A macrocyclic bis-acridine shifts the equilibrium from duplexes towards DNA hairpins
A macrocyclic bis-acridine shifts the equilibrium from duplexes towards DNA hairpinsA. Slama-Schwok*, F. Peronnet1, E. Hantz-Brachet, E. Taillandier, M.-P. Teulade-Fichou2, J.-P. Vigneron2, M. Best-Belpomme1 and J.-M. Lehn2
Laboratoire de Spectroscopie Biomoléculaire, URA CNRS 1430, UFR de Santé, Médecine et Biologie Humaine, Université Paris XIII, Bobigny, France, 1Groupe de Génétique Moléculaire et Cellulaire, URA CNRS 1135, Université Pierre et Marie Curie, Paris, France and 2Chimie des Interactions Moléculaires, UPR CNRS 285, Collège de France, Paris, France
Received April 7, 1997;Revised and Accepted May 16, 1997
ABSTRACT
Nucleic acids can undergo dynamic conformational changes associated with the regulation of biological processes. A molecule presenting larger affinities for alternative structures relative to a duplex is expected to modify such conformational equilibria. We have previously reported that macrocyclic bis-acridine binds preferentially to single-stranded regions, especially DNA hairpins, due to steric effects. Here, we show, using gel electrophoresis, fluorescence and melting temperature experiments, that the macrocycle bis-acridine shifts an equilibrium from a duplex towards the corresponding hairpins. Competition experiments enlighten the higher affinity of the macrocycle for hairpins compared with double-stranded DNA. The macrocycle bis-acridine destabilizes a synthetic polynucleotide, by the formation of premelted areas. By extrapolation, the macrocycle bis-acridine should be able to disrupt, at least locally, genomic DNA duplexes and to stabilize unpaired areas, especially palindromic ones forming hairpins. Such macrocyclic compounds may have potential applications in the therapy of diseases involving hairpins.
INTRODUCTION
Nucleic acids can adopt alternative structures to the double helix (1 ). These structures are sometimes conserved through evolution. On the one hand, these structures are often specifically recognized by proteic factors. For example, the nuclear high mobility group (HMG) 1 protein recognizes cruciform DNA (2 ,3 ) and other HMG box transcription factors, like SRY, can bind to four-way junctions (4). Single-stranded structures are also involved in protein binding (5). On the other hand, the binding of proteic factors (6 -8 ) sometimes promotes conformational changes (9 ).
Alternative structures can be dynamic motifs involved in the regulation of biological processes, such as transcription (10), translation (11) and replication (12 ,13 ). Cruciform extrusion or hairpin formation can regulate transcription (14 ) of several human genes (15-17). Hairpins may also be associated with the fragile X chromosome syndrome and human diseases (18 -20 ). These structures are involved in the biology of retroviruses at different levels. The initiation of HIV dimerization is promoted by a `kissing complex' between hairpin loops (21-22 ).
Antiviral compounds, as do antisense oligonucleotides (23-24), often bind poorly to these structures unless they do so by strand invasion (25-27), which disrupts the target structure. Attempts to exploit the structure instead of avoiding it led to the construction of tethered antisense oligonucleotides (28) or structure-specific peptides or aptamers (29 -32 ). The involvement of single-stranded motifs in most of these alternative structures has motivated the design of molecules that specifically recognize single-stranded versus double-stranded nucleic acids. At the present time only a limited number of organic compounds which present specific binding to single-stranded DNA are known. A Ni2+ macrocyclic complex cleaves unpaired guanines located in mismatches, bulges and loops (33). Intercalators linked to an appropriate base bind to hairpin bulges (34). Neocarzinostatin, a compound of the enedyine family, selectively binds and cleaves DNA bulges by a thiol-independent mechanism (35 ,36 ).
The macrocycle bis-acridine 1 (Scheme 1 ) was designed with the expectation that compounds of the bis-cyclointercaland type would bind preferentially to single-stranded rather than to double-helical nucleic acid domains, based on steric effects (37). The macrocycle 1 specifically stabilizes a model hairpin. This contrasts with the lack of stabilization of this structure obtained with a reference compound that possesses an acridine subunit and similar amine arms but is not included in a bulky macrocyclic structure. The macrocycle 1 presents a high affinity for model hairpins compared with other DNA structures, double helical or `unstructured' single-stranded oligomers (38 ). These results suggest that the macrocycle may disrupt a double helical fragment.
