Is a Closing “GA Pair” a Rule for Stable Loop-Loop RNA Complexes?*

RNA hairpin aptamers specific for the trans-activa-tion-responsive (TAR) RNA element of human immunodeficiency virus type 1 were identified by in vitro selection (Duconge´, F., and Toulme´, J. J. (1999) RNA 5, 1605–1614). The high affinity sequences selected at physiological magnesium concentration (3 m M ) were shown to form a loop-loop complex with the targeted TAR RNA. The stability of this complex depends on the aptamer loop closing “GA pair” as characterized by preliminary electrophoretic mobility shift assays. Thermal denaturation monitored by UV-absorption spectroscopy and binding kinetics determined by surface plasmon resonance show that the GA pair is crucial for the formation of the TAR-RNA aptamer complex. Both thermal denaturation and surface plasmon resonance experiments show that any other “pairs” leads to complexes whose stability decreases in the order AG > GG > GU > AA > GC > UA >> CA, CU. The binding kinetics indicate that stability is controlled by the off-rate rather than by the on-rate. Comparison with the complex formed with the TAR* hairpin, a rationally designed TAR RNA ligand (Chang, K. Y., and Tinoco, I. (1994) Proc. Natl. Acad. Sci. U. S. A. 91, 8705– 8709), demonstrates that the GA pair is a key determinant which accounts for the 50-fold increased stability of the TAR-aptamer complex ( at a volume and a concentration suitable for each experiment and stored at 2 20 °C. Before the experiments, each RNA sample used was heated at 95 °C during 1 min and then put on ice for 10 min to avoid the formation of intermolecular complexes. Thermal Denaturation of RNA Complexes— Thermal denaturation of RNA complexes in 20 m M sodium cacodylate buffer, pH 7.3, at 20 °C, with 140 m M potassium chloride, 20 m M sodium chloride and 3 m M magnesium chloride (R 9 buffer) was monitored on a Cary 1 spectropho- tometer interfaced with a Peltier effect device that controls temperature within 6 0.1 °C. Denaturation of the samples was achieved by increas- ing the temperature at 0.4 °C/min from 5 °C to 90 °C and was followed at 260 nm. A cuvette that contained R 9 buffer was used as the reference. Except cacodylate, which replaced the temperature-sensitive HEPES, the buffer used for these thermal denaturation experiments was equiv- alent to the one used during the in vitro selection process (25). As the absorbance of the TAR RNA at 260 nm is too large to allow accurate monitoring of the absorption change resulting from the denaturation of the bimolecular complex between the HIV-1 RNA and the RNA aptamer, the experiments were carried on with miniTAR, a 27-mer oligonucleotide, instead of the entire TAR hairpin, which is 59 nt. RNA samples were prepared at 1 m M final concentration in the mixture. They were mixed at room temperature and allowed to interact 30 min before cooling down to 5 °C. The experiment then started after 1 h atthis temperature. The enthalpy change, D H , for the formation of the bimolecular com- plex, was deduced from the total RNA concentration dependence of the T m according to Equation 1.

Intermolecular interactions between structured RNA play key roles in the regulation of gene expression. In Escherichia coli, the replication of the ColE1 plasmid is regulated by the interaction of two RNA transcripts, RNA I and RNA II, which fold as hairpins (1,2). The interaction starts with base pairing between the complementary loops of these RNAs and leads to the formation of a double-stranded RNA along the entire length of RNA I, thus disrupting the RNA II hybridization with the template DNA required for replication (3,4). Even if the sta-bility of these so-called "kissing" complexes is primarily based on loop complementarity (2), factors such as the orientation of the loops (5), the loop closing base pair, and the sequence of each stem next to the loop (6), are crucial for stability. For instance, the loop inversion 5Ј to 3Ј induces a 350-fold increased stability of the RNA I-RNA II complex.
