully used as efficient antisense agents, even in vivo (for examples, see Refs. [11][12][13], and also as decoys (14), aptamers (15), LNAzymes (16), and DNA-correcting agents (17).
In the present work we report the triplex forming properties and intracellular activity of G-containing TFO/LNAs. We characterized the mechanism of triplex stabilization induced by LNA modifications in G-containing TFOs, compared with nonmodified isosequential phosphodiester TFOs and with pyrimidine TFO/LNA directed against the same oligopyrimidine⅐ oligopurine target sequence. The dependence of triplex stability on pH was also explored. To address these questions, kinetic and thermodynamic parameters of triplexes were evaluated by surface plasmon resonance. Specificity of TFO/LNA binding was also analyzed, using UV melting experiments, electrophoretic mobility shift assays, and restriction enzymatic protection assay. Finally, we evaluated the capacity of G-containing TFO/LNAs to interfere with biological processes in vitro and in cells, using experimental settings designed to demonstrate a triplex-based mechanism in a quantitative manner. We showed that G-containing TFO/LNAs were active, as measured by inhibition of transcription elongation. Our results support TFO/LNA activity in a cellular context at submicromolar concentrations, which has never been reported before, and represent the first step toward further development of TFO/LNAs as artificial modulators of DNA-associated biological functions.

MATERIALS AND METHODS
Oligonucleotides-Oligonucleotide analogues with LNA residues were synthesized as previously described (3) or purchased from Proligo (France SAS). Phopodiester oligonucleotides were synthesized by Eurogentec; for cellular experiments, the phosphodiester oligonucleotide (15TCG sequence, see Fig. 1) was 3Ј-modified by incorporation of a propylamine group (named 15TCG*/po) to resist nuclease-mediated degradation.
Plasmids-The pSP-F47 plasmid was constructed by insertion of a 780-bp fragment of the human immunodeficiency virus type 1/nef gene containing the oligopyrimidine⅐oligopurine PPT target sequence between the T 7 and SP 6 promoters in the pSP73 host vector (Promega), as previously described in detail (18). This system allowed us to run bidirectional in vitro transcription assays.
The pCMV(ϩ)PPT/luc plasmid derives from the bidirectional expression vectors (pBI Tet vectors; Clontech). These vectors contain the tet-responsive element between two identical minimal CMV promoters in opposite directions. This system was used to express two reporter genes, the firefly luciferase (Photinus pyralis) gene (luc) and the GFP gene. Expression of both genes is co-regulated by doxycycline. The activator protein (reverse Tet repressor protein, rTetR) binds the tetresponsive element in the presence of doxycycline and activates transcription. This protein was produced from the pTet-On expression plasmid (Clontech). Two 55-bp inserts containing either the wild-type human immunodeficiency virus type 1 polypurine tract target sequence (PPT, 5Ј-AAAAGAAAAGGGGGGA-3Ј) or a mutated sequence (PPTmut, 5Ј-AAAAGAAGGGGAGGAA-3Ј; the four mutations are shown in bold) were cloned in the 5Ј-transcribed but untranslated regions of the luciferase and GFP genes, downstream of the transcription start site.
The pRL-CMV vector (Promega) contains Renilla luciferase (from the marine organism Renilla reniformis) under the control of the CMV promoter. pRL-CMV was used to monitor transfection efficiency.
UV Absorption Melting Experiments-All thermal denaturation experiments were performed in a 10 mM sodium cacodylate buffer (at the indicated pH value) containing 10 mM MgCl 2 and 150 mM NaCl. The sample contained 1 M duplex (PPT or mutPPT duplex: 29-bp-long intramolecular hairpin duplexes) and 1.5 M TFO (see the sequences in Fig. 1). Melting profiles were recorded at 260 nm, using an UVIKON 940 spectrophometer, as previously described (19). Corrections for spectrophotometric instability were made by subtracting the absorbance at 540 nm from that at 260 nm. The temperature was decreased from 92°C to 0°C and increased again to 92°C at a rate of 0.2°C/min. All the cooling and heating profiles presented here were reversible. The melting temperature (T m in°C) was evaluated from melting curves, either directly or after subtraction of the duplex absorption from that of the triplex. The triplex T m was estimated within Ϯ0.5°C accuracy, except when the melting profiles of triplex transition presented a weak hypochromism (Ϯ1°C accuracy in T m ).
The standard molar enthalpy and entropy changes (⌬H a 0 and ⌬S a 0 ) associated with triplex formation were determined according to a twostate model and a van't Hoff analysis (20).
SPR Experiments-SPR measurements were performed on a BIAcore 2000 TM (BIAcore AB) using a carboxymethylated dextran-coated sensor chip (CM5), as previously described (21). Briefly, a controlled amount of streptavidine (ϳ5500 resonance units) was immobilized on the chip after activation by N-ethyl-NЈ-(dimethylaminopropyl)-carbodiimide/ N-hydroxysuccinimide. The unused activated sites were capped by injecting 100 l of 1 M ethanolamine. One micromolar solutions of 3Ј-biotinylated hairpin duplexes (PPT duplex and control duplex 5Ј-GCTAAAGAGAGAGAGAAATCGTTTTCGATTCTCTCTCTCTT-TAGCTTTTTTT-Biotin) were prepared in a buffer purchased from BIAcore (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, and 0.0005% surfactant P20), and duplex injection was controlled to obtain a 1500resonance unit increase. Serial dilutions of TFO were prepared in 10 mM sodium cacodylate, 10 mM MgCl 2 , 150 mM NaCl, and 0.0005% surfactant P20 at different pH values (pH 6, 6.5, or 7). These triplexforming buffer conditions are the same as the ones used in UV absorption melting experiments. G-containing TFOs, such as the 15TCG sequence, can self-associated to form G-quadruplex structures, thus decreasing the effective concentration of TFO. To prevent G-quadruplex formation, stock solutions of 15TCG/LNA and 15TCG/po were alkalitreated (with 50 mM NaOH for 15 min at 25°C and neutralized with 50 mM HCl) before SPR or spectroscopic experiments. Serial TFO injections were performed (100 l at a flow rate of 10 l/min) simultaneously on the hairpin duplex containing the target PPT sequence and on the control duplex lacking the PPT sequence. All the sensorgrams were corrected by subtraction of the control duplex signal, reflecting mainly bulk index changes. For thermodynamic analyses, SPR experiments were carried out at four fixed temperatures (20°C, 25°C, 30°C, and 37°C).
