ABSTRACT
The three-dimensional structure of the hairpin formed by d(ATCCTA-GTTA-TAGGAT) has been determined by means of two-dimensional NMR studies, distance geometry and molecular dynamics calculations. The first and the last residues of the tetraloop of this hairpin form a sheared G-A base pair on top of the six Watson-Crick base pairs in the stem. The glycosidic torsion angles of the guanine and adenine residues in the G-A base pair reside in the anti and high-anti domain (~-60o) respectively. Several dihedral angles in the loop adopt non-standard values to accommodate this base pair. The first and second residue in the loop are stacked in a more or less normal helical fashion; the fourth loop residue also stacks upon the stem, while the third residue is directed away from the loop region. The loop structure can be classified as a so-called type-I loop, in which the bases at the 5'-end of the loop stack in a continuous fashion. In this situation, loop stability is unlikely to depend heavily on the nature of the unpaired bases in the loop. Moreover, the present study indicates that the influence of the polarity of a closing A[middot]T pair is much less significant than that of a closing C[middot]G base pair.
For many years considerable effort has been put into conformational studies aimed at establishing the structural principles that underly DNA and RNA hairpin loop stability (reviewed in 1 ). From these investigations, which were mainly focused on hairpins with tetraloops, it has emerged that stable loops very often possess a base pair between the first and last loop residue. The amount of stabilization seems to depend on the exact nature of this base pair, which can be either canonical (C[middot]G) or non-canonical. Recently, a particular group, i.e. loops with a G-A base pair, has gained much interest (2 -9 ). Sequences which have the potential to form such a loop are generally remarkably stable and, moreover, fairly common. Tetraloops with the consensus sequence GNRA (in which N is any nucleotide and R is purine), for example, are abundant motifs in rRNA (5 ,10 ). DNA hairpins comprising G-A base pairs in the loop have been related to telomeric and centromeric structures (11 ,12 ).
An important part of the current knowledge on G-A base pairs has resulted from studies performed on (partly) self-complementary oligonucleotide strands, which form duplex structures with a tandem G-A mismatch (13 -18 ). In particular, it has been found that structures with two successive, so-called sheared G-A base pairs formed by a 5'-GA-3'/3'-AG-5' sequence are exceptionally stable. This stability has primarily been attributed to very favourable inter-strand GG and AA stacking interactions (15 ). According to heuristic rules, postulated on the basis of these studies, the formation of this stable structural unit is promoted if the guanine residue is preceded by a pyrimidine residue (19 ,20 ).
Remarkably, conformational studies on hairpins with the potential to form a G-A base pair in the loop have thus far revealed, in all instances, the existence of the same sheared G-A base pair configuration (4 ,5 ,7 -9 ). Based on the results of these investigations it has therefore been suggested that favourable stacking interactions between the G-A and the underlying base pair in the stem, preferably C[middot]G (8 ,21 ), contribute to the stability of these hairpins. In addition, RNA GNRA loops seem to be further stabilized by the formation of various hydrogen bonds involving phosphate oxygen and hydroxyl oxygen atoms (5 ,7 ,22 ).
The influence of the unpaired bases in the loop has thus far been a matter of debate. Experiments by Miura and co-workers (2 ,3 ) indicate that this base(s) does play a substantial role. For instance, they observed melting temperatures for hairpins formed by d(GC-GAAA-CG) and d(GC-GAA-CG) which are ~7oC higher than that of d(GC-GTTA-GC). NMR studies on the GAA loop support this view. Extensive stacking interactions were found between the base of the second loop residue and the G-A base pair formed by the first and third loop residues (4 ). Interesting results have also been obtained by Reid and co-workers, who report completely different conformations for d(CAAT-GCA-ATG) and d(GAAT-GGA-ATG). The first sequence forms a hairpin with a sheared G-A base pair within a three-membered loop, whereas the second forms a bimolecular duplex, involving a (GGA)2 motif (9 ). At the same time, this precludes definitive conclusions about a possible role of the unpaired bases, because the hairpin has to compete with the dimeric form. Others, including our group, have not observed such a large effect of the nature of unpaired bases in the loop. Zuiderweg and collaborators compared the melting temperatures of a variety of hairpins containing the sequence d(C-GNNA-G) and found a negligible influence of the N bases on stability (8 ). Thermodynamic studies performed in our laboratory on d(ATCCTA-GAAA-TAGGAT) and d(ATCCTA-GTTA-TAGGAT) revealed that these hairpins are about equally stable, with melting temperatures of 53 and 55oC respectively (23 ).