In HIV-1, 1 the dimerization of the genomic RNA involves the formation of a loop-loop complex between two structured regions (7). The dimerization initiation site of HIV-1 RNA folds as a hairpin, which is closed by a noncanonical AA pair. The 9-nt-long loop contains a 6-nt self-complementary sequence flanked by two 5Ј and one 3Ј purines, which, together with loop complementarity, are crucial for the dimerization process (8). Non-Watson-Crick interactions in RNA molecules have been also reported in the viral RNA element bound by the Rev protein of HIV-1 (9), in GRNA tetraloops (10,11), tRNAs (12)(13)(14), and tandem mismatches within duplexes (15)(16)(17). All these results indicate that interactions other than canonical base pairs contribute to the structural diversity displayed by RNAs, which, as a matter of fact, is crucial for activity. The contribution of such noncanonical interactions can actually be conveniently explored by in vitro selection, since neither the structure of the target nor the interactions between the target and the interacting aptamer need to be known (18,19). This strategy was successfully used in our laboratory to identify DNA aptamers against DNA (20 -22) and RNA hairpin structures (23).
Recently, RNA aptamers specific for the trans-activationresponsive (TAR) RNA (24) were selected by in vitro selection (25). The TAR RNA element is a 59-nt-long imperfect stem loop structure located at the 5Ј end of the retroviral RNA. A 3-nt bulge in the upper part of the hairpin constitutes part of the binding site of the viral protein Tat, which recruits cyclin T 1 . Together with additional TAR-bound cellular proteins, this complex prevents abortion of the transcription of the retroviral genome. Therefore, TAR plays a key role in the life cycle of HIV-1 and constitutes a valid target for the development of ligands, which could inhibit its interaction with viral and cellular proteins, thus ultimately preventing the development of the virus. The isolated high affinity anti-TAR aptamers were shown to fold as imperfect hairpins and form kissing complexes with the targeted RNA at physiological magnesium concentration. The apical loop of all these aptamers corresponds to the 5Ј-GUCCCAGA-3Ј consensus sequence, the six central bases of which are complementary to the entire TAR loop whereas a conserved "GA pair" closes the loop. As for ColE1 RNA I and RNA II, and HIV-1 dimerization initiation site RNA motif, interactions other than loop complementarity are crucial for stability. Indeed, several mutations of the GA pair that closes the loop of the identified RNA aptamers decrease the stability of the TAR-aptamer complex, as shown by preliminary electrophoretic mobility shift assays (EMSA).
In the work presented herein, the role of this GA pair was investigated at physiological concentration of magnesium by systematically mutating the loop closing pair of the aptamer. The effects of these mutations on the stability of the TAR RNA-aptamer complex were analyzed by thermal denaturation monitored by UV-absorption spectroscopy (26,27). The binding kinetics were determined by using surface plasmon resonance (SPR). This physical phenomenon is used to follow in real time the interaction between a molecule in a continuous flow and an immobilized one (28). Numerous studies have been published on protein-protein interactions, protein-ligand interactions, and protein-nucleic acid interactions, but a much reduced number of investigations of nucleic acid-nucleic acid interactions are available (29 -31). As a matter of fact, no RNA-RNA complexes have been analyzed up to now.
Our results demonstrate that the aptamer loop closing GA pair is crucial for the stability of the TAR-RNA aptamer complex once formed. This likely explains the higher stability of the loop-loop complex formed by TAR with the aptamer, at physiological magnesium concentration, compared with the one formed with the rationally designed hairpin, TAR*, whose loop is closed by a UA pair (32). The increased stability of a loop-loop complex formed between RNA IIЈ and a RNA IЈ mutant in which the GC pair next to loop was replaced by GA suggests that closing GA pair could be preferred for kissing complexes.

MATERIALS AND METHODS
Oligonucleotides-All RNA molecules including the biotinylated TAR RNA were synthesized on an Expedite 8908 synthesizer and purified by electrophoresis on denaturing polyacrylamide gels. The pure samples were desalted on Sephadex G-25 spin columns. To avoid repeated thawing and freezing of the stock solutions, the samples were aliquoted at a volume and a concentration suitable for each experiment and stored at Ϫ20°C. Before the experiments, each RNA sample used was heated at 95°C during 1 min and then put on ice for 10 min to avoid the formation of intermolecular complexes.
