Skip Navigation

This Article
Right arrow Abstract Freely available
Right arrow Print PDF (1131K) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (18)
Right arrowRequest Permissions
Right arrow Commercial Re-use Guidelines
for Open Access NAR Content
Google Scholar
Right arrow Articles by Asensio, J. L.
Right arrow Articles by Lane, A. N.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Asensio, J. L.
Right arrow Articles by Lane, A. N.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Nucleic Acids Research Pages 3677-3686  

Comparison of the solution structures of intramolecular DNA triple helices containing adjacent and non-adjacent CG·C+ triplets
Introduction
Materials And Methods
   Materials
   Methods
Results
Discussion
Acknowledgements
References

Comparison of the solution structures of intramolecular DNA triple helices containing adjacent and non-adjacent CG·C<sup>+</sup> triplets

Comparison of the solution structures of intramolecular DNA triple helices containing adjacent and non-adjacent CG·C+ triplets

Juan Luis Asensio+, Tom Brown1 and Andrew N. Lane*

Division of Molecular Structure, National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 1AA, UK and 1Department of Chemistry, University of Southampton, Highfield, Southampton SO17 1BJ, UK

Received May 14, 1998; Revised and Accepted July 2, 1998

ABSTRACT

The solution conformations of the intramolecular triple helices d(AGAAGA-X-TCTTCT-X-TC+TTC+T) and d(AAGGAA-X-TTCCTT-X-TTC+C+TT) (X = non-nucleotide linker) have been determined by NMR. 1H NMR spectra in H2O showed that the third strand cytosine residues are fully paired with the guanine residues, each using two Hoogsteen hydrogen bonds. Determination of the 13C chemical shifts of the cytosine C6 and C5 and their one-bond coupling constants (1JCH) conclusively showed that the Hoogsteen cytosine residues are protonated at N3. The global conformations of the two molecules determined with >19 restraints per residue are very similar (RMSD = 0.96 Å). However, some differences in local conformation and dynamics were observed for the central two base triplets of the two molecules. The C N3H were less labile in adjacent CG·C+ triplets than in non-adjacent ones, indicating that the adjacent charge does not kinetically destabilize these triplets. The sugar conformations of the two adjacent cytosine residues were different and the 5[prime]-residue was atypical of protonated cytosine. Hence, there are subtle effects of the interaction between two adjacent cytosine residues. The central two purines in each sequence showed non-standard backbone conformations, averaging between [gamma] [ap] 60° and [gamma] [ap] 180°. This may be related to the difference in the dependence of the thermodynamic stability on pH observed for these two sequences.

INTRODUCTION

The thermodynamic stability of parallel DNA triple helices depends on the content of CG·C+ triplets, pH and ionic strength (1-3). For triplexes that consist entirely of TA·T triplets, the stability is independent of pH in the range 5-9 (3), but it is greatly enhanced by increasing ionic strength (4) and especially by divalent metal ions such as magnesium (5). This is presumably by electrostatic interaction with the negative charge on the phosphate groups. However, as the content of CG·C triplets in the molecule is increased, the thermodynamic stability becomes increasingly pH dependent and below pH ~6, it becomes much higher than that of an all TA·T triplex of the same length (3,6). The net stability at low pH increases in proportion to the number of GC·C+ triplets. In these cases, the triplexes are destabilized by increasing ionic strength (2), presumably because at low ionic strength the positive charge on the cytosines in the Hoogsteen strand is effective in neutralizing the backbone charge.

It has been reported that introducing several contiguous CG·C+ triplets is destabilizing (7), perhaps because charge repulsion between adjacent cytosines has an adverse effect on base stacking. However, we have recently shown that the thermodynamic stability of intramolecular triple helices containing two adjacent cytosine residues is very similar to ones where the cytosines are separated by one or more TA·T triplets. Furthermore, the CN3H protons were observable by NMR, showing that both cytosines were protonated in the adjacent case (3). This result is perhaps surprising, so we have looked for a structural explanation. We report here the solution structures of two triplexes, d(AGAAGA-X-TCTTCT-X-TC+TTC+T) and d(AAGGAA-X-TTCCTT-X-TTC+C+TT), where X is the di-octylphosphate linker (8).

MATERIALS AND METHODS

Materials

Intramolecular triplexes and duplexes were synthesized using the di-octylphosphate linker (X) as previously described (8): d(AGAAGA-X-TCTTCT-X-TC+TTC+T) (AGAAGA) and d(AAGGAA-X-TTCCTT-X-TTC+C+TT) (AAGGAA); underlining denotes the Watson-Crick pairs in the triplexes.

Methods

Purified samples were dissolved in 0.1 M KCl, 10 mM sodium phosphate, pH 5.0, containing 0.1 mM 2,2[prime]-(dimethyl)silapentane-5-sulphonate for NMR analysis.

1H spectra were recorded at 11.75 and 14.1 T on Varian UnityPlus and Varian Unity spectrometers respectively. Spectra in 1H2O were recorded using the gradient echo Watergate read pulse (9) at 2°C. NOESY spectra were recorded in both 1H2O and 2H2O. DQF-COSY spectra were recorded in 2H2O. Spectra in 2H2O were recorded at 30 (AAGGAA) or 35°C (AGAAGA). The spectra for the former were slightly better at 30 than at 35°C. Control experiments showed that this has only slight effects on chemical shifts and other NMR parameters. Pure absorption mode spectra were obtained using the hypercomplex method (10). 13C-1H HSQC (11) NMR spectra were recorded at 14.1 T with and without 13C decoupling to provide both the 13C assignments from the known 1H assignments and the one-bond C-H coupling constants in the bases.

All data were transferred to a Silicon Graphics Indigo 2 and processed using Felix v.95.0 (MSI, San Diego, CA). The free induction decays were apodized with a shifted sine squared function in t2 and t1 and zero filled in both dimensions to give a final 8192 × 1024 real data matrix.