The sensorgrams were analyzed using BIAevaluation 3.0 software. The association rate constant k on was determined by the linear fitting of the apparent association rate constant (k app ϭ k on ϫ [TFO] ϩ k off ) at different TFO concentrations ([TFO]), in agreement with a two-state model. The dissociation profile could be fitted by a bi-exponential model. The fast component was neglected because it could arise from a fast relaxation of the sensor chip matrix, as previously discussed (21). The mass transport effects were negligible under our experimental conditions.
Electrophoresis Mobility Shift Assay-PPT or mutPPT hairpin intramolecular duplexes were 5Ј-end-labeled with [␥-32 P]ATP (3000 Ci/ mmol; Amersham Biosciences) by T4 polynucleotide kinase (Promega). The duplex (50 nM) was incubated with increasing concentrations of TFO in a buffer containing 50 mM HEPES, pH 7.2, 150 mM NaCl, 10 mM MgCl 2 , 0.5 mM spermine, and 10% sucrose (at room temperature overnight). The non-denaturing polyacrylamide gel (15%) was run at 37°C in a buffer containing 50 mM HEPES, pH 7.2, and 5 mM MgCl 2 . Gels were scanned with a PhosphorImager, and results were quantified using the ImageQuant software (Amersham Biosciences). The level of complex formation was estimated by the TFO concentration at which 50% of complex was formed (C 50 ).
Restriction Enzyme (DraI) Protection Assay-The pCMV(ϩ)PPT/luc plasmid contains the PPT sequence overlapping one of the seven DraI sites. It was used as a substrate for a restriction enzyme protection assay. For the cleavage assay, the pCMV(ϩ)PPT/luc plasmid was incubated at 37°C for 20 min with increasing amounts of oligonucleotides in 50 mM HEPES, pH 7.2, 50 mM NaCl, 10 mM MgCl 2 , and 0.5 mM spermine in the presence of DraI enzyme. The fragments generated by DraI cleavage in the absence of TFO are 3654-, 1936-, 1718-, 692-, 683-, 534-, and 19-bp long. A 3654-bp fragment corresponding to the addition of the 1936-and 1718-bp fragments was obtained when TFO-induced inhibition of DraI cleavage occurred. The extent of triplex-mediated inhibition of DraI cleavage was assessed by gel electrophoresis (0.8% agarose gel) and quantitated. The TFO concentration that gave 50% inhibition (IC 50 ) was then evaluated.
In Vitro Transcription Assay-Transcription assays were performed using the pSP-F47 plasmid that contains the PPT sequence between T 7 and SP 6 promoters. The plasmid was linearized by BspEI for T 7 transcription and synthesis of purine/PPT-containing RNA (660-nt long) or by Bsu36I for synthesis of pyrimidine/PPT-containing RNA (597-nt long). The linearized plasmids (0.  After 5 min at 37°C, transcription reactions were terminated by ethanol precipitation (10 volumes of ethanol was added to samples with 0.3 M sodium acetate and 15 g of glycogen). The transcription products were analyzed by electrophoresis on 6% polyacrylamide gels, and quantifications (Ϯ10%) were obtained by PhosphorImager analysis. The percentages were corrected for transcript length effects, taking into account that the transcripts were uniformly radiolabeled.
Cell Cultures and Transient Expression Assay-The P4-CCR5 cells were derived from HeLa cells (22) and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum.
P4-CCR5 cells were used for transfection of oligonucleotides and reporter vectors. Transfections were performed using the cationic activated dendrimer Superfect (Qiagen). Typically, 0.1 g of pCMV(ϩ/Ϫ)-PPT/luc plasmid, 0.1 g of pTet-On activator plasmid, 0.003125 g of pRL-CMV plasmid, and various amounts of oligonucleotides were mixed with Superfect (1.5 l) in a total volume of 12 l (serum-free medium). The mixture was prepared for triplicates and added to P4-CCR5 cells (13,750 cells/well in a 96-well plate in 88 l of serumcontaining medium in the presence of doxycycline, which was necessary for CMV promoter induction). After cell lysis (in 30 l of Passive Lysis Buffer; Promega), activities of both luciferases (firefly and Renilla) were measured in the same cell extract using the dual-luciferase assay kit (Promega), and GFP expression was measured as well (15 l of lysate in 80 l of phosphate-buffered saline). Luciferase and GFP expressions were measured with a luminometer/fluorometer (Victor TM -Wallac). The modulation of firefly luciferase expression by oligonucleotide treatment was quantitated by evaluation of the firefly/Renilla ratio; the specificity of triplex-induced inhibition was evaluated by the GFP/Renilla ratio. Each experiment was repeated at least three times, and values are presented as the mean of a triplicate (ϮS.D.) from a representative experiment.
RNA Analysis-Total cellular RNA was prepared (RNeasy Mini; Qiagen). Firefly and Renilla luciferase RNAs were subjected to competitive reverse transcription-PCR (for details, see Ref. 23). 2 Briefly two sets of two PCR primers were designed for each luciferase RNA; two competitor DNAs (firefly and Renilla) were used and co-amplified with the RNA sample. For each luciferase, the two amplified products (from sample and competitor, which are 153-and 192-bp long for firefly and 250-and 294-bp long for Renilla, respectively) were separated using an 8% PAGE in Tris borate-EDTA and quantitated by PhosphorImager analysis.

