J. Am. Chem. Soc. 1999, 121, 11063-11070
11063
Conformational and Thermodynamic Properties of Parallel Intramolecular Triple Helices Containing a DNA, RNA, or 2′-OMeDNA Third Strand Juan Luis Asensio,†,§ Reuben Carr,‡ Tom Brown,‡ and Andrew N. Lane*,† Contribution from the DiVision of Molecular Structure, National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 1AA, UK, and Department of Chemistry, UniVersity of Southampton, Highfield, Southampton SO17 1BJ, UK ReceiVed June 10, 1999
Abstract: The thermodynamic stability and solution conformational properties of three intramolecular triple helices based on the sequence AGAAGA-x-TCTTCT-x-TCTTCT (x is a non-nucleotide linker) comprising a DNA duplex and DNA, RNA, or 2′-OMeDNA third strands have been compared. The most stable triple helix contains the 2′-OMe third strand, followed by the triplex containing RNA in the third strand. Comparison of the NMR spectroscopic data for the RNA hybrid triplex with those of the all-DNA triplex shows that the duplex parts of the structure are very similar; the major difference is that the RNA strand is characterized by C3′-endo sugars (except the two terminal residues). In the all-DNA triplex γ has a substantial fraction of the trans rotamer for both of the internal adenine residues (A3 and A4), whereas in the free duplex γ is g+ for these residues. In the RNA-containing triplex, only A3 shows the presence of γ(t), and in the 2′-OMe state, both A3 and A4 are γ(g+). In addition, the 2′-OMe triplex shows conformational heterogeneity. Thus, there are sugar-dependent differences in the degree of distortion in the purine strand imposed by the third strand binding. The helical parameters for the underlying duplexes are very similar in all three triplexes. However, the helical parameters for the third strands are different for the DNA versus the RNA and 2′-OMe strands, reflecting their different sugar conformations. The lower degree of distortion of the underlying duplex in the presence of the 2′-OMe third strand is consistent with higher thermodynamic stability of this triplex compared with the greater distortion of the duplex induced by both DNA and RNA third strands.
Introduction Oligonucleotides that would target the major groove of DNA, forming a triple helix and prevent transcription are possible antigene agents.1 As such, they would be very effective agents for external regulation of gene activity, since this strategy bypasses the amplification steps inherent in transcription and translation that lead to the more traditional drug targets.1,2 Triplexes can be assembled from mixtures of nucleic acid strands ranging from all-DNA to all-RNA structures. As DNA and RNA duplexes adopt radically different conformations in solution, there may be different global conformations of triple helices, depending on the composition of their strands. However, not all of the possible “hybrid” structures are stable.3,4 Of the eight possible combinations of DNA (D) and RNA (R) strands of a given sequence, two of them (rPudPy‚dPy and rPurPy‚ dPy) were unstable. Furthermore, the triplexes rPurPy‚rPy and rPudPy‚rPy were less stable than the other four possibilities, which were of comparable stability.3 On the basis of the observed affinity cleavage patterns, it was proposed that there * Corresponding author. † National Institute for Medical Research. ‡ University of Southampton. § Present address: Dr. Juan Luis Asensio, Instituto de Quimica Organica General, Juan de la Cierva 3, Madrid 28006, Spain. (1) Sun, J.-S.; He´le`ne, Curr. Opin. Struct. Biol. 1993, 3, 345-356. (2) Gowers, D. M.; Fox, K. R. Nucleic Acids Res. 1999, 27, 15691577. (3) Han, H.; Dervan, P. B. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 38063810. (4) Han, H.; Dervan, P. B. Nucleic Acids Res. 1994, 22, 2837-2844.