Sugar puckers were analysed using sums of coupling constants ([Sigma]1[prime], [Sigma]2[prime] and [Sigma]2[prime][prime]) derived from DQF-COSY, NOESY and in some cases from 1D spectra. The scalar coupling information was supplemented where possible with the H1[prime]-H4[prime] distance determined from NOESY build-up rates and the data were analysed with the program pfit as previously described (12). Standard deviations for the coupling data were typically 0.5 Hz for [Sigma]1[prime] and 1 Hz for [Sigma]2[prime] and [Sigma]2[prime][prime]. In some instances, an estimate of [Sigma]3[prime] was obtained, which was used only qualitatively to define possible ranges of sugar conformation. Sugar conformations were systematically searched with P values from 0 to 200°, pucker amplitudes from 30 to 42° and fraction of the S state from 0.4 to 1 in steps of 4°, 3° and 0.02 respectively. Those residues for which [Sigma]1[prime] > 14.5 Hz and r(H1[prime]-H4[prime]) > 2.8 Å were loosely constrained during the refinement to the S range (110° < P < 180°).

Further information about the nucleotide conformations was obtained from measurements of [Sigma]4[prime], which were used to derive a constraint for the torsion angle [gamma] (12-14). In addition, some experimental information about the backbone angle [epsis] was obtained from the H3[prime] linewidths. If the [epsis](t) rotamer is significantly populated, a large 3JHP (>20 Hz) makes [Sigma]3[prime] >>20 Hz (14). In all cases [Sigma]3[prime] were in the 12-18 range, which rules out this possibility. These data were not used as direct constraints in the structure refinement, but as a check on the final structures.

Cross-peak volumes in the NOESY spectra were measured using Felix by integrating the footprint. Where there was significant spectral overlap, most notably in the homopyrimidine strands, the Gaussian line-fitting function was used on cross-sections through each cross-peak in both F1 and F2. The volume was taken as proportional to the product of the peak areas divided by the average of the two peak heights. This procedure was verified for resolved peaks. These volumes were then normalized to the average cross-peak volume of the cytosine H6-H5 pairs at each mixing time.

Interproton distances were derived using MARDIGRAS (15) with NOESY spectra recorded with mixing times from 50 to 300 ms. Starting models for the MARDIGRAS calculations were the A-form and B-form of both molecules. The A-form was built from an A-DNA triplex model based on fibre diffraction studies, using the Biopolymer module within InsightII (MSI, San Diego, CA). The B-form was built from the A-form using restrained molecular dynamic simulations in which all dihedral angles in the duplex part were constrained to their standard values in B-DNA. All models comprised three separate strands; the linkers, which showed few unambiguous NOEs, were not included in the calculations. Standard distance constraints for the hydrogen bonds (both Watson-Crick and Hoogsteen) were used. The glycosidic torsion angles in the third strand were constrained within the anti range of -70 to -140°.

In order to determine the interproton distance bounds for the restrained molecular dynamics refinement, the RANMARDI modification of MARDIGRAS was used (16). Thirty different intensity sets were derived from each experimental data set and MARDIGRAS calculations were performed on all of them. This procedure includes the effects of experimental noise in the relaxation matrix calculations. A noise level of one fifth the integrated intensity of the smallest cross-peak was used. The dynamic range of observed cross-peak intensities was 750.

The procedure was repeated using two different correlation times (2 and 3 ns) and two starting models (B-form and A-form) for both molecules. For every single RANMARDI calculation, the upper and lower bound for any particular distance is determined as the average value ± SD. The bounds files from all mixing times, starting models and correlation times were combined in a single file from which the final restraint file was generated. Upper bounds comprised the mean of all individual upper bounds plus the SD. Lower bounds were the mean of all individual lower bounds minus the SD. These constitute the quantitative distance restraints.

Those NOEs that could not be measured due to extreme overlap were classified by visual inspection as strong (2-2.6 Å), medium (2-3.5 Å) or weak (3-5 Å) using a short mixing time NOESY and used to create qualitative constraints. Qualitative constraints were classified in a similar way for NOEs involving exchangeable protons. Those proton pairs whose NOEs were absent in the 300 ms NOESY spectrum were constrained with a lower bound of 4.5 Å (or 5 Å when a methyl group was involved).

The distance and angle restraints obtained in this way were incorporated into a restrained molecular dynamics procedure using Discover 95.0 (Molecular Simulations Inc., San Diego, CA) with the Amber force field. The scaling factor for non-bonded interactions was 0.5. All calculations were made with no charges and a cut-off for the non-bonded interactions of 12 Å. No backbone constraints based on 31P chemical shifts were used. The geometry of the base triplets was restrained with standard hydrogen bond distances. In addition, planarity restraints with a very low weight factor of 2 kcal/mol/rad2 were applied to each triplet at all refinement steps. This favours overall planarity of the triplets if no specific distance restraints cause out-of-plane tilting of particular bases. The NOE and dihedral restraint functions used were standard square-well effective potentials.

The starting structures were first subjected to 3000 steps of restrained energy minimization. Then initial velocities were assigned with a Maxwell distribution at 800 K. After an equilibration period of 1 ps, the temperature was maintained constant for another 4 ps. In this initial stage the force constant for the hydrogen bond constraints was 25 kcal/mol/Å2, 2 kcal/mol/Å2 for the NOE constraints and 80 kcal/mol/rad2 for the dihedral constraints. The system was then slowly cooled to 200 K for 19 ps. During the cooling phase the NOE constraint force constant was gradually increased to a final value of 50 kcal/mol/Å2. The structures were then subjected to a final 3000 steps of energy minimization.