MATERIALS AND METHODS
Materials
The oligonucleotides GCG(AAA)CGC, GCG(AAAAA)CGC, GCG(TTTTT)CGC and GCG(CTGGGA)CGC are referred to as sA3, sA5, sT5 and sTAR (38 ) respectively, where s represents the stem (GCG/CGC). GCTGGCGTAG(GGC)CTGCGTCAGC (16 ,17 ) is referred to as B23. The oligomers d(A)4, d(A)9 and d(A)30 are referred to as (A)n, where n = 4, 9 or 30. CGCGAATTCGCG (39) is referred to as D12. These oligonucleotides were purchased from Eurogentec or Genosys, (A)4, poly- (dA[middot]dT) and poly(dA)[middot]poly(dT) from Pharmacia and poly(A) from Sigma. The chemicals used were of the highest commercial purity. All aqueous solutions utilized purified water (MilliQ, Millipore). The solutions contained 1 mM cacodylate buffer, pH 6.0, 4 mM NaCl (buffer A) or 10 mM PIPES [piperazine N,N'-bis(2-ethanesulfonic acid)] buffer, pH 6.0, (buffer B) or 30 mM cacodylate buffer, pH 6.0, 100 mM KCl (buffer C). The synthesis of the macrocyclic acridine 1 has been previously reported (37 ).
DNA-1 binding assays
The oligonucleotides were labeled by T4 polynucleotide kinase by phosphorylation with [[gamma]-32P]ATP and purified on sequencing gels. The appropriate bands were cut and the DNA eluted in 100 mM Tris-HCl, pH 8.0, containing 0.5 mM EDTA and 0.5 M NaCl. The DNA was desalted on Sephadex G-10 columns and lyophilized. Probes were resuspended in water and prepared for DNA-1 binding assays. Aliquots of 1 [mu]l oligonucleotide probe (105 c.p.m.), 1 [mu]l of a 10* oligonucleotide solution (10 times more concentrated relative to the required * concentration), 1 [mu]l 10 mM cacodylate buffer, pH 6.0, and 6 [mu]l water were mixed and boiled for 2 min (except in experiments including D12). The samples were then kept for 30 min at room temperature or 4oC as indicated; 1 [mu]l 10* macrocycle solution was added to obtain the final concentration of 1. After a 1 h incubation at room temperature or 4oC as indicated, 4 [mu]l 10% Ficoll (Pharmacia) were added. The samples were loaded on a 20% acrylamide/bis- acrylamide gel (29:1) in buffer B and run at 3 mA for 24 h at room temperature, except when specified. The gel and the glass plate covering the gel were autoradiographed with Kodak XOmat-AR films. At high [1] the radioactive material found in the gel wells was also adsorbed on the cover plate. Thus the photographs shown in Figures 3 -5 present superimposition of the radioactivity observed on both the gel and on its glass plate cover.
Absorption measurements
The measurements were performed with Kontron 933 or 942 spectrophotometers. The temperature was usually regulated at 20.0oC and measured within the cell using a temperature sensor. Job's continuous variation method was applied to determine binding stoichiometries by following the absorption at 362 nm (40 ). The Tm measurements were performed using constant heating rates of 0.2 oC/min for oligomers and 1 .0oC/min for polynucleotides. Tm values were calculated from the first derivative of the melting curves. The melting profiles were monitored at 270 and 260 nm and corrected for baseline fluctuations by subtracting the absorbance at 350, for the free DNA, 362 and 270 nm and corrections of the baseline were done by subtraction of the optical density at 500 nm in the presence of 1 . A wavelength of 270 nm was chosen because of the low contribution of 1 to the total absorbance based on the [epsilon] value of free 1 .