are two major families of conformation, one including dPudPy‚ dPy, dPudPy‚rPy, dPurPy‚dPy, dPurPy‚rPy and the other comprising rPudPy‚rPy + rPurPy‚rPy.4 These results may reflect the conformational properties of the target duplexes. It has been shown for short duplexes that dPu‚dPy is in the B form and rPu‚rPy is in the A form. In addition, the DNA‚RNA hybrids have distinct properties intermediate between A and B forms, with dPu‚rPy less A-like than rPu‚dPy.5,6 NMR studies of RNA triple helices (RR‚R) have indicated that the sugar puckers are C3′-endo and that these triplexes have features most in common with an A form structure.7 In contrast, several studies of all-DNA triplexes have shown that they adopt a conformation more similar to the B-form, with C2′-endo sugar puckers, although with some A-like characteristics.8-16 Thus, (5) Gyi, J. I.; Conn, G. L.; Lane, A. N.; Brown, T. Biochemistry 1996, 35, 12538-12548. (6) Gyi, J. I.; Conn, G. L.; Lane, A. N.; Brown, T. Biochemistry 1998, 37, 73-80. (7) Holland, J. A.; Hoffman, D. W. Nucleic Acids Res. 1996, 24, 28412848. (8) Macaya, R. F.; Schultze, P.; Feigon, J. J. Am. Chem. Soc. 1992, 114, 781-783. (9) Radhakrishnan, I.; Patel, D. J. Structure 1993, 1, 135-152. (10) Radhakrishnan, I.; Patel, D. J. Structure 1994, 2, 17-32. (11) Radhakrishnan, I.; Patel, D. J. Biochemistry 1994, 33, 11405-11415. (12) Bartley, J. P.; Brown, T.; Lane, A. N. Biochemistry 1997, 36, 14502-14511. (13) Tarko¨y, M.; Phipps, A. K.; Schultze, P.; Feigon, J. Biochemistry 1998, 37, 5810-5819. (14) Asensio, J. L.; Brown, T.; Lane, A. N. Structure 1999, 7, 1-11. (15) Nunn, C. M.; Trent, J. O.; Neidle, S. FEBS Lett. 1997, 416, 8689.
10.1021/ja991949s CCC: $18.00 © 1999 American Chemical Society Published on Web 11/13/1999
11064 J. Am. Chem. Soc., Vol. 121, No. 48, 1999 the all-DNA and all-RNA triplexes probably do form two distinct classes of structures. A DNA.DNA‚RNA hybrid triplex (DD‚R) was shown to be stable, with the RNA strand in the expected position in the major groove.17 In a recent study it was concluded that the underlying duplex was essentially in the same conformation for either the DNA or the RNA third strand, and that the sugars of the RNA third strand were mixed C2′-endo/C3′-endo.18 Furthermore, the triplex containing the RNA third strand was slightly (∆Tm ) 2.5 K) more stable than the analogue containing a DNA third strand.18 In contrast, molecular mechanics calculations19 indicated a potential energy for the DD‚D triplex lower than that for DD‚R, and differences in conformation. Thus, the relationship between strand composition, conformation, and stability is not clear. One of the problems with parallel triplexes is their relatively low thermodynamic stability at physiological pH. A simple modification that is generally thermodynamically stabilizing is methylation of the 2′-O of ribose. O2′-methylation of one of the strands of an RNA duplex is slightly stabilizing or destabilizing,20,21 whereas methoxylation of a DNA strand in a DNA duplex or the DNA strand in a DNA‚RNA hybrid duplex is generally stabilizing.21,22 The stabilization is greatest when a DNA pyrimidine strand is methoxylated.21 Methoxylation at the C2′ in the Hoogsteen strand of a parallel triple helix is stabilizing and more so than replacing a Hoogsteen DNA strand with the RNA analogue.23 Hence, this simple modification can significantly enhance the stability of parallel triplexes. Although there is a considerable body of information about the conformation of all-DNA triple helices from a variety of methodologies,3-16 relatively little is known about the structures of the hybrid triple helices.4,17-19,23 To understand the relationship between strand composition, conformation