Tight and Specific Binding of LNA-modified TFOs in the (T,C,G)-Motif-
In the present work we have studied the binding and biological properties of a series of LNA-modified TFOs (TFO/LNAs) directed against a 16-bp-long oligopyrimidine⅐ oligopurine sequence (named PPT). All the sequences and their LNA content are shown in Fig. 1. The TFOs (15TCG and 16TC) have been described previously (25). Both bind parallel to the purine-containing strand of the duplex by Hoogsteen hydrogen bonding. Because fully modified TFO/LNA failed to hybridize to double-stranded DNA (7), only a few LNA modifications (five to nine) were introduced. For the (T,C,G)-containing sequence, three TFO/LNAs were designed: the 15TCG(1)/LNA is composed of alternating LNA and DNA nucleotides all along the sequence (8 LNA/15 nt); the 15TCG(2)/LNA contains LNA modifications only in the T-rich part of the sequence (5 LNA/15 nt); and the 16TCG/LNA is a modified version of the 15TCG(2)/ LNA with two additional LNA modifications at the 3Ј-end of the sequence to increase resistance to 3Ј-exonucleases (26), and it was used in cellular experiments. The 16TC/LNA sequence is the pyrimidine analogue of the 15TCG(1)/LNA and was designed as a reference to compare the (T,C)-and (T,C,G)-motifs.
The thermal stability of the different triplexes formed by these TFO/LNAs and the PPT target sequence was first determined by UV absorption melting experiments at neutral pH. Representative UV melting profiles of the triplexes formed by 15TCG/LNA and 16TC/LNA are shown in Fig. 2; the melting temperature values (T m ) are summarized in Table I. The presence of LNA modifications strongly increases the triplex-forming ability of the oligomers compared with the isosequential phophodiester in the (T,C)-motif (as described previously) (9, 27) and in the (T,C,G)motif. At neutral pH and physiological concentrations of monovalent cations (150 mM), the highest thermal stability was ob-  tained with the sequences 15TCG(1)/LNA and 15TCG(2)/LNA (T m ϭ 58°C and 55°C, respectively): a 21°C and 18°C increase (respectively) in the melting temperature was observed compared with the corresponding pyrimidine sequence 16TC/LNA (T m ϭ 37°C). The presence of three additional LNA modifications in the G-rich portion of the 15-TCG sequence slightly enhanced triplex stability (⌬T m (15TCG(1)/LNA Ϫ 15TCG(2)/ LNA) ϭ 3°C, i.e. 1°C/(LNA-G nucleotide)). Compared with the isosequential phospodiester 15TCG/po, the presence of LNA modifications resulted in a strong increase in T m (4.9°C/LNA nucleotide and 7.2°C/LNA nucleotide for 15TCG(1)/LNA and 15TCG(2)/LNA, respectively). Two additional LNA modifications at the 3Ј-end of the 15TCG(2)/LNA sequence did not significantly influence triplex stability as measured by the T m (⌬T m (16TCG/ LNA Ϫ 15TCG(2)/LNA) ϭ 1°C). This result is consistent with the hypothesis that the 3Ј-end nucleotide of the 16TCG/LNA sequence is not involved in base triplet formation as already observed for the 16TCG/po (28).
In order to assess the specificity of duplex recognition by the 15TCG/LNA sequences, we performed three different assays (Table II). In two of them, we compared the TFO/LNA binding to the PPT duplex target and to the mutPPT duplex containing two mutations (5Ј-AAAAGAAAAAGGAGGA-3Ј; mutations are shown in bold). First, using UV absorption melting experiments, we showed that 15TCG(1)/LNA and 15TCG(2)/LNA could discriminate between the wild-type and the mutated target with a large decrease in T m (⌬T m ϭ 27°C and 19°C, respectively). It should be noted that nine contiguous triplets could theoretically be formed with the A 4 GA 4 sequence on both of the targets (PPT and mutPPT). Second, the specificity of triplex formation was confirmed by gel shift experiments. Increasing amounts of the different TFO/LNAs were incubated in the presence of the PPT or mutPPT duplexes at 37°C and neutral pH (see "Materials and Methods"). TFO concentrations required for 50% of complex formation (C 50 ) were estimated. Under these conditions, the two 15TCG/LNA sequences present affinities for the PPT target in the same range (C 50 (15TCG(1)/LNA) ϭ 0.4 M; C 50 (15TCG(2)/LNA) ϭ 0.15 M). Under our experimental conditions (150 mM NaCl), the 15TCG sequences might self-associate; a different degree of self-association for 15TCG(1)/LNA and 15TCG(2)/LNA could explain the slight difference in apparent binding affinities. No binding was detected on the mutPPT duplex up to 2.5 M TFO/LNAs (at this concentration, the triplex was completely formed on PPT duplex), ensuring specificity of DNA target recognition. Third, we determined the capacity of TFO/LNAs to interfere with restriction enzyme cleavage specifically at the triplex site. Indeed, the sequence at the 5Ј-side of the PPT oligopurine strand provides a recognition site for the restriction enzyme DraI, which overlaps the triplex site on 3 bp (Fig. 1). This enzyme cleaves at the junction of the triple helix site (TTT2AAA), and triplex formation inhibits DNA cleavage (29). The pCMV(ϩ)-PPT/luc plasmid contains the PPT site and seven DraI sites. In the presence of TFO, DraI cleavage was impaired only on the site overlapping the PPT sequence: inhibition of cleavage at this site led to the appearance of a unique longer 3654-bp fragment and the disappearance of the 1936-and 1718-bp fragments. TFO concentrations required for 50% inhibition (IC 50 ) of DraI cleavage were determined from the concentration dependence of cleavage inhibition for the two 15TCG/LNA TFOs (IC 50 Under the same conditions, the isosequential phosphodiester TFO did not inhibit cleavage at 37°C up to a concentration of 10 M. The level of DraI cleavage inhibition specifically at the triplex site induced by TFOs reflects both the specificity and the stability of triplex formation.