Fluorescence measurements
The fluorescence measurements were performed with a Fluoromax instrument (Spex) equipped with an Hamamatsu 931 photomultiplier (PM) and a thermostated cell holder. The data are corrected for the PM response and for variation of absorption at the excitation wavelength. The DNA hairpins used for fluorescence measurements were futher purified by ethanol precipitation and desalted on Sephadex G10 columns. Repeated fluorescence spectra of bound 1 to oligomers induce spectral modifications under continuous irradiation due to a photochemical reaction that was avoided by taking short integration times and short excitation pathlengths l = 0.4 or 0.2 cm. The relative fluorescence yields were obtained from the ratio of the integrated fluorescence spectrum of the complex 1-DNA to that of the free macrocycle. This ratio was corrected for the difference in absorption at the excitation wavelength. Fitting of the relative fluorescence yields as a function of the [1]/[(sA5[middot]sT5)] concentration ratios was performed using Mathematica software (Stephen Wolfram) that numerically solves sets of equations, searching for roots by the method of Newton (42 ,42 ).
RESULTS
Macrocycle 1 shifts the equilibrium from a duplex towards hairpins
The oligomers sA5 and sT5 adopt hairpin structures with different loops and the same self-complementary stem (38 ). Thus sA5 and sT5, when mixed in an equimolar ratio, can adopt a duplex form (sA5[middot]sT5) in equilibrium with the hairpin structures depending on the salt and temperature conditions. The experiments below show that macrocycle 1 modifies this equilibrium.Fluorescence measurements. The fluorescence properties of the macrocycle enable single-stranded and double-stranded DNA to be distinguished (38 ). The fluorescence maximium is red shifted when the macrocycle binds to duplexes such as poly(dG[middot]dC) and poly(dA[middot]dT) compared with single-stranded oligomers. These differences are exploited to monitor the sA5 and sT5 equilibrium reaction. Figure 1 presents the fluorescence spectra of the macrocycle in the presence of a constant concentration of sA5 and sT5. In the presence of 0.2 macrocycle equivalents in buffer B, [1] = 0.2 * [sA5] = 0.2 * [sT5], the fluorescence maximum is observed at 512 nm (Fig. 1 A, spectrum a). A similar peak is also obtained with poly(dA)[middot]poly(dT) in the presence of 1. In addition, melting temperature experiments of macrocycle-free (sA5@sT5) solutions show that duplex is the predominant structure at 20oC (not shown). Therefore, the 512 nm maximum in Figure 1 A(a) is consistent with the macrocycle mainly bound to the duplex form (sA5[middot]sT5). An increase in the macrocycle concentration induces a clear blue shift of the fluorescence maximum (compare spectra a-c in Fig. 1 A).
When lower ionic strength conditions were used, namely buffer A, the melting temperature of the macrocycle-free (sA5[middot]sT5) solution was 24oC, implying that ~50% of the duplex melted at 20oC. Therefore, the fluorescence spectrum observed at low macrocycle concentrations is expected to be intermediate between that of 1 bound to the duplex and to hairpins. This is indeed the case, as shown in Figure 1 B, spectrum a. The presence of higher macrocycle concentrations induces a shift of the peak towards shorter wavelengths, following the same trend as in Figure 1 A: the spectrum obtained with 1 macrocycle equivalent resembles that in the presence of the sT5 hairpin, with some contribution of that of 1-sA5 (Fig. 1 B, spectra b and c respectively).
The blue shift of the macrocycle fluorescence is concomitant with a decrease in the relative fluorescence yield, shown in Figure 2 for the data obtained in the PIPES buffer. This figure also presents the calculated values according to the equilibrium equations (below) and law of mass conservation for sA5, sT5 and 1. The data could only be fitted if one assumes a 1:1 stoichiometry for the binding of 1 to (sA5[middot]sT5) duplex. The formation of 1:1 complexes between 1 and the hairpins is determined by Job plots. The binding constant for the sA5 hairpin has been measured by titration (absorption spectroscopy), KA = (1.5 +- 0.2) * 107 M-1. The equilibrium constant K1 is derived from melting temperature experiments, K1 = 1 * 107 M-1. The seven equations are used to fit the experimental data by numerical calculation. Three best sets of binding constants among the possible roots are found; the deviations of the calculated points from the measured ones are within the experimental error (Fig. 2 ).