Here we present the results of a detailed determination of the loop structure of d(ATCCTA-GTTA-TAGGAT) by means of NMR. This DNA fragment, which was designed to form a hairpin with a G-A base pair in the four-membered loop, differs from that in an earlier NMR study, d(TGC-GGCA-GCA) (8 ), in two essential aspects. First, in the hairpin studied in this paper the first unpaired residue in the loop is a pyrimidine (T) instead of a purine (G) residue and, second, the stem is closed by a purine[middot]pyrimidine (A[middot]T) instead of a pyrimidine[middot]purine (C[middot]G) base pair. The present study shows that the loop structure can be classified as a so-called type-I loop (24 ), in which the bases at the 5'-end of the loop stack in a more or less continuous fashion. In this situation, the stability of the loop is unlikely to depend heavily on the nature of the unpaired bases in the loop. In addition, the present study indicates that the polarity of a closing A[middot]T base pair influences the loop stability much less significantly than does a C[middot]G pair.
The oligonucleotide d(ATCCTA-GTTA-TAGGAT) was synthesized by the phosphotriester method (25 ). NMR samples were prepared by dissolving three times lyophilized oligonucleotide to a concentration of 3 mM in a D2O solution containing 25 mM sodium phosphate, 1 mM sodium cacodylate and 0.1 mM EDTA, pH 6.8. NaCl was added to a total sodium concentration of 200 mM. A similar sample in a 95% H2O/5% D2O solution was prepared. Parameters and conditions for the NMR experiments, recorded on Bruker AM-400 and AM-600 spectrometers interfaced to ASPECT 3000 computers, are summarized in Table 1 . Data processing was carried out using either UX-NMR software running on a Bruker-X32 computer or the M-NMR software package running on a Silicon Graphics Indigo2tm work station.
Volumes of NOE cross-peaks of the NOESY spectra with mixing times of 75, 100, 150 and 300 ms were integrated using UX-NMR and served as input for the NO2DI program (26 ), running on a CONVEX-C120 computer. The NO2DI algorithm calculates inter-proton distances from NOE intensities, taking spin diffusion into account. The calculations were performed with the rotation correlation time set to 2.0 ns, as estimated with the aid of the Stokes-Einstein relation (27 ). The NOEs belonging to the H5-H6 proton pairs of cytosines were used for calibration. Error bounds of -10 and +20% of the calculated distances were assumed to produce lower and upper distance bounds respectively.
JH1'H2' and JH1'H2'' coupling constants were derived by simulation of the one-dimensional proton spectrum, using PANIC (Bruker). Sums of J-couplings, [Sigma]2' and [Sigma]2'', were estimated from multiplet patterns in a DQF-COSY spectrum. For J1'2' and J1'2'', errors of 0.3 Hz were assumed, based on the digital resolution of the NMR spectrum. For [Sigma]2' and [Sigma]2'', the errors were estimated as 5.0 Hz for residues C3 and G13, 3.0 Hz for T2, G7, A12 and G14 and 1.5 Hz for the remaining residues. The allowed pseudorotation parameters, P and [Phi]m, of the sugars were derived from these J-couplings using the program MARC (28 ,29 ). PS was sampled between 117 and 211.5o in steps of 4.5o, [Phi]mS was sampled between 32 and 42o in steps of 2o and the molar fraction fS was sampled between 0.5 and 1.0 in steps of 0.005. PN and [Phi]mN were kept fixed at 9 and 35o respectively.
Ranges allowed for the backbone dihedral angles [beta], [gamma] and [epsilon] as well as the glycosidic torsion angle [chi] were derived from NOE data combined with homo- and heteronuclear J-coupling constants (vide infra).
Table 2
Table 3 . Chemical shifts of d(ATCCTA-GTTA-TAGGAT) at 298Ka
Table 4 J-coupling constants and sugar pucker parameters of d(ATCCTA- GTTA-TAGGAT)
The thus derived distance and dihedral angle ranges served as input for the variable target function algorithm DIANA, version 2.8 (30 ,31 ), adapted in our laboratory to incorporate variable sugar puckers. Calculations were performed on a Silicon Graphics Indigo2tm workstation. For the stem of the hairpin standard B-DNA distance restraints, hydrogen bond restraints and dihedral angle ranges (29 ,32 ,33 ) were used, as well as restraints to preserve base pair planarity.