Trial runs were performed to refine the restraints. The co-ordinates of two restrained structures obtained from the A-form and from the B-form were taken as input data for the program CORMA (17). The back-calculated two-dimensional NOEs were compared with the experimental NOEs to determine how well the model structure fits the experimental data. A deviation of a factor of 2.0 between the calculated cross-peaks and the experimental values was taken as a violation. In order to minimize these differences, the distances corresponding to the violations were progressively adjusted. Since the violations in most cases were due to the use of restraints which were too loose, the adjustment required only tightening of the initial bounds.

Table 1. 1H NMR assignments of AAGGAA and AGAAGA
  H6/H8 H5/Me/H2 H1[prime] H2[prime] H2[prime][prime] H3[prime] H4[prime] NH NH2
(a) AAGGAA
A1 8.13 7.63 6.20 2.73 3.04 4.98 nd   nd
A2 7.75 7.73 6.06 2.55 2.89 5.07 4.39   7.97/7.67
G3 7.22   5.82 2.28 2.83 4.85 4.52 13.51 nd
G4 7.35   5.87 2.42 2.95 4.91 4.41 12.90 nd
A5 7.54 7.66 5.97 2.42 2.82 4.94 4.39   7.86/7.76
A6 8.01 8.04 6.22 2.51 2.52 4.90 nd   7.78
T7 7.82 1.97 6.28 2.47 2.76 4.94 4.30 14.78  
T8 7.66 1.84 6.22 2.38 2.76 4.95 nd 14.55  
C9 7.69 5.74 6.10 2.33 2.62 4.88 4.30   8.46/6.93
C10 7.66 5.63 6.01 2.22 2.63 4.81 4.25   8.59/7.29
T11 7.58 1.73 6.11 2.18 2.64 4.90 4.22 14.41  
T12 7.65 1.73 6.32 2.35 2.50 4.92 nd 14.35  
T13 7.88 2.00 6.32 2.58 2.74 4.97 4.35 14.56  
T14 7.68 1.81 6.31 2.40 2.78 4.92 nd 3.28  
C15 7.93 5.96 6.24 2.28 2.74 4.89 nd 15.95 10.06/8.97
C16 7.96 5.91 6.13 2.27 2.71 4.84 4.42 15.51 10.29/9.76
T17 7.69 1.76 6.24 2.29 2.71 4.92 2.29 13.46  
T18 7.53 1.66 6.36 2.29 2.32 4.58 4.12 12.49  
(b) AGAAGA
A1 8.02 nd 6.13 2.70 2.91 4.99 nd   nd
G2 7.50   5.95 2.50 2.96 5.06 4.35 12.86 nd
A3 7.43 7.36 5.95 2.29 2.92 4.91 4.51   7.75/7.68
A4 7.53 7.58 5.95 2.34 2.80 4.89 4.46   8.10/7.74
G5 7.41   5.83 2.53 2.81 4.93 4.39 13.24 nd
A6 7.89 8.11 6.25 2.44 2.54 4.84 4.39   7.87/7.89
T7 7.76 1.94 6.26 2.50 2.69 4.94 4.30 14.61  
C8 7.74 5.83 6.16 2.32 2.73 4.80 nd   8.68/7.52
T9 7.64 1.78 6.16 2.27 2.77 4.94 nd 14.57  
T10 7.56 1.77 6.13 2.28 2.67 4.95 nd 14.07  
C11 7.67 5.71 6.13 2.15 2.56 4.82 nd   8.40/7.26
T12 7.61 1.73 6.29 2.33 2.49 4.89 4.23 14.38  
T13 7.81 1.94 6.34 2.56 2.61 4.98 4.39 nd  
C14 8.02 6.06 6.13 2.36 2.76 4.67 nd 15.69 10.11/9.88
T15 7.85 1.83 6.36 2.45 2.78 4.94 4.39 14.02  
T16 7.60 1.72 6.16 2.31 2.68 4.88 nd 12.3  
C17 7.83 5.84 6.16 2.17 2.67 4.81 nd nd 10.26/9.27
T18 7.61 1.70 6.40 2.31 2.32 4.61 4.13 12.87  
Assignments were obtained as described in the text at 30 (AAGGAA) and 35°C (AGAAGA) for the non-exchangeable protons and 2°C for the exchangeable protons.

A final ensemble of 10 structures (five starting from A-form and five starting from B-form) was calculated for AAGGAA and AGAAGA using the refined constraint set.

All backbone torsion angles, sugar conformations and helical parameters were calculated with CURVES 5.1 (18).

RESULTS


Figure 1. NOESY spectra of AGAAGA and AAGGAA in H2O. Spectra were recorded at 14.1 T and 2°C as described in Materials and Methods. (A) AGAAGA, tm 100 ms; (B) AAGGAA, tm 100 ms.

Figure 1 shows the low field region of the NOESY spectrum in H2O of the two triple helices, showing interactions between the imino protons and the amino protons within a base triplet, NH and C2H in Watson-Crick A·T base pairs and NH of the third strand to the H8 of the purines. These latter NOEs are typical of the parallel orientation and Hoogsteen pairing of the third strand to the purine strand (8,19). The assignment of the imino and amino protons was straightforward and they are given in Table 1.

At pH 5, the N3H+ signal of the two Hoogsteen cytosine residues integrate to one proton each, showing that both cytosines are fully protonated (3). The N3H resonances are narrow in AAGGAA, but one is broader in AGAAGA. In DNA triplexes, exchange of the CN3H is determined mainly by a conformation fluctuation that allows chemical exchange to occur (20). The degree of magnetization transfer from the imino proton to water can be measured in NOESY and ROESY experiments and this provides an estimate of the exchange rate constant. The degree of exchange in AAGGAA at 10°C in 100 ms was ~50%, indicating an exchange rate constant of >5/s. The values in AGAAGA were larger, especially for C17. Because the degree of magnetization transfer in the related molecule AGAGAA was comparable with that in AGAAGA (data not shown), position alone is not the deciding factor in the exchange process. This indicates that the AAGGAA triplex is kinetically more stable than the AGAAGA triplex. The influence of exchange can also be seen in Figure 1. The NOE between the N3H and the H8 of the purine residues appears much more intense in AAGGAA than in AGAAGA, because of the exchange broadening of C17, which is also apparent for the amino protons of C17 in AGAAGA (3).