Aggregation of the DNA oligomers induced by the macrocycle
The lack of migration observed in Figure 3 may be related to aggregation and subsequent precipitation. The latter is evidenced by an absorption decrease of oligomers belonging to the (A)n series, where n = 4, 9, 30 and ~220; it also occurs with all the hairpins tested when measurements are performed at constant [DNA] (data not shown). This effect takes place on a days timescale, if it is not accelerated by heating or centrifugation. This macrocycle-induced absorption decrease is observed from a [1] threshold relative to the [DNA], which strongly decreases with molecular weight up to 104, then stays approximately constant. It is observed at (1.4 +- 0.2), (0.35 +- 0.02), (0.12 +- 0.02) and (0.09 +- 0.01) macrocycle equivalents per phosphate for (A)4, (A)9, (A)30 and poly(A) respectively. The macrocycle threshold for hairpin aggregation follows a similar decrease with molecular weight as that observed with the (A)n series. The greater extent of sA5 aggregation relative to that of sT5 is consistent with the higher amount of radioactivity in the wells for the former than for the latter (Fig. 3 , lanes 20 and 14 respectively).
Figure 4 shows that migration of the DNA fragments is abolished above a macrocycle threshold, which seems to depend on the DNA molecular weight: a much lower intensity of radioactivity is observed with (A)4 relative to (A)9 (compare lanes 28-30 with 24-26). The same holds for sA3 compared with sA5 (lanes 20-22 relative to 2-6). A previous spectroscopic work showed that the macrocycle forms a soluble complex with the sA3 hairpin (38 ) at a 1:1 stoichiometry, i.e. one macrocycle per sA3 strand (type I complex). A second complex (type II complex) was observed for an excess of 1 over sA3, presenting a low solubility. Figure 4 shows that migration of the type I complex between sA3 and 1 cannot be distinguished from that of the free probe, while the type II complex appears as a non-migrating band in the wells (lanes 19-22). The lack of migration may originate from electrostatic interactions between a negatively charged DNA and a +4 charged macrocycle. Electrophoresis performed on agarose gels shows that the type II complex between the macrocycle and sA3 does not migrate, neither to the negative nor to the positive electrode (data not shown). This suggests that the global electric neutrality of the aggregates results in non-migration in the electric field. The hairpins sA5, sTAR and B23 and the single-stranded oligomers (A)9 and (A)4 follow the same trend as sA3 in the presence of the macrocycle, pointing to a common and general process. Their type II complexes are therefore likely to be neutral aggregates, as in the case of sA3. (Designation of the bands as type I or type II complexes is not intended to specify their stoichiometry, but rather describes the difference in migration between the species.)
Competition between different structures for binding of macrocycle 1
Competition experiments enlighten the selectivity of the macrocyclic acridine for hairpins relative to double helical DNA.Binding of the macrocycle to hairpin B23 as compared with double-stranded D12: competition between (sA3[middot]sA3) duplex, sA3 hairpin and (A)4on binding of 1. The B23 hairpin and the D12 duplex are well-characterized sequences (11 ,25 ,27 ) possessing similar molecular weights. This ensures a comparable electrostatic attraction to 1. Figure 5 A compares the ability of the macrocycle to retain the B23 hairpin and the D12 oligomer in the wells during gel electrophoresis. Migration of each oligomer alone is first checked. When a low concentration of D12 is used ([D12] = 0.3 [mu]M) the dodecamer adopts mainly a single-stranded (hairpin) structure (lane 1) (11 ,25 ). At 30 [mu]M strand concentration ~50% duplex and 50% single-stranded conformation are formed (lane 2). These last conditions were used to analyze the binding of 1 in a mixture containing D12 and B23. Alternative labeling of D12 oligomer (lanes 4-10) or B23 (lanes 11-17) was performed because the duplex band of D12 migrates similarly to the B23 hairpin, as expected from their similar molecular weights. Migration of the (D12[middot]D12) duplex is abolished by [1] = 120 [mu]M (lane 9), whereas that of the B23 hairpin band vanishes when [1] = 30 [mu]M (lane 14), in spite of a 10-fold molar excess of D12 compared with B23. These results suggest a preference of the macrocycle for a hairpin structure as compared with a duplex of similar size.