These results obtained with G-containing TFO/LNAs in different experimental settings support a strong increase in stability induced by LNA modification and a specific binding of TFO/LNAs to the PPT duplex target.
Kinetics and Energetics of Triplexes Formed with LNA-modified TFOs in the (T,C,G)-Motif-To investigate the origin of the high stability of triplexes formed with G-containing TFO/ LNAs, we performed kinetic and thermodynamic analyses.
The kinetics of triplex formation was studied by SPR experiments for the 15TCG/LNAs and compared to both the 16TC/ LNA and 15TCG/po. The PPT duplex was captured on the surface of a sensor chip as well as a control duplex, and the different TFOs were injected at increasing concentrations (see "Materials and Methods"). Fig. 3A shows the corrected sensorgrams obtained with the 15TCG(1)/LNA at 37°C: 10 concentrations of TFO (0.5, 0.75, 1, 1.5, 2, 3, 4, 6, 8, and 12 M) were used in injection experiments to determine the association (k on ) and dissociation (k off ) rate constants. The association rate constant was deduced from the apparent association rate constant k app (k app ϭ k on [TFO] ϩ k off ). For all the studied TFOs, we obtained a linear dependence of the k app with respect to the concentration of TFO (data not shown), consistent with a simple two-state model. The calculated values of the kinetic constants are reported in Table III.
The association and dissociation rate constants of the triplexes formed with the G-containing 15TCG(1)/LNA (and 15TCG(2)/LNA) were 0.53 ϫ 10 3 (and 1.1 ϫ 10 3 ) M Ϫ1 s Ϫ1 and 4.5 ϫ 10 Ϫ5 (and 1.3 ϫ 10 Ϫ4 ) s Ϫ1 , with these latter k off values corresponding to a lifetime ( ϭ ln2/k off ) as long as 9.5 h (and 1.5 h). Compared with the pyrimidine 16TC/LNA, triplex formation with the 15TCG/LNAs exhibited an equivalent association rate (k on (16TC/LNA) ϭ 0.9 ϫ 10 3 M Ϫ1 s Ϫ1 ), but a much  Fig. 1). Data were obtained in 10 mM cacodylate, 150 mM sodium chloride, and 10 mM magnesium chloride, pH 7. Strand concentrations: 1 M PPT duplex and 1.5 M TFO. Estimated error in T m is Ϯ0.5°C (or Ϯ1°C for cases indicated by an asterisk, which correspond to a triplex transition presenting a weak hyperchromism). b TFO concentrations for 50% triplex formation estimated from gel shift assays (C 50 values). Increasing concentrations of TFO/LNAs were incubated in 150 mM sodium chloride, 50 mM HEPES, pH 7.2, 10 mM magnesium chloride, and 0.5 mM spermine, with either PPT or mutPPT duplex. The samples were analyzed by electrophoresis in 15% non-denaturing polyacrylamide gel (50 mM HEPES, pH 7. slower dissociation (ϳ10 -35-fold decrease in k off (k off (16TC/ LNA) ϭ 1.5 ϫ 10 Ϫ3 s Ϫ1 ). Thus, this decrease in the dissociation rate constants is the major reason for the enhanced stability of triplexes formed with the two G-containing TFO/LNAs compared with the one formed with the pyrimidine TFO/LNA The comparison of the two 15TCG/LNAs (15TCG(1)/LNA and 15TCG(2)/LNA), which contain the same 5Ј-sequence (5Ј-tTtTcTtTt; LNA is shown in lowercase letters) and differ only by the presence of three LNA modifications in the G-rich 3Јportion (GgGgGg-3Ј), gave information concerning the specific effect of these modifications. The presence of LNA-G nucleotides in the 3Ј-end region resulted in a slightly slower association rate (2-fold decrease in k on ) and in a longer lifetime of the triplex (3.5-fold decrease in k off ). As a result, the two 15TCG/ LNAs exhibited very similar equilibrium constants (K d (37°C) Ϸ 0.1 M). The observed 2-fold decrease in k on could be explained by unfavorable conformations of the single-strand 15TCG(1)/LNA, possibly including structures generated by self-association.
The pH-induced effects on kinetic and equilibrium constants of triplexes formed with TFO/LNAs (15TCG(2)/LNA and 16TC/ LNA) were studied (Fig. 3B). Changes in pH values between pH 6.5 and pH 7 affected the stability of triplex formed with 15TCG(2)/LNA (6-fold increase in K d /ϩ0.5 unit of pH) less than that of the one formed with 16TC/LNA (14-fold increase in K d /ϩ0.5 unit of pH), as expected considering the different C content of the TFOs (one in 15TCG and seven in 16TC). In this pH range (pH 6.5-7), K d appeared to be mainly driven by k off for both types of triplexes (formed with 15TCG(2)/LNA or 16TC/LNA). Such a trend has already been reported for phosphodiester pyrimidine TFOs (21).