Figure 2. Possible protonation states in adjacent CG·C+ triplets. Red bases are protonated and positively charged, green bases are the canonical neutral forms.

In principle, it is possible that the protonated cytosine can exist in different tautomeric states and in a CG·C triplet, the proton could be either on the N3 of C or the N7 of G, the location being largely determined by the difference in the pK values in the triplex state (21). Thus, although the pK of free cytosine is ~4.5, it is >9 in these triplexes (3) as a consequence of the change in environment and the formation of a hydrogen bond to the N7 of G. However, the N7 of G has a pK of ~2.5 in the free nucleoside (22), so if that were perturbed in the triplex, it might become similar to that of the cytosine, in which case the proton would be shared between the C N3 and the G N7 and the charge further delocalized (Fig. 2). This arrangement might be more important for adjacent CG·C+ triplets. The 1H spectra show that the Hoogsteen cytosine residues are in an amino form (both hydrogen atoms are present on N4). The exact location of the additional proton and delocalization of the positive charge, however, is less certain, because the NOEs to neighbouring protons would be similar whether the proton is bonded to either the N3 of C or to the N7 of G. We have used 13C NMR to obtain more information about the location of the proton, as both the chemical shift and the one-bond coupling constants are affected by charge density in the ring. Figure 3 shows a 13C-1H coupled HSQC spectrum, from which the values of 1JCH can be determined with a precision of better than ±0.5 Hz. The chemical shifts and coupling constants of the base protons are given in Table 2. The Hoogsteen cytosine H6, H5 and C6 are shifted downfield and the C6 is shifted upfield compared with Watson-Crick cytosines. This is a consequence both of the ring currents and changes in the local electron density in the ring due to protonation. Changes in electron density can also be expected to change 1JCH, which is observed for the cytosine C5 (difference between Hoogsteen and Watson-Crick = 6.5 Hz) and C6 ([Delta]J = 4 Hz). The Watson-Crick cytosines also have coupling constants similar to those observed in RNA duplexes (23) and DNA duplexes (A.N.Lane, unpublished data). Comparison with the coupling constants measured for 5[prime]-CMP and 5[prime]-GMP at neutral (24) and low pH, moreover, indicates that these differences are mainly due to the positive charge on the cytosine ring (Table 2). We conclude, therefore, that both cytosines are fully protonated at N3 in AAGGAA at pH 5. As the pH is raised, the thermodynamic stability decreases, yet both cytosines retain the same degree of protonation (3). This indicates that such short triplexes are not stable unless both cytosines are protonated; significant deprotonation of the Hoogsteen cytosines only occurs above pH 9 (3).


Figure 3. HSQC spectrum of AAGGAA. The 1H coupled HSQC spectrum was recorded at 14.1 T and 30°C as described in Materials and Methods. The acquisition times in t1 and t2 were 0.05 and 0.5 s respectively. The free induction decays were zero filled once in t2 and twice in t1. 13C assignments were obtained by correlating to the proton shifts in the decoupled HSQC spectrum. (A) Correlations for the base 8, 6 and 2 positions; (B) correlations for the cytosine 5 position.

Figure 4. NOESY in D2O of AGAAGA. The spectra were recorded at 35°C, pD = 5 and 11.75 T. The NOESY spectrum was recorded with a mixing time of 300 ms. (A) The lower panel shows the base to H1[prime] NOEs, with the assignments of the purine strands. The upper panel shows the base to H2[prime]/H2[prime][prime] region. (B) DQF-COSY spectrum showing the H1[prime]-H2[prime]/H2[prime] regions. The spectrum was recorded with acquisition times of 0.05 s in t1 and 0.6 s in t2 respectively.


Figure 5. Presence of [gamma](t) for A3 and A4 in AGAAGA. NOESY and DQF-COSY spectra were recorded at 11.75 T, pD = 5, 35°C. The top left panel shows the H4[prime]-H3[prime], H4[prime]/H5[prime][prime] region of the NOESY spectrum and the top right panel shows the corresponding DQF-COSY spectrum. The bottom left panel shows the H4[prime]-H1[prime][prime] region of the NOESY spectrum.

Figure 4 shows regions of NOESY and DQF-COSY spectra of AGAAGA in D2O. The NOESY spectrum demonstrates the good resolution obtainable with these short intramolecular triplexes. The sequential NOE connectivities in the Watson-Crick strands are shown, giving the assignments of the non-exchangeable protons (Table 1). The cross-strand H8(R) to H1[prime](i+1) (YH), which are characteristic of parallel triplexes (19), are also shown boxed. The DQF-COSY cross-peaks (Fig. 4B) H1[prime]-H2[prime][prime] and H1[prime]-H2[prime] provide information about the sugar conformations. Thus, for the purines, the H1[prime]-H2[prime] cross-peak shows a + + - - fine structure, whereas the H1[prime]-H2[prime][prime] cross-peak shows a + - + - fine structure. This is characteristic of a sugar conformation in the S domain with P values between ~120 and 160°. In contrast, although it appears that all sugars show both cross-peaks to H1[prime] (ruling out a pure N conformation), some of the pyrimidines show much less fine structure (i.e. a simple + - pattern) and nearly equal intensity for the two cross-peaks. This is characteristic either of a sugar conformation with P in the range 70-100° (or near O4[prime]-endo) or of a mixed sugar conformation (8). The relatively weak H1[prime]-H4[prime] NOESY cross-peaks, however, are inconsistent with an O4[prime]-endo conformation.