Thermal Denaturation of RNA Complexes-Thermal denaturation of RNA complexes in 20 mM sodium cacodylate buffer, pH 7.3, at 20°C, with 140 mM potassium chloride, 20 mM sodium chloride and 3 mM magnesium chloride (RЈ buffer) was monitored on a Cary 1 spectrophotometer interfaced with a Peltier effect device that controls temperature within Ϯ0.1°C. Denaturation of the samples was achieved by increasing the temperature at 0.4°C/min from 5°C to 90°C and was followed at 260 nm. A cuvette that contained RЈ buffer was used as the reference. Except cacodylate, which replaced the temperature-sensitive HEPES, the buffer used for these thermal denaturation experiments was equivalent to the one used during the in vitro selection process (25). As the absorbance of the TAR RNA at 260 nm is too large to allow accurate monitoring of the absorption change resulting from the denaturation of the bimolecular complex between the HIV-1 RNA and the RNA aptamer, the experiments were carried on with miniTAR, a 27-mer oligonucleotide, instead of the entire TAR hairpin, which is 59 nt. RNA samples were prepared at 1 M final concentration in the mixture. They were mixed at room temperature and allowed to interact 30 min before cooling down to 5°C. The experiment then started after 1 h at this temperature.
The enthalpy change, ⌬H, for the formation of the bimolecular complex, was deduced from the total RNA concentration dependence of the T m according to Equation 1.
The number of magnesium ions that bind upon formation of the complex, ⌬Mg 2ϩ , was determined from the ion concentration dependence of the melting temperature, T m , according to Equation 2.
Surface Plasmon Resonance Kinetic Measurements-SPR experiments were performed on a BIAcore 2000 apparatus (Biacore AB, Sweden) running with the BIAcore 2.1 software. Biotinylated TAR RNA, 59 nt long, was immobilized on CM5 sensorchips coated with streptavidin according to the procedure described in the BIA applications handbook. In these conditions, 5000 resonance units (RU) of streptavidin (Sigma), equivalent to 5 ng/mm 2 , were immobilized on the chip which subsequently was allowed to equilibrate at 23°C, the temperature of the in vitro selection, in the selection buffer: 20 mM HEPES, pH 7.3, at 20°C containing 20 mM sodium acetate, 140 mM potassium acetate, and 3 mM magnesium acetate (R buffer). Biotinylated TAR RNA (10 -50 nM) was prepared in this buffer and then injected at a flow rate of 5 l/min. The injection was stopped as soon as 500 -600 RU of bound TAR RNA was reached. This amount was shown to be appropriate to keep the pseudo-first order kinetic condition and to allow good reliability of recorded sensorgrams (RU versus time) in terms of signal to noise ratio. One noncoated or one streptavidin-coated flow-cell was used to check for nonspecific binding of RNA aptamers. The signals from these control channels served as base lines and were subtracted to the RU change observed when an injected RNA aptamer interacts with the immobilized TAR RNA. SL1 RNA, a nonrelevant hairpin from the hepatitis C virus RNA, was used as a negative control (33). The sensorchip surface was successfully regenerated with three 5-l pulses of 25% formamide, followed by one 5-l pulse of distilled water and finally one 10-l pulse of R buffer.
Nonlinear regression analysis of single sensorgrams at five concentrations, at least, of injected RNAs was used to determine the kinetic parameters of the complex formation. The data were analyzed with the BIA evaluation 2.2.4 software, assuming a pseudo-first order model, according to Equations 3-5, for the association and dissociation phases, respectively, where R is the signal response, R max the maximum response level, C the molar concentration of the injected RNA molecule, k on the association rate constant, and k off the dissociation rate constant.
To check for self-consistency of data, k obs , derived from nonlinear analysis, was plotted as a function of the RNA concentration according to Equation 5.