Figure 5 B presents a similar competition between the sA3 hairpin and bulged duplex and (A)4 on binding of the macrocycle. This experiment was run at 4oC, using 10 [mu]M sA3 and (A)4. The low temperature favors formation of (sA3[middot]sA3) bulged duplex (band d) together with some hairpin form (band b, lane 1). Addition of 1 macrocycle equivalent causes the disappearance of the hairpin form (lane 2). A further increase in [1] induces a decrease in the intensity of the (A)4 band, while the (sA3[middot]sA3) bulged duplex band remains constant. Thus, Figure 5 B qualitatively indicates that the affinity of 1 for the sA3 hairpin is larger than that for (sA3[middot]sA3), in spite of double the number of phosphate groups in the latter bulged duplex relative to the hairpin. This is in agreement with the results of Figure 5 A.Premelting of poly(dA[middot]dT). The above results suggest that the macrocycle may induce local unpairing of a long double helix such as poly(dA[middot]dT). Melting temperature experiments were performed with this polymer with increasing macrocycle concentrations in buffer C. Figure 6 A shows that under these conditions, the Tm of free poly(dA[middot]dT) is 62 +- 1oC (curve a). In the presence of 0.1 macrocycle equivalent, a biphasic melting profile is observed (curve b). One macrocycle equivalent produces a multiphasic profile that becomes much less cooperative than that of the free polymer (Fig. 6 A, curve c). These lower derivatives could correspond to local unpairing of poly(dA[middot]dT), creating `bubbles'. The single-stranded nature of the binding site for 1 is further confirmed by fluorescence measurements (Fig. 6 B). The increase in the macrocycle concentration induces a progressive blue shift of the 512 nm peak; this fluorescence maximum corresponds to a double-stranded environment. The spectrum obtained with 0.8 1 equivalent peaks at 444 nm, consistent with a single-stranded adenine-rich environment (38 ).
DISCUSSION
Evidence forDNA aggregation above a threshold[1]relative to the [DNA]
The interactions of the macrocycle with DNA fragments, monitored by gel electrophoresis, give rise to two types of complexes. The type I complex migrates as the free probe. This is in agreement with previous spectroscopic results showing the formation of a soluble 1:1 complex between sA3 and 1. The experiment depicted in Figure 4 shows the appearance of a non-migrating band in the wells (type II complex) at the expense of a band migrating as the free oligomer, occuring above a macrocycle concentration threshold. Absorption data provide evidence for DNA aggregation in the presence of 1. This process presents some similarity with DNA condensation induced by polyamines (45 -48 ). Onset of aggregation by 1 seems to depend on the molecular weight of the DNA fragments, as previously suggested for DNA condensation by spermine (45 ,48 ) Aggregation of DNA fragments induced by spermine, spermidine and cobalt hexaamines may be due to short-range electrostatic attraction between a highly charged polyelectrolyte like DNA and the condensing ions (45 ). These processes probably originate from the diethylene triamine arms of macrocycle 1 (37 ,38 ).
The macrocycle shifts the equilibrium from duplexes towards hairpins or single-stranded structures
The macrocycle shifts the equilibrium from a duplex to hairpins. The complementarity of sA5 and sT5 sequences allows the establishment, at low ionic strength, of an equilibrium between a duplex form and the corresponding hairpins (Scheme 2 ). According to previous results (38 ), it is expected that the macrocycle can shift the (sA5[middot]sT5) duplex towards the corresponding hairpins. However, electrostatic interaction between the duplex and 1 is greater than that with the hairpins.
We have tested the ability of the macrocycle to disrupt the (sA5[middot]sT5) duplex by two techniques. The first is fluorescence spectroscopy, performed under conditions where [1] <= [(sA5[middot]sT5)]. Therefore, the fluorescence data are obtained under true equilibrium conditions, where type I complexes mainly exist. In contrast, the gel electrophoresis experiments used an excess of [1] compared with [(sA5[middot]sT5)], under the same ionic conditions. These data follow the formation of type II complexes upon increasing [1].