Table 2. 13C chemical shifts and one-bond C-H coupling constants in AAGGAA
Resonance [delta](13C)
(p.p.m.)		 [delta](1H)
(p.p.m.)		 1JCH (Hz)
G3 8 136.7 7.22 215.0
G4 8 138.1 7.35 214.5
A5 8 140 7.54 214
C9 6 144 7.69 181
5 97.9 5.74 174
C10 6 143.6 7.66 181
5 98.0 5.63 174
C15 6 146.7 7.93 185
5 96.3 5.96 180.4
C16 6 147 7.96 185
5 96 5.91 180
5[prime]-GMP 8     215 (pH 6.5), 223.5 (pH 1.5)
5[prime]-CMP 6     185 (pH 6.5),188 (pH 2.9)
5     176 (pH 6.5),182 (pH 2.9)
13C chemical shifts were determined from HSQC spectra using indirect referencing to the 1H (34). 1JCH values were determined from coupled HSQC spectra as described in Materials and Methods.

This conclusion was confirmed by analysis of the coupling constants, which showed that the non-terminal purines on average are >90% S (Table 3). The Watson-Crick pyrimidines have a higher fraction of the N state, but are still predominantly S. The Hoogsteen nucleotides, especially the cytosines, show large contributions from the N state, as has been previously observed in these kinds of triplexes (8,25; J.L.

Asensio et al., unpublished data). There is, however, an important difference in the coupling patterns of the two molecules. For the Hoogsteen cytosines [Sigma]1[prime] is typically 11-13 Hz (8,25), which is the case for both cytosines in AGAAGA and for C16 in AAGGAA. In contrast, [Sigma]1[prime] for C15 in AAGGAA is significantly larger, ~14 Hz, showing that the sugar conformations on average are not the same for the two adjacent cytosines. This difference in behaviour of the adjacent cytosines could be a consequence of an interaction between the neighbouring charges.

Table 3. Coupling constants and sugar conformations in AGAAGA and AAGGAA
  [Sigma]1[prime] [Sigma]2[prime] [Sigma]1[prime][prime] Ps fs
AAGGAA
A1 15.6 31.2 22.8 130 0.93
A2 15.1 31 20.3 140 0.95
G3 15.1 29 19.6 155 1.0
G4 15.1 30 20 155 1.0
A5 15.1 30.6 21.5 135 0.9
A6 14.4 nd nd S 0.75
T7 14.6 nd nd S 0.77
T8 15.0 30.9 nd 120 0.85
C9 14.0 29 nd 135 0.7
C10 13.4 28 24.2 140 0.65
T11 14.0 29 nd 135 0.7
T12 14.1 nd 22 160 0.84
T13 15.1 29 nd 170 1.0
T14 13.9 nd nd S 0.67
C15 14.0 nd nd S 0.71
C16 12.6 28.5 24.4 199/130 0.68/0.52a
T17 13.7 nd nd S 0.63
T18 13.9 nd nd S 0.67
AGAAGA
A1 15.1 nd 21.9 136 0.9
G2 15.2 30.5 nd 136 0.88
A3 14.9 nd nd S 0.83
A4 14.9 nd 20 160 1.0
G5 14.9 29 20.4 172 1.0
A6 14.1 28.7 nd 156 0.78
T7 15.1 nd 20.2 152 0.92
C8 nd 28.7 nd S nd
T10 14.2 nd nd S 0.72
C11 13.7 28.7 25 128 0.64
T12 14.1 nd nd S 0.70
T13 14.6 nd nd S 0.78
C14 11.4 nd nd S 0.27
T15 13.4 28 23 155 0.73
T16 12.4 nd nd S 0.43
C17 11.7 nd nd S 0.32
T18 13.8 nd nd S 0.67
Sums of coupling constants were determined from cross-peaks in both NOESY and DQF-COSY ([Sigma]1[prime]) or DQF-COSY ([Sigma]2[prime] and [Sigma]2[prime][prime]) as described in Materials and Methods. The pseudorotation phase angle Ps and the fraction of the S state fs were determined using the program pfit (12). Where only [Sigma]1[prime] was available, the value of fs was calculated as ([Sigma]1[prime] - 9.8)/5.9 (35). The error on Ps is approximately ±20° and on fs ±0.05.
aTwo sets of values compatible with the data.

The coupling between H4[prime] and H5[prime]/H5[prime][prime] is sensitive to the orientation about the C4[prime]-C5[prime] bond. In the usual g+ ([gamma] [ap] 60°) rotamer, the two scalar couplings are both small (<3 Hz), whereas in the other two rotamers, one coupling is small (<3 Hz) and the other large (>11 Hz) (16). The H4[prime] is also scalar coupled to H3[prime]. Thus, the linewidth in the presence of unresolved couplings will be determined by the natural linewidth and the sum of all the couplings. In the g+ rotamer, the unresolved coupling, assuming a C2[prime]-endo conformation, will be ~4-6 Hz. Hence, an apparent linewidth of <10 Hz and no splitting is expected for g+. In the t and g- rotamers, the sum of the couplings should be >15 Hz. Furthermore, as one of the couplings is large (~10-12 Hz), splitting may be observed (14). In both triplexes, there are unusual chemical shifts of the H4[prime] resonances of the two central purines. The H4[prime] in the DQF-COSY (Fig. 5) shows only one cross-peak to the H5[prime]/H5[prime][prime], whereas two cross-peaks are visible in the NOESY (and ROESY, not shown) spectrum. This indicates that one coupling is large and the other small. Furthermore, in cross-sections of NOESY spectra these H4[prime] resonances appear as partially resolved doublets, with a width at half height of ~15 Hz, whereas other resolved H4[prime] resonances appear as sharp singlets (width at half height <10 Hz). Because the scalar coupling data showed that all purines are predominantly in the S conformation (Table 3 and see above), the H3[prime]-H4[prime] coupling must be small. The large width at half height for A3 and A4 and the partially resolved splitting must therefore arise from a significant H4[prime]-H5[prime]/H5[prime][prime] coupling, in agreement with the COSY spectra. Hence, both the central purines must have a substantial contribution from g- or t. Because the apparent width at half height is only ~15 Hz and the splitting is ~6 Hz, it is probable that there is an exchange between g+ and t (see below). Similar behaviour was observed for G3 and (less pronounced) for G4 in AAGGAA (not shown).