RESULTS
The RNA molecules used for this study are shown in Fig. 1. MiniTAR (27 nt long), the upper half part of TAR, was shown to be the mininal domain necessary and sufficient for responsiveness in vivo (34). It interacts with the anti-TAR selected aptamers without changes in the affinity compared with the fulllength TAR (35). R-06 24 (GA) is the aptamer of highest affinity identified by in vitro selection against TAR, with the consensus motif 5Ј-GUCCCAGA-3Ј, LR-06 8 , in the apical loop. The six central bases of the consensus sequence are complementary to the entire TAR loop. TAR* is a hairpin rationally designed to interact with TAR (32). Its loop is fully complementary to the TAR one. RNA IЈ and RNA IIЈ are two structured RNAs derived from the two RNA transcripts, RNA I and RNA II, involved in the control of the ColE1 plasmid regulation. The sequences of RNA IЈ and RNA IIЈ were modified in the stem to avoid the formation of an extended duplex, as seen with the biological RNAs once the kissing complex is formed (6).

Detection of the Loop-Loop interaction by Thermal
Denaturation-The derivative of UV melting curves of miniTAR with R-06 24 (GA) aptamer, with TAR* RNA and with LR-06 8 are reported in Fig. 2. The melting profiles obtained with mixtures of miniTAR and R-06 24 (GA) RNA at two different concentrations in RЈ buffer display two transitions ( Fig. 2A). Only one transition is observed for the melting profiles of the RNA hair-pins alone. On diluting the miniTAR and R-06 24 (GA) aptamer mixture 4-fold (from 2 M to 0.5 M), the T m (the maximum of the derivative plots) of the lower transition decreases from 47.5°C to 40°C as expected for a bimolecular complex, whereas the T m of the higher transition remains unchanged and thus can be ascribed to the melting of the RNAs alone.
We then compared the stability of the miniTAR and R-06 24 (GA) aptamer mixture to that of two reference complexes: miniTAR with either TAR* RNA or LR-06 8 (Fig. 2B). Under the ionic conditions used for the in vitro selection, T m for the complex with TAR* is equal to 30.7°C (Table I), i.e. 16.8°C below that of the miniTAR-R-06 24 (GA) complex. Finally, T m for the complex with LR-06 8 , the 8-mer RNA 5Ј-GUCCCAGA-3Ј consensus motif of the aptamers, is equal to 20.3°C. Clearly, interaction of miniTAR with the aptamer gives rise to the most stable bimolecular complex. On one hand, the higher stability of the TAR-aptamer complex over that of the 8-mer RNA can be ascribed to the hairpin structure of the ligand as previously demonstrated (36). On the other hand, the comparison between TAR* and the selected aptamer strongly suggests that the sequence outside the loop plays a crucial role in the extra stability displayed by TAR-aptamer complexes compared with the TAR-TAR* one.
The stability of RNA structures and complexes is well known to depend on magnesium ion. We then wanted to address the question whether the differential behavior between TAR* and the aptamer would originate in the number of associated Mg 2ϩ ions. This can be achieved by measuring the variation of T m as a function of the Mg 2ϩ concentration if the enthalpy change, ⌬H, for the transition is known. The enthalpy change for the formation of the complex, ⌬H, was deduced from the slope of the total RNA concentration dependence of the T m (Fig. 3), according to Equation 1. In these experiments, the concentration of Mg 2ϩ was decreased from 3 mM in RЈ buffer to 1 and 0.1 mM, for the miniTAR-TAR* and miniTAR-aptamer complexes, respectively, to accurately measure the T m values of bimolecular complexes in the total RNA concentration range chosen, 0.5-16 M, with no interference with the melting of the hairpins alone. Under these conditions, ⌬H is equal to Ϫ42.8 Ϯ 1.4 kcal/mol and Ϫ39.2 Ϯ 1.3 kcal/mol for miniTAR-R-06 24 (GA) and miniTAR-TAR* complexes, respectively. The variation of T m as a function of the Mg 2ϩ concentration is presented in Fig.  4. For both complexes, linearity of 1/T m versus ln[Mg 2ϩ ] plots indicates a site binding mode of the magnesium ion. The number of ions, ⌬Mg 2ϩ , which bind was deduced according to Equation 2 using the enthalpy change. This gives ⌬Mg 2ϩ ϭ 1.7 Ϯ 0.1 and 1.4 Ϯ 0.1 for the complexes formed with R-06 24 (GA) and TAR*, respectively. It demonstrates that the different stability of these complexes does not originate in a difference in the number of magnesium ions that each complex binds. From this ⌬H and the T m values measured at 3 mM magnesium (Table I), the equilibrium dissociation constant, K d , in solution at 23°C, for both complexes can be easily determined. K d , under the in vitro selection conditions, is equal to 2.0 Ϯ 0.4 nM and 92.5 Ϯ 5.2 nM for miniTAR-R-06 24 (GA) and miniTAR-TAR* complexes, respectively. Clearly, under this condition, the miniTARaptamer complex is more stable than the one formed with TAR*, the rationally designed RNA.