The DIANA structures with the lowest target function were used as input for the program CORMA, version 2.25 (34 ), which calculates a dipole-dipole relaxation matrix for a system of protons and converts that into intensities expected for a NOESY experiment. A number of cycles of DIANA and CORMA calculations were performed in the same way as described before (24 ,35 ), to obtain a model for the hairpin with a minimum number of violations of the NMR data.
The final DIANA input included 72 upper and lower distance restraints. An additional set of 65 lower bound distance restraints of 4.0 Å was added for the loop region. They followed from the first cycles of DIANA calculations, which generated structures with short inter-proton distances for which no corresponding experimental NOE cross-peak was observed while at the same time other NOEs involving the same protons ruled out local motion. A total of 50 lower and upper distance restraints were used to preserve the base pairs, including four lower and upper distance restraints for the H bonds of a G-A base pair. Furthermore, 98 dihedral restraints in the stem and stereospecific assignments for 25 out of 32 CH2 groups (H2'/H2'' or H5'/H5'') were used in the calculations.
Two separate DIANA calculations, representing two distinct, possible ranges for the [beta] dihedral angle of residue 9 (vide infra), were performed, each producing a total of 200 structures. For each conformation the target function was minimized at the so-called levels 0-15 (see 31 ), whereby minimization at level 15 was done three times. The weighting factors for explicit upper and lower distance limits were set to 1.0 and the weighting factor for dihedral angle restraints was set to 5.0. For levels 0-14 and for the first minimization step at level 15 the weighting factor for the van der Waals restraints was 0.2, for the second minimization step at level 15 it was increased to 0.5 and for the final minimization step at level 15 to 1.0. For minimization levels 0-14 the maximal number of target function calculations was set to 100; for level 15 it was set to 200.
The 15 structures with smallest target function values (<3.0 Å2) of each set were used as input for X-PLOR, version 3.1 (36 ), which was run on a Silicon Graphics Indigo2tm workstation. They were subjected to simulated annealing and refinement protocols, not substantially different from those delivered with the X-PLOR software, of which the details are given in Table 2 . For each input structure five new structures were generated by random variation of the initial velocities, resulting in 75 structures in total. As in the DIANA calculations, two possible ranges for [beta](9) were distinguished. The force field used for these calculations consisted of geometric terms accounting for bond stretching, bond angle bending and maintaining planar ring structures and (pro)chiral centres. The standard parameter and topology files were supplemented by `improper' terms to maintain prochirality around the C5' atoms. Performing the calculations in vacuo, a soft repulsive term was used for the non-bonded interactions. The NMR-derived distance and dihedral restraints were added as (soft) square well potentials. The base pairs of the stem were kept flat using weak planarity restraints. After these calculations a structure was accepted as a reasonable conformation if the sum of energies resulting from NMR-derived distance and dihedral restraints did not exceed 75 kcal/mol.
As a last step, the accepted structures were subjected to restrained energy minimization in a force field including terms for electrostatic and dihedral contributions. The effect of atomic charges was reduced by the application of partial charges and by choosing a distance-dependent dielectric constant, [epsilon] = [epsilon]0[middot]r. During this stage, the soft repulsive non-bonded energy term was replaced by a Lennard-Jones potential; the distance restraints used earlier to preserve the base pairs of the stem and the H bonds of the G-A base pair were omitted, as were the planarity restraints.
Several specific interaction energy contributions were calculated using the Vpert function in X-PLOR. Quanta (Polygen), running on a Silicon Graphics Indigo2tm workstation, was used to visualize the results of the structure calculations. Helical parameters were calculated with the program NEWHELIX93 (kindly provided by R.E.Dickerson).
The imino proton spectrum of the hairpin with GTTA loop, recorded at several temperatures, is shown in Figure 1 . The assignments of the imino proton and amino proton resonances are summarized in Table 3 . They were derived from NOESY spectra recorded for the sample dissolved in H2O (not shown). Between 12.0 and 15.0 p.p.m. the resonances originating from base paired residues are detected and account for all base pairs in the stem. The integrated peak intensity in the spectrum near 10.9 p.p.m. accounts for three resonances. The imino proton resonances of the unpaired T8 and T9 are observed at this position and the imino proton of G7 accounts for the intensity of the third resonance. The resonance position and temperature-dependent behaviour of the imino proton of G7 indicate that this proton is not involved in hydrogen bonding.