To investigate the kinetic mechanisms of LNA-induced triplex stabilization, we compared the 15TCG/LNA TFOs to the isosequential phophodiester 15TCG/po. Due to weak triplex stability with the unmodified TFO (T m ϭ 19°C), SPR measurements were performed at 10°C, as compared with 37°C with TFO/LNAs. The decrease in K d for the 15TCG/LNAs was mainly associated with a decrease in the dissociation rate constant (Table III). For 15TCG (2)/LNA, we interpolated the Arrhenius plots of k off and K d (Fig. 4) to estimate their values at 10°C (k off (10°C) ϭ 1.9 ϫ 10 Ϫ5 s Ϫ1 , K d (10°C) ϭ 2.4 ϫ 10 Ϫ8 M). We could thus evaluate the LNA-induced effect compared with the isosequential phosphodiester TFO, namely, a 68-fold increase in affinity, associated with a 162-fold decrease in dissociation rate constant.

Kinetics of triplexes formed with different TFO/LNAs: kinetic rate constants k off and k on and the corresponding calculated equilibrium constant K d ϭ (k off /k on )
The data were obtained in a buffer containing 10 mM cacodylate, 150 mM sodium chloride, 10 mM magnesium chloride, and 0.005% P20. Specific conditions are indicated. These values were obtained at 37°C for the TFO/LNAs and at 10°C for the 15TCG/po ($); in the latter case the triplex is unstable at 37°C under the conditions applied. All data were obtained in a buffer containing 10 mM cacodylate, 150 mM sodium chloride, 10 mM magnesium chloride, and 0.0005% P20.
Our kinetic data demonstrate that the tight stabilization observed for triplexes formed with G-containing TFO/LNAs was mainly due to a longer lifetime of the triplexes (in the range of 1 h, at 37°C) compared with phosphodiester TFO (in the range of 100 s). In contrast, the association rate was marginally affected. These results support the data obtained with another pyrimidine TFO/LNA sequence (9) and extend them to G-containing TFO/LNAs.
To further understand the origin of the triplex stabilization observed with 15TCG(2)/LNA, the entropy and enthalpy changes were evaluated and compared with those for both the isosequential phosphodiester (15TCG/po) and 16TC/LNA. For the TFO/LNAs, UV melting profiles did not allow an accurate determination of thermodynamic parameters: indeed, either the melting of the triplex was too close to that of the duplex (15TCG/LNA) or the transition was poorly cooperative (16TC/ LNA) (see Fig. 2). Instead, SPR experiments were carried out at different temperatures (20°C, 25°C, 30°C, and 37°C) under conditions allowing triplex formation with the 15TCG/LNA and 16TC/LNA (see Fig. 2). The values of the equilibrium dissociation constant K d were calculated as the ratio of the kinetic rate constants (K d ϭ k off /k on ); the enthalpy and entropy changes (⌬H a 0 and ⌬S a 0 ) associated with triplex formation were calculated from a linear fitting of lnK d versus T Ϫ1 (according to the equation lnK d ϭ (1/RT)⌬H a 0 Ϫ (1/R)⌬S a 0 , where R ϭ gas constant and T ϭ temperature in Kelvin) (Fig. 4B). The estimated enthalpy and entropy changes of the association of TFO/ LNAs to the PPT target are reported in Table IV. The binding of the studied TFOs was enthalpically driven (⌬H s 0 Ͻ 0 and ⌬S a 0 Ͻ 0). Nevertheless, we could observe significant differences in the enthalpic and entropic contributions depending on LNA-modifications and G-content in the TFO (Table IV).
Compared with the isosequential phophodiester 15TCG/po, the binding of 15TCG(2)/LNA was associated with an 8-fold lower entropy change (less negative ϭ more favorable) and a 3-fold lower enthalpy change (less negative ϭ less favorable). The pyrimidine 16TC/LNA was also compared with the isosequential phosphodiester 16TC/po. Due to the lack of stability of the pyrimidine phosphodiester triplex at pH 6.5, thermodynamic data were obtained from experiments at pH 6.0. To compare the data obtained at different pH values, we take into account the fact that, in this pH range, increase in pH was mainly associated with an entropic penalty (more negative ⌬S a 0 and equal ⌬H a 0 , as suggested previously (30)). For 16TC/LNA, the same energetic effects associated with the presence of LNA modifications were observed (as the ones described previously for the 15TCG(2)/LNA): the LNA modifications confer an entropic gain but an enthalpic loss, compared with the nonmodified TFOs. These thermodynamic properties seem to be intrinsic to pyrimidine or G-containing TFO/LNAs, thus extending a previous study on another pyrimidine TFO/LNA (9). They are consistent with favorable preorganization of TFO/ LNAs (entropic contribution) and unfavorable base stacking in the triplex containing TFO/LNA (enthalpic contribution), as suggested by a recent NMR study (31).
Finally, we compared the thermodynamic parameters of the triplexes formed with TFO/LNAs in the (T,C)-and (T,C,G)motifs (i.e. 16TC/LNA and 15TCG(2)/LNA), under conditions in which the corresponding triplexes presented nearly the same stability at 37°C (namely, pH 7.0 for 15TCG(2)/LNA and pH 6.5 for 16TC/LNA; see K d in Fig. 3B). The enthalpy of triplex formation was less favorable (less negative) for the G-containing TFO/LNA than for the pyrimidine TFO/LNA (Ϫ15 and Ϫ24 kcal mol Ϫ1 , respectively): such an enthalpic effect has already been described for phosphodiester (T,C)-and (T,G)-containing TFOs (32). This less favorable enthalpic contribution for triplexes containing (C⅐GXG) base triplets was compensated by a less unfavorable (less negative) entropic change (Ϫ17 and Ϫ45 cal K Ϫ1 mol Ϫ1 , respectively). The small enthalpic gains (low ͉⌬H a 0 ͉) associated with triplex formation with LNA-modified

TABLE IV Thermodynamic constant parameters of triplexes formed with TFO/LNAs and TFO/pos in the (T, C, G)-and (T, C)-motifs
For TFO/LNAs, data were deduced from SPR experiments (Arrhenius plots; Fig. 4B). For TFO/pos, the thermodynamic parameters were deduced from UV melting experiments (van't Hoff plots: r 15TCG/po 2 ϭ 0.996; r 16TC/po 2 ϭ 0.997). Enthalpy (⌬H a 0 ) and entropy (⌬S a 0 ) values of TFO association are reported. T m SPR values were estimated from Arrhenius plots as the temperatures at which K d ϭ 1 M (Fig. 4B). The measured T m UV values from Table I  and G-containing TFO, as obtained by SPR experiments, are consistent with the corresponding UV melting profiles, featuring weakly cooperative transitions (Fig. 2). It can be noted that SPR experiments did allow determination of ͉⌬H a 0 ͉ (and ͉⌬S a 0 ͉) values as low as 10 -20 kcal mol Ϫ1 (and cal K Ϫ1 mol Ϫ1 , respectively); it could have been hard to obtain these values from UV spectroscopic experiments or calorimetry measurements.