The base pair closing the loop was a marked difference between TAR* and the aptamer molecule. Then, in a second set of  Table I. Clearly, any change of the closing pair induces a decrease in T m that reflects a decrease in the stability of the miniTAR-aptamer complex.
Surface Plasmon Resonance Detection of RNA Complexes-SPR was used to follow, in real time, the interaction of the immobilized full-length TAR RNA on streptavidin-coated sensorchip with various RNA hairpins. Sensorgrams, obtained when R-06 24 (GA) or TAR* analytes were injected over the sensorchip surface at two different concentrations of magnesium ion, are reported in Fig. 6. In all cases, as expected for a pseudo-first order model, the dissociation phase does not show significant dependence on the concentration of the injected analyte whereas the association phase increases with it. Furthermore, these kinetics fulfill pseudo-first order conditions as checked from linearity of plots of the observed rate constant, k obs , versus the analyte concentration (insets). R-06 24 (GA) aptamer binds to TAR RNA at either 3 or 10 mM magnesium, as seen in Fig. 6 (A and B, respectively). In contrast, whereas TAR* binds also at 10 mM (Fig. 6D), it shows a poor affinity at 3 mM (Fig. 6C). The rate constants, k on and k off , deduced from these sensorgrams are listed in Table II. For R-06 24 (GA), the on-rate changes from 6.3 ϫ 10 4 M Ϫ1 s Ϫ1 to 17 ϫ 10 4 M Ϫ1 s Ϫ1 upon increasing the concentration of magnesium from 3 to 10 mM. The off-rate remains constant and is equal to about 10 Ϫ3 s Ϫ1 . For TAR*, at 3 mM magnesium, the rate constants cannot be reasonably determined. At 10 mM, k on and k off are equal to 38 ϫ 10 4 M Ϫ1 s Ϫ1 and 5.3 ϫ 10 Ϫ3 s Ϫ1 , respectively. Also listed in Table II are the equilibrium dissociation constants deduced from these kinetic measurements, K d , equal to k off /k on , which confirms that the TAR-aptamer complex is much more stable in the in vitro selection buffer than the one with the rationally designed TAR* molecule, in agreement with EMSA (25). As expected, at 10 mM magnesium, i.e. the concentration used to investigate the TAR-TAR* complex (32), the stability of both complexes increases. The stabilizing effect observed on the complex with R-06 24 (GA) resides essentially in the association  step, as expected for larger screening of electrostatic repulsions.
The role of the GA closing pair of R-06 24 RNA was further examined by SPR in R buffer (3 mM magnesium). Mutants of the closing pair were injected, and RU variations that resulted from complex formation were monitored as a function of time. The rate constants, k on and k off , deduced from nonlinear regression analysis of sensorgrams are listed in Table III as well as the equilibrium dissociation constant, K d . Except for the AG inversion, which is equivalent to the wild-type aptamer in terms of stability, with a K d of about 20 nM, all other mutations have a negative effect on the stability of the TAR-RNA complex. CU and CA mutations destabilize the TAR-aptamer complex to such a degree that the rate constants cannot be determined. The effects of the other mutations are in between. Closer analysis shows that the equilibrium constant is mainly controlled by the off-rate, which increases when the complex is destabilized, whereas the on-rate shows limited variation.