Most sequential NOEs expected for helical regions, e.g. between H6/H8 and H1' resonances (Fig. 2 a) and between H6/H8 and H2'/H2'' resonances, were also visible for the GTTA loop region. Although the H1'(T8)-H6(T9), H2'/H2''(T8)-H6(T9) and H2''(T9)-H8(A10) cross-peak intensities are unusually weak, only the H1'(T9)-H8(A10) and H2'(T9)-H8(A10) NOEs could not be observed. Figure 2 b summarizes these results.
In addition to the standard sequential sugar-base connectivities, the NOESY spectra showed a large number of unusual contacts between the various loop residues. Figure 4 schematizes the full set of inter-residue NOEs observed for the GTTA loop region. The connectivities found for the A6-G7, G7-T8 and A10-T11 steps indicate that the first and bases of the loop, G7 and A10 respectively, are stacked upon the terminal base pair of the stem region, A6-T11, and that residue T8 is more or less stacked upon residue G7. The lack of intense NOEs to T9 suggests that this residue must be directed away from the loop.
A J-coupling analysis of the resonances of the sugar moieties was performed and the results were used to determine the sugar conformations. The results are listed in Table 4 . The sugar rings are in an S-type conformation, although for almost all residues N-type conformations are admixed to some extent. It should be noted that the J-coupling information for residues A1, T8, A10, G14, A15 and T16 is rather limited. The sugar conformations of these residues are consequently not well defined (see Table 4 ). A qualitative analysis of H3'-H4' cross-peak intensities in the TOCSY spectrum, which depend on the values of the JH3'H4'' coupling constants, corroborated the results. For A6, G7, A10, T11 and A12 these cross-peaks are either absent or show very low intensities. This suggests that the sugar rings of these residues are present almost exclusively in the S-type form. For the remaining residues, among which are included T8 and T9, the TOCSY data indicate some conformational averaging.
The 1H-31P heterocorrelation spectrum (Fig. 3 ) provided information on the [beta] angles (P-O5'-C5'-C4'). The cross-peaks observed for P(G7), i.e. H3'(A6)-P(G7), H4'(G7)-P(G7) and H5'(G7)-P(G7), imply that the [beta] angle of G7 is in the trans conformation, ~150o. P(T8) shows cross-peaks to H3'(G7), H4'(T8) and H5''(T8), as in standard B-type DNA, showing that [beta](8) adopts a value of ~210o. For P(T9) the cross-peaks to H5'(T9) and H5''(T9) are both present, while the cross-peak to H4'(T9) is absent. From cross-sections through the multiplets it was estimated that JH5'P as well as JH5''P are ~7-10 Hz. This indicates that [beta] for T9 adopts a value of ~110 or ~250o, dependent upon the stereospecific resonance assignment of H5'/H5'' of this residue. P(A10) shows cross-peaks to H3'(T9) and H5''(A10). It was estimated from the cross-section through the A10(H5'') multiplet that JH5''P is ~7-10 Hz. The standard cross-peak H4'(A10)-P(A10) is absent, which implies that [beta] may still adopt a value in the trans region of ~210o, but [gamma] must deviate from the standard gauche(+) value (vide infra).
Analysis of the intra-residue NOE cross-peaks showed that the [chi] angles of all residues fall in the anti domain normally found in B-DNA, except for residue A10. A high intensity of the H8-H1' cross-peak was observed, together with H8-H2' and H8-H2'' cross-peaks of the same intensity as the corresponding cross-peaks of other residues. Earlier this observation suggested to us that A10 is in the syn orientation (23 ). A more detailed analysis provides a better definition of this angle, i.e. integration of the NOE volumes of the intra-residue cross-peaks, subsequent distance calculation using the NO2DI program (26 ) and consideration of the P-[chi] correlation plots for H8-H1', H8-H2' and H8-H2'' proton pairs (29 ) shows that [chi] of residue A10 adopts a value between 0 and -60o.
Additional evidence for this orientation of A10 comes from the chemical shift position observed for the H2' sugar proton resonance of this residue (2.67 p.p.m.). The ring current effect experienced by this proton is very sensitive to the value of the glycosidic torsion angle. As a consequence, [chi] in the syn domain would have resulted in a substantial downfield shift of the H2' resonance, which is definitely not the case (Wijmenga et al., manuscript in preparation).