Kinetic SPR measurements were validated by comparison with data from UV melting experiments. UV melting experiments were carried out at duplex and TFO concentrations so that the equilibrium dissociation constant at the melting temperature was 1 M (K d UV (T m UV ) ϭ 1 M). By interpolating the Arrhenius plots obtained from SPR measurements (lnK d SPR versus T Ϫ1 , where K d SPR ϭ k off /k on ), we extrapolated the temperature T m SPR at which K d SPR ϭ 1 M. The temperatures deduced from SPR (T m SPR ) agreed with those obtained by UV melting experiments (T m UV ) ( Table IV). The consistency of the data obtained separately by the two techniques provides a validation of parameters measured by SPR.
TFO/LNAs for Arrest of Elongating RNA Polymerases in Vitro-We evaluated the capacity of TFO/LNAs to interfere with DNA-associated proteins, such as elongating RNA polymerases.
A plasmid with the T 7 polymerase promoter upstream of the PPT site was used as a template for transcription assays (Fig.  5). After appropriate plasmid linearization, full-length transcripts (660-nt long) were produced. In the presence of TFO/ LNAs, truncated products did appear. To establish whether the truncated fragments corresponded to the arrest of transcription at the PPT site, the site of transcription arrest was compared with that obtained when the DNA template was cleaved with DraI on the 5Ј-side of the PPT target sequence (DraI RNA marker: 332-nt long, lane M in Fig. 5B). The truncated transcripts migrated at the same position as the DraI RNA marker, consistent with a physical blockage of RNA synthesis at the PPT site, i.e. the TFO/LNA binding sequence. In the presence of TFO/LNAs, the percentage of truncated fragments increased with TFO concentration. For T 7 transcription at 37°C, the plateau (reached at 10 M TFO/LNA) corresponded to a percentage of truncated fragments of about 50% with 15TCG(1)/ LNA, 70% with 15TCG(2)/LNA, and 80% with the 16TCG/LNA. Under these transcription conditions, the phosphodiester TFO (15TCG/po) was unable to inhibit transcription.
To evaluate the influence of the nature of the coding strand (purine-or pyrimidine-containing strand), SP 6 RNA polymerase was used to initiate transcription of the pSP-F47 plasmid in the opposite direction as compared with the T 7 RNA polymerase (Fig. 5A): T 7 transcription generated the purine-containing RNA (copy of the pyrimidine-containing strand), and SP 6 transcription generated the pyrimidine-containing RNA (copy of the purine-containing strand). SP 6 transcription in the presence of TFO/LNAs also produced a truncated product that corresponded to an arrest of RNA synthesis at the PPT site, localized by the DraI marker as described above (data not shown). All these results were consistent with a physical blockage of the RNA polymerase by the TFO/LNAs, independent of the nature of the coding strand, and support a triplex-based mechanism for inhibition of transcription.
These data are consistent with data from UV melting experiments and SPR analyses described above (strong triplex stabilization induced by LNA modifications, specificity of triplex formation, and equivalent binding of the different G-containing TFO/LNAs to the double-stranded PPT target).
Triplex-based Activity of G-containing TFO/LNAs in Cells-The physico-chemical properties and in vitro biological activity of TFO/LNA described above allowed us to envision the use of these molecules in a cellular context, which has never been described previously. We evaluated the cellular activity of TFO/ LNA using a transient expression assay designed to quantitatively and rapidly test drugs directed against the oligopyrimi-dine⅐oligopurine PPT target sequence and to demonstrate triplex involvement in the TFO-induced effect. The assay used two different DNA templates (pCMV(ϩ)PPT/luc and pCMV-(Ϫ)PPT/luc) coding for the firefly luciferase; they either contained or lacked (respectively) the PPT sequence in the transcribed and untranslated region of the luc gene (Fig. 6). More precisely, these constructs differed by the presence or absence of a 55-bp insert containing the wild-type PPT target sequence (see "Materials and Methods"). P4-CCR5 cells were transiently transfected with the pCMV(ϩ/Ϫ)PPT/luc plasmid together with pTet-On (expressing the activator protein necessary for transcription from the CMV-inducible promoter) and pRL-CMV plasmid (expressing Renilla luciferase and used for correction of transfection efficiency) and various TFO/LNAs. Both luciferase activities were measured on the same cell extracts. The ratio of luminescence (firefly/Renilla) permits quantification of the modulation of the firefly luciferase expression induced by TFO treatment. In addition, the pCMV(ϩ)PPT/luc plasmid did contain a second CMV promoter in the opposite direction controlling the expression of a GFP gene and a mutated PPT sequence (PPTmut, 5Ј-AAAAGAAGGGGAGGAA; four mutations are shown in bold) inserted upstream of the GFP coding sequence in the transcribed region (see Fig. 6): evaluation of GFP expression (corrected by Renilla expression; GFP/Renilla ratio) as a function of TFO treatment will be used to estimate the specificity of triplex-induced inhibition.