Figure 6. Structures of AGAAGA and AAGGAA. Overlay of the 10 best structures in two orthogonal views. The purine strand is shown in green, the Watson-Crick pyrimidine strand in yellow and the Hoogsteen strand in white. (Left) AGAAGA; (right) AAGGAA.

Structures were calculated as described in Materials and Methods using 19-20 restraints per residue (Table 4) and are shown in Figure 6. The definition of the molecules is relatively high: the NMR R factors show that the calculated structures give good agreement with the data. Furthermore, the R factors are lower for these structures than those calculated for initial models constructed from standard A or B structures or one in which the duplex was B with an A-like third strand. The R factors show that the final structures are closer to a B conformation than to an A conformation. The pairwise RMSD values (heavy atoms) are also small, 0.41 Å for AGAAGA and 0.63 Å for AAGGAA. This is further demonstrated by the R values in which restraints for AGAAGA were applied to the optimized AAGGAA structures and vice versa (Table 4). These R factors are slightly poorer than those calculated for the structures using their own restraints, but show that the two structures must be very similar. The RMSD values between the structures is only 0.96 Å, which is only slightly larger than the RMSD values for each individual family of structures (Table 4).

Selected torsion angles and helicoid parameters obtained for the triplexes are summarized in Table 5. As expected from the analyses of the coupling constants, the backbone torsion angle [delta] lies mainly in the range 110-120°. Where extensive averaging was expected and there were insufficient coupling constants to determine the phase angles, no restraints in the sugar conformations were used. In these instances, values of [delta] ranging from 82° to >140° were obtained (C11 and C17 of AGAAGA), which means that the experimental data did not determine their conformations. Hence, any conformation from N to S is compatible with the local structures, although the H1[prime]-H4[prime] interaction rules out a significant population in the O4[prime]-endo state. This is consistent with the ability of these residues to undergo extensive N[harr]S averaging. The average of the values of [delta], however, does not agree with the observed values of [Sigma]1[prime] (Table 2). The angle [gamma] is generally near to +60° (g+), though for residues that were not constrained, an alternative conformation was found, namely [gamma] [ap] 180° (t). In these cases, the angle [alpha] also changed from its value of ~290° for [gamma] = 60° to 130-160° for [gamma] = 180°, showing the well-known correlation of these two torsions (26). This shows that both torsions are stereochemically reasonable and do not cause significant violations of the input data. The experimental data do indicate that the central two purines in both AGAAGA and AAGGAA undergo averaging about [gamma] (see above). The glycosidic torsion angles are on average higher (approximately -130°) than in B-DNA (-110° to -120°), as previously observed (8). The parameters for the terminal bases are less well determined in general. For these short sequences, the helicoid parameters of the duplex moiety have been averaged over the central 4 bp (Table 5). On average, the parameters are similar between the two molecules; the twist is equivalent to 10.5-11 base pairs/turn, with an axial rise of ~3 Å. The base inclinations are small and positive and the helix displacement is between -2.5 and -3 Å. The propeller twist values (not shown) are substantial and negative. Thus, although the nucleotides are clearly more typical of those found in B-DNA duplexes, the helicoid parameters are intermediate between standard B and A structures. This has been observed for other parallel DNA triplexes (8,27-31). In our calculations, the axis displacement is larger than in some other structures and the average number of base pairs per turn is lower. However, as all of the triplexes solved to date are quite short and many contain lesions or mismatches (27-30), some of these parameters are quite poorly determined by the data.

Table 4. Restraints and statistics of the structure determination
  AGAAGA AAGGAA
NOESY 300 ms 180 intensities 173 intensities (tm 250 ms)
NOESY 175 ms 162 intensities 160 intensities
Quantitative distances 171 166
Qualitative distances 90 110
Lower bounds (>4.5Å) 41 42
H bonds 26 26
Total distances 328 344
[gamma] constraints 7 3
v1,v2 constraints 10 10
Total/residue 19.2 19.8
  10R1 100R1/6 10R1 100R1/6
AGAAGA 2.70 4.86 AAGGAA 2.86 5.12
B.B 4.18 7.55   4.46 8.42
B.A 5.25 9.14   5.44 10.27
A.A 8.69 16.12   8.64 16.00
AGAAGA(X) 4.05 7.13 AAGGAA(X) 4.70 7.75
Pairwise RMSD (Å) 0.41 0.63
  n > 0.2A Max n > 0.2A Max
Violations 3-5 0.24-0.28 0  
R1 = [Sigma]¦NOEc - NOEobs¦/[Sigma]¦NOEc¦; R1/6 = [Sigma]¦NOEc1/6 - NOEobs1/6¦/[Sigma]¦NOEobs1/6¦; AGAAGA(X) refers to R factors calculated with the NOE restraint file for AAGGAA. Pairwise RMSD values were calculated for 10 structures. B.B is an initial structure with a B-DNA duplex and a B-DNA third strand, B.A is a B-DNA duplex with an A-DNA third strand and A.A is an all A-DNA triplex.