RNA IЈ-RNA IIЈ Loop-Loop Complex from ColE1 Plasmid-In an attempt to establish whether a closing GA pair is thermodynamically favorable for loop-loop interactions and thus might constitute a rule, we replaced the GC pair next to the 7-nt-long loop of RNA IЈ by a GA pair (Fig. 1). The seven central bases of the new loop, closed by a GA pair, are complementary to the entire RNA IIЈ. The effect of this mutation was analyzed by thermal denaturation. The results obtained for the complex with the GA mutant and the unmodified RNA IЈ are reported in Fig. 7. A clear transition with a T m equal to 23.8°C is seen for the complex with the GA mutant, whereas no significant transition is observed for the wild-type complex. This transition is concentration-dependent (data not shown), as expected for a bimolecular process. The transitions above 60°C are ascribed to the melting of the RNAs alone (concentrationindependent). Interestingly, this new transition is observed despite the lower stability of the GA RNA IЈ mutant compared with the wild-type molecule. This is actually in agreement with the removal of a GC pair. T m is equal to 76°C and 70°C for the latter and the former, respectively (Table I). This indicates that the GA pair is a key structural determinant for the stability of the RNA IЈ-RNA IIЈ complex as well.   24 or TAR* The experiments were performed either in R buffer or in R buffer ϩ 7 mM Mg 2ϩ . K d(BIAcore) was calculated as k off /k on and K d(EMSA) was derived from electrophoretic mobility shift assays (Ducongé & Toulmé, 1999). k on , k off , and K d are the averages and standard deviations of at least five sensorgrams where R-06 24

DISCUSSION
In vitro selection against the trans-activation-responsive RNA of HIV-1 identified RNA hairpin aptamers, which form kissing complexes with the targeted RNA at physiological concentration of magnesium (3 mM) (25). Together with loop complementary, the loop closing GA pair of the aptamer of highest affinity, R-06 24 (GA), is critical for the stability of such complexes, as shown by electrophoretic mobility shift assays. In this work, the contribution of the GA pair to the thermodynamics and the kinetics of the loop-loop interaction was investigated by using UV-absorption spectroscopy and surface plasmon resonance. The results obtained with the 8-nt sequence of the aptamer loop, 5Ј-GUCCCAGA-3Ј, the six central bases of which are complementary to the TAR RNA loop, compared with those obtained with the R-06 24 (GA) aptamer clearly suggest a role for the stem in loop-loop complexes. It supports a view where the higher stability of the complex with the aptamer over the one with the antisense sequence (⌬T m ϭ ϩ20°C) results from the interaction of two structured motifs in agreement with previous results (36). NMR studies on the complex formed between HIV-1 TAR and TAR*, the rationally designed hairpin RNA with a fully complementary loop, have shown that there is a continuous stacking from the 3Ј-side of one stem helix through the loop-loop helix to the other stem helix (37). Similar stacking was observed in the loop-loop interaction between RNA I and RNA II, the two RNA hairpins involved in the regulation of the ColE1 plasmid replication (38). The three-dimensional structure of TAR-R-06 24 complex is not known yet, but one can reasonably expect a similar structure as indicated by enzymatic footprints and by the fact that loop mutants of R-06 24 or TAR RNA no longer formed a complex with the wild-type partner as checked by EMSA (25).