The activities of the three different G-containing TFO/LNAs were evaluated in dose-response experiments (see the sequences in Fig. 1). Typical histograms are reported in Fig. 6. In the presence of 16TCG/LNA, an inhibition of firefly luciferase expression, dependent on TFO concentration, was observed: 50% inhibition was obtained at 1.5 M TFO. To assess triplex involvement in the inhibitory activity, three types of experiments were conducted. First, we used a control LNA oligonucleotide (16TCG-cont/LNA) presenting the same base composition and LNA content as 16TCG/LNA and preserving the G 3 tract that could be implicated in the formation of multistranded structures: no decrease of firefly luciferase expression was de-tected in the presence of the 16TCG-cont/LNA. Second, we used the pCMV(Ϫ)PPT/luc plasmid: deletion of the PPT target sequence abolished the inhibitory effect of 16TCG/LNA. Third, we measured GFP expression: TFO treatment did not change the level of GFP expression, consistent with no or low TFO binding to the mutated PPT oligopyrimidine⅐oligopurine sequence upstream of the GFP gene. These results demonstrated that the inhibition was due to an interaction between 16TCG/ LNA and the PPT target sequence. Moreover, quantification of the mRNA level showed a decrease of firefly luciferase mRNA (data not shown), excluding a direct interaction between 16TCG/LNA and the PPT sequence present on the mRNA and supporting transcriptional inhibition at the DNA level. The same type of results was obtained with the three studied TFO/ LNAs (15TCG(1)/LNA, 15TCG(2)/LNA, and 16TCG/LNA) with an IC 50 around 1.5 or 2 M.
Finally, the efficacy of triplex-induced inhibition was studied with 16TCG/LNA linked to a psoralen derivative (Pso-16TCG/ LNA). The psoralen moiety should intercalate at the duplextriplex junction, as previously described for other TFO modifications (25), and thus increase triplex stability, in the absence of photoactivation. The psoralen derivative enhanced the inhibitory effect compared with unconjugated TFO, with a 4-fold decrease in the IC 50 (IC 50 (Pso-16TCG/LNA) ϭ 0.5 M).
In contrast, the phosphodiester TFO (15TCG*/po, 3Ј-modified to resist nuclease-mediated degradation) was inactive up to 2 M, even when attached to a psoralen molecule (Pso-15TCG*/po) (Fig. 6). DISCUSSION In this work, we have characterized the binding properties and cellular activity of triplexes formed with G-containing FIG. 6. Cellular activity of G-containing TFO/LNAs: inhibition of transcription elongation. A, experimental design. Schematic representation of the pCMV(ϩ)PPT/luc plasmid. This plasmid contains two identical modified CMV promoters in opposite directions (see "Materials and Methods" for details) and allows expression of two reporter genes (firefly luciferase (luc) and GFP). The PPT sequence was inserted just upstream of the luc gene and a mutated version (PPTmut, 5Ј-AAAAGAAGGGGAGGAA-3Ј, the four mutations are shown in bold) was inserted upstream of the GFP gene. Alternatively, there was no insertion of the PPT-containing fragment in the pCMV(Ϫ)PPT/luc plasmid. B, cell extracts were analyzed for firefly and Renilla luciferases activities 24 h after transfection by a cationic dendrimer (Superfect). The expression vector (pCMV(ϩ/Ϫ)PPT/luc) was introduced together with the pRL-CMV plasmid, in the presence or absence of TFO/LNAs. Normalized firefly/Renilla luminescence intensity ratios are reported in histogram with different TFO/LNA concentration treatments. Data are derived from 3 (for 16TCG-cont/LNA and 15TCG*/po) to 10 experiments (for other TFO/LNAs). GFP expression was also evaluated on the same cell lysates, and the GFP/Renilla ratio was unaffected by TFO treatments (data not shown). Equivalent results were obtained with the two 15TCG/LNAs (15TCG/ LNA(1) and 15TCG/LNA(2)). 15TCG*/po, 15TCG/po 3Ј-modified by incorporation of a propylamine group. TFO/LNAs. These triplexes were compared with the ones formed with isosequential phosphodiester TFO (15TCG/po) and also with pyrimidine TFO/LNA (16TC/LNA) directed against the same oligopyrimidine⅐oligopurine PPT target sequence, with the aim of determining the specific influence of LNA modifications in the context of G-containing TFO/LNAs. Actually, thus far, TFO/LNAs have only been used to form triplexes in the (T,C)-motif. Our data, obtained in conditions simulating physiological ones (37°C, neutral pH), showed a strongly enhanced DNA binding of G-containing TFO/LNAs compared with 15TCG/po and 16TC/LNA. For the G-containing TFO/ LNA, 15TCG(2)/LNA, an increase in T m as high as 7.2°C per LNA modification was observed, the highest ever reported for modified triplexes. It is noteworthy that, in the present study, all the data were obtained at physiological monocation concentrations (150 mM NaCl). These conditions are known to decrease the stability of triplexes formed with G-containing TFOs: indeed, TFO self-association can compete with triplex formation. For some G-rich LNA-modified oligonucleotides (33), the formation of G-quadruplexes has been observed, which could act as competing structures.
To understand the origins of triplex stabilizations observed with 15TCG/LNAs compared with 15TCG/po and 16TC/LNA, we performed kinetic and thermodynamic analyses.