Table 5. Selected torsion angles and helicoid parameters of the structures
Molecule [alpha] (°) [gamma] (°) [delta] (°) [chi] (°)
AGAAGA
R 293 ± 4 57 ± 4a 116 ± 9 -130 ± 9
Y1 284 ± 8 56 ± 2 109 ± 14 -131 ± 4
Y2 289 ± 4 58 ± 6 106 ± 13 -131 ± 5
AAGGAA
R 290 ± 4 57 ± 7a 113 ± 5 -133 ± 5
Y1 287 ± 4 59 ± 2 130 ± 4 -128 ± 5
Y2 289±4 57±3 126±9 -131±5
B 320 38 140 -98
A 310 41 79 -154
  AAGGAA AGAAGA B A
Twist (°) 34 ± 3.5 32.5 ± 1.5 36 33
Rise (Å) 3.02 ± 0.1 3.05 ± 0.13 3.4 2.6
Inclination (°) 3.4 ± 0.5 7.4 ± 1.6 -6 19
Dx (Å) -2.82 ± 0.16 -3.05 ± 0.1 -0.7 -5.4
Helicoid parameters and torsion angles were determined using Curves (18). Terminal base triplets were not included in the calculations.
aDoes not include trans conformers.

Conformational averaging certainly occurs at the nucleotide level, most notably in the sugars of the cytosines, but also in the [gamma] values of the two central purines (see above). The exchange between t and g+ is associated with a compensating change in [alpha] (and smaller changes in [beta]). This complication has not been fully accounted for in the structure calculations and the restraints file represents the dominant conformation. Where extensive averaging was observed, no constraints were supplied for the sugars. This reduces the occurrence of virtual structures such as O4[prime]-endo sugar puckers and unlikely values of [gamma] intermediate between g+ and t (i.e. [gamma] = 120°), at the expense of loss of definition. The current set of structures accounts very well for the NOE data (cf. Table 4).

The NH exchange data and the scalar coupling data for the Hoogsteen cytosines clearly show that there are differences in local conformation and dynamics in these residues. The static structures (Fig. 6) that have been calculated do not show this level of detail, which requires an ensemble averaging model to be used (32,33). This represents the next level of refinement. Nevertheless, the ensemble of structures includes some in which [gamma] = t is present, which induces a small local change in the position of the backbone.

DISCUSSION

The lack of large structural differences is consistent with the observation that the thermodynamic stability of these two duplexes is very similar at low pH (3), despite the presence of two positively charged residues next to one another. The main difference in the thermodynamic behaviour is that the pH sensitivity is slightly greater for AAGGAA than for AGAAGA or AGAGAA; the value of d(1/Tm)/dpH is 2.0 and 1.84 for the latter two and 2.34 for the former (3). As the pK values in the triplex state are >9 (3), the pH dependence of the stability at pH < 8 depends only on the enthalpy change for the dissociation and the pK values of cytosine ionization in the strand state according to the formula:

d(1/Tm)/dpH = -2.303R/[Delta]H[(2h2 + Ka1h)/(h2 + Ka1h + Ka1 Ka2) - 2]

where h = 10-pH and Ka1 and Ka2 are the ionization constants of the two cytosines. A plot of 1/Tm versus pH will be linear between pH 5 and pH 8 (3) only if both pKa1 and pKa2 are <4.5. It is possible that the pK values for the two ionizations in the strand state are quite different for adjacent cytosines than for when they are separated by 1 or 2 bases. Indeed, in the AAGGAA molecule in particular, the adjacent positive charges would be expected to depress the pKa values, rather than raise them. Hence, the slope is determined primarily by the factor R/[Delta]H. The van't Hoff enthalpy for AAGGAA is significantly smaller than for AGAAGA (3) and accounts in large part for the observed difference in pH sensitivity. This smaller enthalpy difference is consistent with poorer base stacking interactions between adjacent protonated cytosine residues compared with a protonated cytosine and a thymine. Clearly, as the overall stability is similar at low pH, there must be a compensating favourable entropic contribution for the adjacent cytosine residues. It is expected that in longer sequences of contiguous CG·C+ triplets, the triplex will become more sensitive to pH compared with, for example, an alternating CG·C+/TA·T sequence.

ACKNOWLEDGEMENTS

This work was supported by the Medical Research Council of the UK and Oswel Research Products Ltd. J.L.A. gratefullyacknowledges a Fellowship from the Spanish Ministry of Education. Accession nos: 1bcb and 1bce.

REFERENCES

1. Sun,J.-S. and Hélène,C. (1993) Curr. Opin. Struct. Biol., 3, 345-356.

2. Plum,G.E. and Breslauer,K.J. (1995) J. Mol. Biol., 248, 679-695. MEDLINE Abstract

3. Asensio,J.L., Dhesai,J., Bergqvist,S., Brown,T. and Lane,A.N. (1998) J. Mol. Biol., 275, 811-822. MEDLINE Abstract

4. Durand,M., Peloille,S., Thuong,N.T. and Maurizot,J.C (1992) Biochemistry, 31, 9197-9204. MEDLINE Abstract

5. Pilch,D.S., Levenson,C. and Shafer,R.H. (1990) Proc. Natl. Acad. Sci. USA, 87, 1942-1946. MEDLINE Abstract

6. Roberts,R.W. and Crothers,D.M. (1996) Proc. Natl. Acad. Sci. USA, 93, 4320-325. MEDLINE Abstract

7. Kiessling,L.L., Griffin,L.C. and Dervan,P.B. (1992) Biochemistry, 31, 2829-2834. MEDLINE Abstract

8. Bartley,J.P., Brown,T. and Lane,A.N. (1997) Biochemistry, 36, 14502-14511. MEDLINE Abstract

9. Piotto,M., Saudek,V. and Sklenar,V. (1992) J. Biomol. Struct., 2, 661-665.

10. States,D.J., Haberkorn,R.A. and Ruben,D.J. (1982) J. Magn. Resonance, 48, 286-292.

11. Bodenhausen,G. and Ruben,D.J. (1980) Chem. Phys. Lett., 69, 185-189.

12. Conte,M.R., Bauer,C.J. and Lane,A.N. (1996) J. Biomol. NMR, 7, 190-206. MEDLINE Abstract

13. Kim,S.-G. and Reid,B.R. (1992) Biochemistry, 31, 12103-12116. MEDLINE Abstract

14. Wijmenga,S.S., Mooren,M.N.W. and Hilbers,C.W. (1993)In Roberts,G.C.K. (ed.), NMR of Macromolecules: A Practical Approach. IRL Press, Oxford, UK, Ch. 8.