The macrocycle presents different fluorescence spectra depending on the structure of the DNA to which it binds. This is clearly seen from the fluorescence peak at 512 nm in the presence of (sA5[middot]sT5) duplex [Fig. 1 A(a)], whereas the maxima are observed at 431 and 442 nm for 1 bound to sT5 and sA5 hairpins respectively [Fig. 1 A(d) and B(c)]. The displacement of this equilibrium was monitored by blue shift of the maximum with increasing [1]. The quantification of these data yielded three sets of solutions. The roots are characterized by calculated binding constants of 1 to the duplex that are lower than the macrocycle affinity for the hairpins by roughly one order of magnitude. In conditions where type I complexes exist, the (sA5[middot]sT5) duplex is disrupted by the macrocycle because of its higher affinity for hairpins. Scheme 2 shows an extreme case of sT5 representing the unique macrocycle binding site. In fact, the probability of the macrocycle binding to sA5 can be as high as that for sT5 according to the first solution. The important point depicted in this scheme is that the macrocycle forms a type I complex under these experimental conditions.
A macrocycle-induced shift of the (sA5[middot]sT5) duplex towards hairpin structures can only be visualized in a gel electrophoresis experiment by a simultaneous decrease in the duplex and one hairpin and an increase in the other hairpin band. If the macrocycle was binding to the duplex and did not shift the equilibrium, a decrease in all migrating bands should then be observed, since 1 also binds to the hairpins. Figures 3 and 4 show that the extent of macrocycle-induced aggregation of sA5 is larger than that of sT5. Therefore, one expects a decrease in the duplex and sA5 bands together with an increase in that of sT5 when [1] increases. This is exactly what is observed. The decrease in sA5 and in the duplex together with the increase in sT5 occur simultaneously (Fig. 3 ). The identity of the sT5 band, which increases with increasing [1], is clearly demonstrated by alternative labeling of sA5 and sT5. The remaining band a is probably not free sT5, although the resolution of the gel does not show the corresponding type I complex (see Scheme 2 ). These results point out qualitatively that the macrocycle disrupts the (sA5[middot]sT5) duplex while stabilizing the hairpins.
Extrapolation of these results suggests that the macrocycle should be able to sequester hairpins within a mixture of structures and to disrupt, at least locally, a long duplex.Hairpins compete successfully with duplexes on binding of the macrocycle. The present competition experiments are based on previous data showing that the macrocycle specifically stabilizes the sA3 hairpin, but not a duplex that mimics the sA3 stem (38 ). Macrocycle 1 is located in/on the loop of hairpins sA3, sA5, sT5 and sTAR. This attribution relies on the variation of the fluorescence spectra of the macrocycle bound to these hairpins possessing different loops but a common stem. There was no spectroscopic evidence for binding of 1 to the stems of these hairpins.
The competition between the sA3 hairpin, (sA3[middot]sA3) bulged duplex and (A)4 illustrates the higher affinity of 1 for the hairpin compared with the duplex and single-stranded oligomers, shown by the disappearance of the sA3 hairpin band, whereas the duplex band intensity stays constant (Fig. 5 B). In this experiment the electrostatic interaction between 1 and the sA3 hairpin is weaker than that between the duplex and 1. Nevertheless, the hairpin competes successfully with the bulged duplex for macrocycle binding.
The competition between the D12 oligomer and the B23 hairpin (Fig. 5 A) supports the above hypothesis. Despite the 10 times higher D12 concentration relative to B23, preferential binding of 1 to the B23 hairpin rather than the D12 duplex is observed. There is no a great difference between the D12 duplex and hairpin forms for macrocycle binding. This may be explained in terms of the lower electrostatic interaction between the D12 hairpin and 1 relative to that of the corresponding duplex. It is also possible that the AATT loop sequence of the D12 hairpin does not favor binding of the macrocycle, in contrast to the above more favorable hairpin loop sequence AAA. Nevertheless, the hypothesis that the macrocycle may disrupt the D12 duplex while sequestering the single-stranded D12 structure cannot be excluded.
The gel electrophoresis results confirm the higher affinity of the macrocycle for hairpins relative to duplexes (of comparable molecular weight) suggested by spectroscopic methods.Relative affinity of the macrocycle for single-stranded fragments compared with a double helix: premelting of poly(dA[middot]dT). The polynucleotide poly(dA[middot]dT) forms a double helix, melting at Tm = 62oC. The melting profile is modified by the macrocycle, which becomes bi- or multiphasic. The main melting temperature is lowered by 16oC compared with the free polymer in the presence of 1.8 [mu]M [1]. This is attributed to a local unpairing of part of this polymer. The fluorescence data confirm the single-stranded nature of the macrocycle binding site, by the progressive blue shift of the fluorescence maxima with increasing [1]. This suggests that the macrocycle should be able to create or to trap transient `bubbles' in a double helix, at least in AT-rich regions, at 100 mM salt, close to physiological conditions.