From a kinetic point of view, triplex stabilization induced by G-containing TFO/LNA was mainly due to a decrease in the dissociation rate constants, with residence time in the range of 1-10 h; the association rates were marginally affected. The decrease in the dissociation rates appeared to be correlated with the number of LNA modifications in the TFO, as shown by the comparison of 15TCG/po (no LNA, as a reference) with 15TCG(2)/LNA (five LNAs) and 15TCG(1)/LNA (eight LNAs). The origin of the decreased dissociation rate might be at the triplet level, as suggested previously (31): a lower propensity for bp opening for single triplets in triplexes formed with TFO/ LNAs would translate into a lower dissociation rate at the strand level. Finally, it has already been shown by UV melting experiments (8,27) that a pH increase destabilizes triplexes formed with cytosine-containing TFO/LNAs. Here we reported pH-induced effects (with pH value close to neutral) on kinetics for triplexes formed with TFO/LNAs in the (T,C,G)-and (T,C)motifs. We showed that the major effect did concern the dissociation process, as reported previously for non-modified triplexes in the (T,C)-motif (21): a pH increase was associated with a faster dissociation.
Energetically, triplex formation with TFO/LNA in the (T,C,G)-motif (15TCG/LNA) was accompanied by a smaller (less negative ϭ unfavorable) enthalpy change that is compensated for by a smaller (less negative ϭ favorable) entropy change, compared with the phopshodiester (T,C,G)-sequence (15TCG/po). Thus, the enhancement in triplex stability was entropic in origin. This feature has already been reported for triplexes formed with TFO/LNA in the (T,C)-motif (9). This could be explained by a conformational restraint of the unbound TFO/LNA (both G-containing and pyrimidine sequences) because the conformational state of free TFO plays an important role in the thermodynamics of triplex formation (34). The increased rigidity of TFO/LNA in the free state, compared with non-modified TFO, may entropically favor its binding on the target duplex. The enthalpic loss induced by TFO/LNA compared with non-modified TFO may be explained in part by a recent NMR structural study (31): the duplex structure may change upon TFO/LNA binding to accommodate nucleotides with different sugar puckerings. These structural changes alter the base stacking interactions and the hydrogen bonding pattern, compared with non-modified TFO. The differences in thermodynamic parameters observed with the LNA (T,C,G)sequence (15TCG(2)/LNA) compared with the LNA (T,C)-sequence (16TC/LNA) could be explained in part by the following: (i) differences in the conformational micro-states of the two types of TFO/LNA when unbound (34); (ii) for entropic contribution, differences in the pattern of solvent water molecules, as already suggested for phosphodiester triplexes in the reverse Hoogsteen (T,G)-and (A,G)-motifs compared with (T,C)-motifs (it is possible that there are more water molecules removed in G-containing triplex motifs than in pyrimidine motifs) (32); and (iii) for enthalpic contribution, a different extent of cytosine protonation.
Finally, LNA modifications facilitated Hoogsteen hydrogen bonding to the duplex when the third strand is in parallel orientation with respect to the the oligopurine target sequence because it was observed with the pyrimidine and G-containing oligomers studied in this report. In contrast, reverse Hoogsteen hydrogen bonding with (G,A)-and (G,T)-containing TFO/LNAs seemed to be disfavored (data not shown). The same type of results (triplex stabilization for parallel Hoogsteen bonding and not for antiparallel reverse Hoogsteen bonding) has already been observed with other TFO modifications associated (as is the case with LNA oligonucleotides) with an RNA-like N-type (C3Ј-endo type) conformation, namely, phosphoramidates and RNA (25,35).
The biological activity of these G-containing TFO/LNAs reported here was consistent with the high triplex stability observed under conditions simulating physiological conditions. We evaluated the in vitro and cellular activities of TFO/LNAs using experimental systems designed to demonstrate quantitatively and rapidly a triplex-based mechanism. Their capacity to interfere with transcription elongation was estimated: all the G-containing TFO/LNAs studied here were active under conditions in which the parent phosphodiester oligonucleotide did not exhibit any inhibitory effect. In different experiments, using control oligonucleotides and targets, we could demonstrate the triplex involvement in transcription arrest, both in vitro and in cells. The three different G-containing LNAs tested showed almost the same activity in our assays. It is noticeable that the 15TCG(2)/LNA, which contains only five LNA modifications, was active in cells. A 3Ј-modification is generally required to protect an oligonucleotide from nuclease degradation. The fact that the 15TCG(2)/LNA, which lacks a 3Ј-modification, was active intracellularly could be explained by a partial selfassociation, which protected it from nuclease degradation but still allowed triplex formation. The efficiency of TFO/LNAs can still be improved by functionalization strategies, as illustrated here with psoralen conjugation: a psoralen conjugate of a TFO/ LNA was active within cells at submicromolar concentrations. Inhibitory concentration of TFO/LNAs would likely be lowered in another context, more favorable than our experimental system: indeed, the target sequence was quite short (16 bp), and the cellular activity to inhibit was transcription elongation, which is known to be difficult to arrest. It is likely that TFO/ LNAs might efficiently interfere with other biological processes at concentrations lower than those obtained in this study for transcription elongation.
A series of oligonucleotide-based techniques do exist and are efficient for gene knock-down, including antisense and small interfering RNA technologies. Nevertheless, TFOs are unique in their nature, as sequence-specific DNA ligands, for modulation of DNA-mediated biological functions. They have already been used to interfere intracellularly with processes occurring on DNA, such as transcription, repair, or recombination. Manipulation of chromatin structure (36, 37) and positioning of DNA-damaging agents using these agents conjugated with TFO (24,38) are other promising applications of DNA codereading molecules; they have been successfully explored with TFOs, but mainly in vitro. However, today, the major limitation of triplex-based applications is the limited intracellular stability of the complex. This is the reason why it is necessary to identify and characterize novel chemically modified TFOs with increased triplex-based cellular activity that are easy to synthesize and compatible with conjugation chemistry, as is the case for the LNA-modified oligonucleotides described here. Our study is the first step for further developments and original use of TFO/LNAs in the anti-gene strategy.