The distance restraints together with the allowed dihedral angle domains and sugar pseudorotation parameters served as input for the distance geometry program DIANA. These data were supplemented by hydrogen bond restraints for the formation of a G-A base pair, the formation of which was suggested by thermodynamic studies performed earlier (23 ). Basically, a number of G-A pair configurations can possibly exist (see, for instance, 13 ,37 ,38 ), but in a first consideration only those in which both glycosidic torsion angles reside inside the anti domain qualify. We tried all of the possible candidates as starting configurations in the structure calculations but only that with hydrogen bonds between the N3 atom of G7 and the amino group of A10 and between the amino group of G7 and the N7 atom of A10 did not lead to severe violations of the NMR constraints. On the contrary. Although we did not have direct NOE evidence for hydrogen bond formation between the G7 and A10 amino protons, the latter base pair fitted the NMR restraints excellently. It should be mentioned in passing that the imino proton of G7 is not involved in base pairing, in accordance with its resonance position and temperature-dependent behaviour (see Fig. 1 ).
The subsequent structure refinement calculations within X-PLOR mainly affected locally unfavourable conformations. Structures with high energies after these calculations were discarded from further analysis. The procedure resulted in a set of 24 refined structures, of which the average root mean square (r.m.s.) deviation for the loop region was 0.33 Å with respect to the mean structure. Figure 5 shows an overlay of the 10 structures, taken from this set, with lowest energies resulting from NMR- derived distance and dihedral restraints. The set comprises structures with both a [beta](9) angle of ~110o as well as ~250o (vide supra). However, energy minimization of these refined structures, in a force field which included terms for electrostatic and dihedral contributions, revealed that the first situation is energetically more favourable, which indicates that it is more likely to occur.
The complementary ends of the d(ATCCTA-GTTA-TAGGAT) fragment form the six base pairs of a B-type stem, connected by a four-membered loop, which changes the direction of the phosphate backbone by 180o (Fig. 6 ). As in all DNA tetraloops studied previously, this change in backbone direction is not spread throughout the entire loop region but takes place within a relatively small portion of the loop, i.e. between the third and fourth loop residues. Up to this point, the backbone essentially follows the general course of the 5'-strand of the underlying helix, although it tends to shift a little in the direction of the helix axis. After the turn the course of the opposite strand is followed.
The GTTA loop exhibits a compact structure, dominated by stacking interactions. Figure 7 a shows that the first base in the loop, belonging to G7, stacks quite extensively on the base of the last stem residue, A6. Compared with the situation encountered for normal helix propagation, the base of G7 is slightly shifted towards the major groove and its sugar moiety lies more inwards with respect to the helix cylinder. The base of residue A10 also stacks on the stem, i.e. on the base of T11, but not at all in a regular fashion. It is drastically shifted towards the minor groove, partly accomplished by the glycosidic torsion angle of this residue being ~-70o. In this way the N6/N7 flank of A10 is in close proximity to the N2/N3 flank of G7, thus enabling the formation of two hydrogen bonds between these residues. The stacking interaction between A10 and T11 is rather poor.
To accommodate the sheared G-A base pair, the value for the helical twist for this base pair is substantially lower than those for the stem, i.e. 13.0o for the G-A base pair versus 36.1o on average for the base pairs in the stem. The small helical twist brings the base of A10 directly over H1' of T11, resulting in the unusual upfield resonance position of this proton (see Table 3 ). A similar effect has been observed in various other compounds comprising one or more sheared G-A base pairs (5 ,7 ,16 ,17 ). Furthermore, we observed values for the tip, inclination and propeller twist for the G-A base pair (14.9, 5.0 and 9.6o respectively) higher than the average values for the stem (0.0, 1.5 and -4.3o respectively).
From a detailed comparison of the accumulated structural data (1 ) it follows that stable hairpin loops tend to fold into a limited number of well-defined, compact conformations. For tetraloops, three distinct types of folding, designated type-I, type-II and type-III folds (24 ), have been observed. Type-I loops, observed exclusively for DNA, adopt a conformation with the first three bases at the 5'-end of the loop forming a more or less continuous stack on the 3'-end of the stem. In type-II loops, found in DNA as well as in RNA, the base of the second loop residue from the 5'-end of the loop is turned into or towards the minor groove and the base of the third residue lies over the base pair formed by residues 1 and 4. The third fold, type-III, is only observed in RNA and is described by a continuous stacking of the loop bases on the 5'-end of the stem.
Of the unpaired loop residues in the GTTA loop, T8 stacks upon its 5'-neighbour (Fig. 7 b), something which is not entirely true for T9. The latter is directed away from the loop and points into the solvent (Fig. 7 c). Nevertheless, the GTTA loop structure established here can be considered as belonging to the type-I category of loop folding. Hence, the results of this study conform to the idea of the existence of a small number of folding patterns for four-membered loops.