These data present some similarity with the destabilization of poly(dA[middot]dT) by the heterogeneous nuclear ribonucleoprotein A1 (49). It is interesting to note that a tetracationic cyclophane stabilizes poly(dA)[middot]poly(dT) and poly(dA[middot]dT) against heat denaturation (50), in contrast to the macrocyclic bis-acridine 1, which disrupts the latter polymer. The ability of an organic compound to disrupt a duplex may not be related only to an affinity difference between double-stranded and single-stranded polynucleotides, but also to the potential structures the latter may adopt in the presence of this compound. We have previously reported for a porphyrin-containing macrotetracyclic receptor two orders of magnitude higher affinity constants for poly(dA) and poly(dT) relative to poly(dA[middot]dT), although the Tm of the duplex was not significantly modified by this compound (51 ).
CONCLUSION AND BIOLOGICAL PERSPECTIVES
Binding of the macrocyclic bis-acridine 1 to DNA is characterized by formation of two types of complexes that differ greatly in their mobility in an electric field. The non-migrating complexes are attributed to aggregates, formed via electrostatic interactions between the macrocycle and the nucleic acid.
Competition experiments demonstrate clearly the higher affinity of the macrocycle for hairpins, compared with duplexes. This difference in affinity enables the macrocycle to shift the equilibrium from a duplex towards the corresponding hairpins. The macrocycle destabilizes a synthetic polynucleotide, creating premelted `bubbles'. By extrapolation of these results, the macrocycle should be able to disrupt, at least locally, a long duplex such as genomic DNA and transiently stabilize unpaired areas, especially hairpins.
The structure of nucleic acids often triggers or modulates important biological processes. DNA hairpins are indeed involved in such different systems as activation (52,53), repression (15 ) or termination (14 ,54 ) of transcription and in the control of replication (12 ,13 ,55 ,56 ). Such structures have also been shown to be directly associated with various diseases, such as retrovirus-induced mouse tumors (57 ), human fragile X syndrome (18 ) and human myotonic dystrophy (19 ).
Organic compounds, such as macrocycle 1, that show specific binding to hairpin structures may be useful probes to trap and/or stabilize transient DNA hairpins, allowing further analysis of their function. Macrocycle 1 may compete with single-stranded DNA proteic factors for binding to response elements, thus providing a good tool to study the role of such proteins. Conformation changes occurring during the course of transcription may be modified by a compound that specifically binds to hairpins (see Introduction). To the best of our knowledge, macrocycle 1is the first compound that displaces the equilibrium from duplex to hairpin structures. Lastly we have to consider the potential photochemical reactivity of 1 in the perspective of specific cleavage of hairpins associated with diseases.
ACKNOWLEDGEMENTS
Dr F.Livolant and Prof. A.Cao are gratefully acknowledged for helpful discussion, T.Bataille for the drawings and Dr Y.Maurin for his help with the photographs.
REFERENCES
1 Van Holde,K. and Zlatanova,J. (1993) BioEssays, 16, 59-68.
2 Bianchi,M.E., Beltrme,M. and Paonessa,P. (1989) Science, 243, 1056-1059.MEDLINE Abstract
41 Wolfram,S. (1991) Mathematica: A System for Doing Mathematics by Computers. Addison Wesley Publishing Co., Redwood, CA.
42 Press,W.H., Flannery,B.P., Teukolsky,S.A. and Vetterling,W.T. (1989) Numerical Recipes: The Art of Scientific Computing. Cambridge University Press, Cambridge, UK.
*To whom correspondence should be addressed at present address: Chimie des Interactions Moléculaires, UPR CNRS 285, Collège de France, 11 Place Marcelin Berthelot, 75231 Paris cedex 05, France. Tel: +33 1 43 25 26 09; Fax: +33 1 43 29 80 88; Email: schwok@ibpc.fr