and thermodynamic stability of parallel triple helices, and the influence of the 2′-OMe modification, we have measured the pH dependence of the thermodynamic stability of three intramolecular parallel triple helices of the general sequence AGAAGA-x-TCTTCTx-TCTTCT where x is the nonnucleotide linker O-(CH2)8OPO2-O-(CH2)8OPO2-O. We have previously analyzed the thermodynamics and structures of the all-DNA analogue.24,25 In addition, we have analyzed their conformations in solution using NMR, and we are able to correlate conformational differences with the observed order of stability. Results Thermodynamics. Figure 1A shows UV melting curves of the all-DNA triplex (DD‚D) and the analogues containing either RNA (DD‚R) or 2′-OMe (DD‚R′) Hoogsteen strands at pH 7. Under these conditions, the melting curves are bimodal. The (16) Soliva, R.; Laughton, C. A.; Luque, F. J.; Orozco, M. J. Am. Chem. Soc. 1998, 120, 11226-11233. (17) van Dongen, M. J. P.; Heus, H. A.; Wymenga, S. S, van der Marel, G. A.; van Boom, J. H.; Hilbers, C. W. Biochemistry 1996, 35, 17331739. (18) Gotfredson, C. H.; Schultze, P.; Feigon. J. J. Am. Chem. Soc. 1998, 120, 4281-4289. (19) Srinivasan, A. R.; Olson, W. K. J. Am. Chem. Soc. 1998, 120, 484491. (20) Inoue, H.; Hayase, Y.; Iwai, S.; Miura, K.; Ohtsuko, E. Nucleic Acids. Res. 1987, 15, 6131-6142. (21) Lesnik, E. A.; Freier, S. M. Biochemistry 1998, 37, 6991-6997. (22) Freier, S. M.; Altmann, K.-H. Nucleic Acids Res. 1997, 25, 44294443. (23) Dagneaux, C.; Liquier, J.; Taillandier, E. Biochemistry 1995, 34, 16618-16623. (24) Asensio, J. L.; Brown, T.; Lane, A. N. Nucleic Acids Res. 1998, 26, 3677-3686. (25) Asensio, J. L.; Dhesai, J.; Bergqvist, S.; Brown, T.; Lane, A. N. J. Mol. Biol. 1998, 275, 811-822.
Asensio et al.
Figure 1. Thermodynamic stability of DD‚D, DD‚R′, and DD‚R. (A) A(260) versus temperature. Continuous lines are best-fit regression lines according to eq (1,2). (9) DD‚D at pH 7.03, (0) DD‚R at pH 6.92, (O) DD‚R′ at pH 7.10. For clarity, only every 10th datum is shown. (B) Dependence of 1/Tm versus pH. The slope of the lines is (2.303R/ ∆H)∆p (see text). (9) DD‚D, (0) DD‚R, (b) DD‚R′.
transition at the lower temperature corresponds to the melting of the triplex into duplex plus third strand, and the higher temperature transition corresponds to the melting of the duplex state into strands. As the pH is decreased to 5.5, the first transition moves to higher temperature and overlaps that of the duplex-strand transition. At neutral pH where the two transitions are well-resolved, the melting curve can be analyzed as a superposition of two sequential unfolding events as previously described,25 from which the Tm values and van’t Hoff enthalpy changes can be determined. As the latter transition is independent of pH (above pH 5), the melting curves at low pH can be analyzed by holding the duplex-strand transition fixed at its high pH values. The thermodynamic parameters for the three molecules are given in Table 1. The thermodynamic parameters for the triplex DD‚D are similar to those previously determined.25 The value of 1/Tm decreases linearly with pH in the range 5.5-8 (Figure 1B), and the triplexes become too unstable to measure at higher pH values. As previously discussed25 this is a consequence of the very high pKa value of the protonated Hoogsteen cytosine residues in the triplex state, reflecting the large stabilization of the triplex by protonation. Below pH 5.5, the value of 1/Tm begins to tail off as the pKa of the cytosines in the strand state ( ∼4.5) is approached. If ∆p is assumed to
Parallel Intramolecular Triple Helices
J. Am. Chem. Soc., Vol. 121, No. 48, 1999 11065
Table 1. Thermodynamics of the Three Triplexesa
DD‚D