The stabilizing role of the loop closing GA pair is unequivocally demonstrated by the thermal denaturation and kinetics experiments. Regardless of the mutation, the new complex is less stable than the one formed with the R-06 24 (GA) aptamer. Obviously, this explains why a hairpin closed by a GA pair was selected by the in vitro selection over all other combinations. The increased stability of the TAR-R-06 24 (GA) aptamer complex compared with the other complexes cannot be attributed to an increased stability of the aptamer stem itself. As seen in Table I, there is no correlation between the stability of the aptamer variants and the stability of the resulting complex with the targeted RNA. For instance, the more stable variant, R-06 24 (GC) (⌬T m ϭ ϩ10.7°C with respect to the wild-type aptamer) does not give rise to the strongest TAR binder (⌬T m ϭ Ϫ15.9°C compared with the wild type complex). The stabilizing role of the GA pair is further emphasized by the comparison with TAR*. This hairpin RNA has a stem that is 3 base pairs shorter than the aptamer, but it is actually equivalent to the R-06 24 (UA) variant in terms of loop complementary and closing pair. Identical T m values (about 30°C) for TAR-TAR* and TAR-R-06 24 (UA) complexes illustrate this point (Table I). Furthermore, a truncated version of R-06 24 (UA), R-06 16 (UA), now also equivalent to TAR* in terms of stem length (5 base pairs), gives a T m for the bimolecular complex equal to about 30°C too (Table I). This indicates that despite the lower intrinsic stability of this truncated mutant (T m ϭ 62.9°C) compared with the full-length mutant (T m ϭ 75.3°C), the higher affinity of R-06 24 (GA) aptamer (K d ϭ 2.0 nM) over the rationally designed TAR* RNA (K d ϭ 92.5 nM) originates essentially in the GA pair.
Kinetics analysis of mutants indicates that the equilibrium binding constant is controlled by the off-rate rather than by the on-rate. Clearly, the GA pair stabilizes the complex between the aptamer and TAR RNA once formed and prevents it from faster dissociation as observed with the aptamer variants. Under the buffer conditions of the in vitro selection (3 mM magnesium), binding of TAR* to the immobilized TAR RNA is so poor that the kinetics could not be determined. However, we could determine rate constants with the aptamer mutant equivalent to TAR* in terms of loop and closing pair, R-06 24 (UA), and even with its truncated version, the TAR* like R-06 16 (UA) (data not shown). Then, TAR* and the UA variants are not kinetically equivalent. TAR* and the truncated R-06 16 (UA) RNA have the same base composition. They only differ in the stem sequence below the two first base pairs next to the loop: GCU-CGA for the former and CAC-GUG for the latter. Influence of base pairs, next to the loop, on kissing complex stability has been reported in the RNA I-RNA II looploop interaction (6). As observed for this complex, the in vitro selection against TAR RNA identified 5Ј-purine-pyrimidine and 5Ј-pyrimidine-purine base pair preference at the first and second positions of the aptamer stem, next to the loop, respectively, which might indicate that this could be crucial for stability. The results obtained with the truncated R-06 16 (UA) aptamer variant suggest that sequence variations further down the aptamer stem has to be considered too. As for TAR*, the rate constants with CA or CU mutants could not be determined, but in this case only the loop closing pair was modified. Thus, both the sequence of the stem close to the loop and the closing pair can be kinetically critical for fast interaction in the time scale of the SPR experiments.
The stability of TAR-aptamer complexes decreases in the order GA ϭ AG Ͼ GG Ͼ GU Ͼ AA Ͼ GC Ͼ UA ϾϾ CA, CU. Roughly the same ranking was observed for hexanucleotide hairpin loops (39) and internal mismatches in RNA (40). Purine stacking interactions at the loop-loop helix/stem junction have been reported for the complex between the two RNA transcripts of the ColE1 plasmid (38). Such interactions might also stabilize the TAR RNA-aptamer complex as, except for the GU aptamer mutant, the PuPu mutations are the less destabilizing ones. Interestingly, the stability of the complexes analyzed with SPR and with UV-absorption spectroscopy follows the one found with EMSA experiments (25). Thermal denaturation experiments monitor the equilibrium in solution, whereas SPR (one of the partners is immobilized on a surface) or EMSA (the molecules migrate in a three-dimensional network) do not. Despite these fundamental technical differences, the results demonstrate that the three techniques are following the same molecular event. Even if the absolute values of the equilibrium binding constant for the complex with the aptamer or with TAR* show a discrepancy, the linear correlation between the binding equilibrium constant determined by SPR and the T m (Fig. 8) validates non-equilibrium techniques and supports our conclusions.