15. Borgias,B.A., Gochin,M., Kerwood,D.J. and James,T.L. (1990) Prog. NMR Spectrosc., 22, 83-100.

16. Schmitz,U. and James,T.L. (1995) Methods Enzymol., 261, 3-44. MEDLINE Abstract

17. Liu,H., Borgias,B., Kumar,A. and James,T.L. (1992) CORMA-Complete Relaxation Matrix Anaysis, Version 3.1. University of California,San Francisco, CA.

18. Lavery,R. and Sklenar,H. (1996) Curves 5.1 Manual. CNRS, Paris, France.

19. Feigon,J., Koshlap,K.M. and Smith,F.W. (1995) Methods Enzymol., 261, 225-255. MEDLINE Abstract

20. Guéron,M. and Leroy,J.-L. (1995) Methods Enzymol., 261, 383-412. MEDLINE Abstract

21. Colominas,C., Luque,F.J. and Orozco,M. (1996) J. Am. Chem. Soc., 118, 6811-6821.

22. Fasman,G.D. (ed.) (1976) Handbook of Biochemistry and Molecular Biology. CRC Press, Cleveland, OH, Vol. I, p. 206.

23. Varani,G. and Tinoco,I.,Jr (1991) J. Am. Chem. Soc., 113, 9349-9354.

24. Ippel,J.H., Wijmenga,S.S., de Jong,R., Heus,H.A., Hilbers,C.W.,de Vroom,E., van der Marel,G.A. and van Boom,J.H (1996)Magn. Resonance Chem., 34, S156-S176.

25. Macaya,R.F., Schultze,P. and Feigon,J. (1992) J. Am. Chem. Soc., 114, 781-783.

26. Cheatham,T.E. and Kollman,P.A. (1997) J. Am. Chem. Soc., 119, 4805-4825.

27. Radhakrishnan,I. and Patel,D.J. (1993) Structure, 1, 135-152. MEDLINE Abstract

28. Radhakrishnan,I. and Patel,D.J. (1994) Structure, 2, 17-32. MEDLINE Abstract

29. Ji,J., Hogan,M.E. and Gao,X. (1996) Structure, 4, 425-435. MEDLINE Abstract

30. Koshlap,K.M., Schultze,P., Brunar,H., Dervan,P.B. and Feigon,J. (1997) Biochemistry, 36, 2659-2668. MEDLINE Abstract

31. Nunn,C.M., Trent,J.O. and Neidle,S. (1997) FEBS Lett., 416, 86-89. MEDLINE Abstract

32. Lane,A.N. (1998) In Leontis,N.B. and SantaLucia,J. (eds),Molecular Modeling of Nucleic Acids, ACS Symposium Series 682. ACS, Washington, DC, pp. 106-121.

33. Gyi,J.I., Lane,A.N., Conn,G.L. and Brown,T. (1998) Biochemistry, 37, 73-80. MEDLINE Abstract

34. Live,D.H., Davis,D.G., Agosta,W.C. and Cowburn,D. (1984)J. Am. Chem. Soc., 106, 1939-1941.

35. van Wijk,J., Huckriede,B.D., Ippel,J.H. and Altona,C. (1992)Methods Enzymol., 211, 286-306. MEDLINE Abstract

*To whom correspondence should be addressed. Tel: +44 181 959 3666; Fax: +44 181 906 4477; Email: alane@nimr.mrc.ac.uk
+Present address: Instituto de Quimica Organica General, Juan de la Cierva 3, Madrid 28006, Spain

This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 30 Jul 1998
Copyright©Oxford University Press, 1998.
Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Nucleic Acids ResHome page
E. Cubero, R. Guimil-Garcia, F. J. Luque, R. Eritja, and M. Orozco
The effect of amino groups on the stability of DNA duplexes and triplexes based on purines derived from inosine
Nucleic Acids Res., June 15, 2001; 29(12): 2522 - 2534.
[Abstract] [Full Text] [PDF]

Home page
 Nucleic Acids ResHome page
O. Mauffret, C. El Amri, F. Santamaria, G. Tevanian, B. Rayner, and S. Fermandjian
A two B-Z junction containing DNA resolves into an all right-handed double-helix
Nucleic Acids Res., November 15, 2000; 28(22): 4403 - 4409.
[Abstract] [Full Text] [PDF]

Home page
 Nucleic Acids ResHome page
R. Soliva, R. Guimil Garcia, J. R. Blas, R. Eritja, J. L. Asensio, C. Gonzalez, F. J. Luque, and M. Orozco
DNA-triplex stabilizing properties of 8-aminoguanine
Nucleic Acids Res., November 15, 2000; 28(22): 4531 - 4539.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Print PDF (1131K) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (18)
Right arrowRequest Permissions
Right arrow Commercial Re-use Guidelines
for Open Access NAR Content
Google Scholar
Right arrow Articles by Asensio, J. L.
Right arrow Articles by Lane, A. N.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Asensio, J. L.
Right arrow Articles by Lane, A. N.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?