For tetraloops containing a Watson-Crick base pair (miniloops) it has been found that the position of the first unpaired base may depend upon its nature. It has been proposed that if this residue is a purine, it tends to stack upon its 5'-neighbour, whereas a pyrimidine residue preferentially folds into the minor groove (39 ). This rule also holds for tetraloops comprising various other types of base pairs (24 ,40 ,41 ). Therefore, one might have expected that the second residue in the GTTA loop would be positioned in the minor groove. The fact that this is not the case can easily be understood by considering the geometry of the sheared G-A base pair. In this arrangement the adenine base has shifted towards the minor groove (see Fig. 7 ), which prevents T8 from folding into this groove. Hence, the results presented here suggest that the overall folding of a loop containing a G-A base pair is not determined by the nature of the second base. A similar stacking pattern might be expected therefore for loops with a purine as the second loop residue. The comparable melting temperatures for d(ATCCTA-GTTA-TAGGAT) and d(ATCCTA-GAAA-TAGGAT), i.e. 55 and 53oC respectively (23 ), as well as the results of Sandusky et al. (8 ), corroborate this.
A number of conformational studies have been performed in which a sheared G-A base pair was found in a four-membered loop. The majority of these investigations concerned loops meeting the consensus sequence GNRA, a motif which is often denoted as unusually stable. However, the actual stability of these loops varies strongly and the question might be raised as to whether all of them really contain a sheared G-A base pair.
Recently, Sandusky et al. (8 ) observed that the stability of GNNA tetraloops strongly depends on the polarity of the C[middot]G base pair lying underneath, a phenomenon also known to occur for other hairpin loops (1 ). The melting temperature of a loop composed of C-GGCA-G was found to be almost 20oC higher than that of a G-GGCA-C sequence and it was concluded that the presence of a sheared G-A pair in the second case was improbable. The authors explained this observation by predicting very poor stacking interactions for the hypothetical situation of a sheared G-A base pair positioned on top of a G[middot]C base pair, especially between the two consecutive purines (GG). These findings might suggest that the rule formulated by Reid and co-workers, i.e. a sheared G-A base pair configuration is only formed when the preceding base pair is C[middot]G or T[middot]A (19 ), also applies to G-A base pairs in loops. However, the results presented here show otherwise. We find a well-defined loop structure, including a sheared G-A base pair, on top of an A[middot]T moiety. The stacking interactions between these two base pairs are definitely much better than those predicted for the 5'-GG-3'/3'-CA-5' unit (8 ). In particular, the A6-G7 stacking encountered for the A-GTTA-T loop seems to be of the same order as the adenine-guanine interaction in the structure adopted by the 5'-CG-3'/3'-GA-5' unit. Similar results as observed here were found earlier for an RNA GAGA tetraloop positioned on top of an A[middot]U base pair (7 ). Unfortunately, we are not aware of thermodynamic or structural data for loops composed of a T(U)-GNNA-A sequence. Still, the structure of the T-GCA-A loop (9 ) indicates that favourable stacking interactions are possible for such a situation.
In conclusion, a picture emerges that the stability of a loop containing a sheared G-A base pair is relatively insensitive to the polarity of an underlying A[middot]T(U) base pair, contrary to the situation in which this is a C[middot]G base pair. This effect is possibly related to the direction of the dipole moment of the guanine base, which is more or less opposite relative to those of thymine, adenine and cytosine (42 ). This might result in a more favourable situation for CG, AG and TG stacks, as compared with a GG stack, and therefore might lead to more favourable stacking interactions in 5'-CG-3'/3'-GA-5', 5'-AG-3'/3'-TA-5' and 5'-TG-3'/3'-AA-5' units and less favourable interactions for 5'-GG-3'/3'-CA-5'.
NMR spectra were recorded at the Dutch National hf-NMR Facility (Nijmegen, The Netherlands), supported by SON. We wish to thank J.Joordens for excellent technical assistance.
*To whom correspondence should be addressed. Tel: +31 243 652 678; Fax: +31 243 652 112; Email: ceesh@sci.kun.nl
Present addresses: +Unilever Research Laboratories, Vlaardingen, The Netherlands and [sect]Department of Medical Biochemistry and Biophysics, University of Umeå, Sweden
Table 1
aFor two-dimensional experiments phase-sensitive detection of the indirectly observed frequency was achieved by the TPPI method (45).
bIf two numbers are given, the first one refers to the direct detected dimension (f2), while the second number refers to the indirect dimension (f1).
REFERENCES
Return