Specific binding of monovalent and particularly divalent cations such as magnesium ions to RNAs is often required to proceed with the biological processes in which these RNAs are involved (41). These cations are trapped locally and interact either directly or through water molecules (42,43). The stabilizing role of magnesium ion on RNA kissing complexes is known (5,6). Strong evidence of direct binding of this cation to the dimerization initiation site of HIV-1 has been reported recently (8). One magnesium would bind at the center of the pocket in the sharp turn that each loop makes. Similar binding was proposed for the ColE1 plasmid (38). The analysis of the three-dimensional structure of TAR-TAR* complex suggests that the two phosphate clusters flanking the major groove of the loop-loop interaction helix may constitute part of two specific metal ion binding sites (37). The number of magnesium ions that bind upon formation of TAR-TAR*, deduced from the ion concentration dependence of the T m agrees well with this hypothesis. In vitro selection identified high affinity RNA ligands for TAR RNA at physiological magnesium concentration (3 mM). According to our results, a similar number of magnesium ions seem to be directly implicated in the formation of the complex with the aptamer. Then the higher affinity of the aptamer (K d ϭ 2.0 nM) for TAR RNA over the rationally designed TAR* binder (K d ϭ 92.5 nM) does not reside in a difference in the number of magnesium ions that bind. Supported by the results obtained with the aptamer variants, we hypothesize that the loop closing pair of the R-06 24 (GA) aptamer is an intrinsic determinant, which explains the higher stability of the TAR-aptamer complex at low magnesium concentration.
Does the loop closing pair of the R-06 24 (GA) aptamer form a non-Watson-Crick base pair, as is frequently observed in other hairpin loops (44)? Of all PuPu combinations of the aptamer closing pair, the AA mutation induced the largest decrease in the stability of the complex, whereas the AG inversion has a minor effect. AA should be exchangeable with AG if a sheared pair is formed (45). Indeed, these two noncanonical base pairs are isosteric. Then the stability of the resulting complexes with TAR RNA should be close. This is not the case, and AG is actually equivalent to GA in terms of stability, likely indicating that the loop closing GA pair is not a sheared pair.
Could a closing GA pair be preferred to increase loop-loop complex stability? Recently, an in vitro selection against yeast tRNA Phe identified an aptamer that folds as a hairpin (46). The seven central bases of the 9-nt loop of this RNA are complementary to the entire tRNA Phe anticodon loop and are also closed by a GA pair. Together with our results, this work suggests that, when all positions are randomized, a GA pair is preferred to close a loop. Thus, this extra pair would be thermodynamically favorable for loop-loop interactions. To check this hypothesis, the Watson-Crick GC pair of RNA IЈ, next to the loop, was replaced by a GA pair. Despite the lower stability that this new hairpin displays (⌬T m ϭ Ϫ5.8°C), the resulting loop-loop complex with RNA IIЈ shows a clear transition at about 24°C, which is not observed with the unmodified RNA IЈ hairpin. The extra-GA pair is clearly favorable even with the stem shortened by 1 base pair that results from this mutation.
In conclusion we showed that the GA pair that closes the loop of the in vitro unidentified R-06 24 (GA) aptamer is critical for the stability of the complex with TAR RNA. Whether the GA pair is a noncanonical pair is not established. The data cannot give direct evidence of that. Clearly the GA pair is a structural determinant, which is fundamental to explain the higher stability of the TAR RNA-aptamer complex at physiological concentration of magnesium ion over any other one, including the complex with the rationally designed TAR* ligand. The role that the aptamer stem might also play is presently under investigation. This validates the usefulness of an in vitro combinatorial approach over a rational one to identify high affinity RNA ligands. The GA pair favors the stability of the complex once formed as the binding equilibrium constant is controlled by the off-rate rather than by the on-rate of the complex formation. Finally, GA pair could be preferred when structural distortions that might increase stability of loop-loop RNA complexes are required. FIG. 8. Correlation between the equilibrium binding constant, K d , and T m . K d for the complexes between the immobilized full-length TAR RNA and the aptamers was deduced from kinetic experiments using the BIAcore. T m was deduced from thermal denaturation experiments, in solution, for the complexes between miniTAR and the aptamers.