DD‚R
DD‚R′
pH
Tm(T) (K)
Tm(D) (K)
∆HTb
∆HDb
∆GT298 b
4.9 5.54 6.10 6.43 6.58 6.85 7.03 7.58 5.08 5.64 6.22 6.45 6.92 7.39 7.73 8.39 5.08 5.62 6.17 6.64 7.10 7.56 7.68 7.93 8.59
333 326.8 315.5 305.4 302.8 295.3 291.4 281 340 333 321 314.6 299.2 291.4 287.6 11
dG(298)/ dpH (kJ mol-1) DD‚D DD‚R DD‚R′
11.4 11.4 11.3
d(1/Tm)/ dpHc (K-1)
d(ln(KTH2)/ d(1/T) (K-1)
2.52 × -13.9 × 2.40 × 10-4 -14.6 × 10-4 2.215 × 10-4 -16.3 × 10-4 10-4
10-4
∆HTH2
∆p
116 121 135
2.00 1.97 1.98
a Tm(T) is the melting temperature for the triplex-duplex, Tm(D) the melting temperature for the duplex, and ∆HT, ∆HD the van’t Hoff enthalpies for the triplex and duplex transition, respectively. DD‚D is the all-DNA triplex, DD‚R′ is the 2′-OMe triplex, and DD‚R is the RNA triplex. ∆GT was calculated from ∆H and Tm as described in the text. Values in parentheses were fixed in the calculation (nearly coincident transitions). b kJ mol-1. c d(1/Tm)/dpH, d(ln(KtH2)/d(1/T) calculated according to eqs 3 and 4 as described in the text. ∆HTH2 was calculated from the slope of the van’t Hoff plot. ∆p was calculated from eq 5 as (∆HTH2 + 2∆Hion)d(1/Tm)/dpH)/2.303R).
be 2 (i.e., above pH 5.5, deprotonation of the strands is complete), then ∆H(app) can be obtained from d1/Tm/dpH (Table 1) using eq 4. ∆H(app) calculated by this method is 152, 173, and 160 kJ mol-1 for DD‚D, DD‚R′, and DD‚R, respectively. These values are in reasonable agreement with the van’t Hoff enthalpy determined from direct fitting (low pH values, and see below) (Table 1). The enthalpy of dissociation of the protonated triplex to the duplex plus protonated strand state can be calculated from the dependence of 1/Tm on pH by noting that at T ) Tm, Kapp ) 1 and using KTH2 eqs 2 and 3. From the slope of the van’t Hoff plots of ln(KTH2) versus 1/Tm we obtain ∆HTH2(DD‚D) ) 116 kJ mol-1, ∆HTH2 (DD‚R′) ) 135 kJ mol-1 and ∆HTH2 (DD‚R) ) 121 kJ mol-1. These values are lower than those obtained from direct fitting or from the slope of 1/Tm versus pH because the latter include the ionization enthalpy of the protonated cytosine in the strand state, which is ∼18 kJ mol-1. Hence, for ∆p ) 2, the van’t Hoff enthalpy determined from direct curve fitting should be ∼36 kJ mol-1 higher than the value estimated from the pH dependence of Tm. These corrected ∆H values are comparable to the average values obtained by curve fitting. It is noticeable, however, that the curve-fitted ∆H is apparently pH dependent, and decreases with increasing pH. At the higher pH values, the fitted ∆H are not consistent with the values obtained from evaluating the Tm data according to eq 3. The discrepancy at high pH arises from the
Figure 2. NMR spectra of DD‚R′ showing chemical exchange in the 2′-OMe strand. 1D spectra were recorded at 600 MHz at the temperatures shown.
incomplete UV curves at low temperature (minimum T ≈ 280 K), so that lower baselines cannot be accurately defined. This gives rise to an erroneously low value of the directly fitted ∆H at high pH values (i.e., when Tm < 290 K). The enthalpy changes calculated from the pH dependence of Tm are likely to be more accurate. The enthalpy changes for the triplex-duplex transition are in the order DD‚R′ > DD‚R > DD‚D (Table 1). The O2′-Me derivative is substantially more stable (∆Tm ≈ 15 K) than the all-DNA triplex over the range 5 < pH < 8, which shows that the O2′-Me group has a substantial stabilizing influence on the triplex state. This stabilization is at least partly enthalpic. The enthalpy change is substantially larger for the O2′-Me strand than for the DNA strand as determined from the van’t Hoff analysis (Table 1). The Tm values are in the order DD‚R′ > DD‚R > DD‚D. However, ∆G is a thermodynamically more appropriate measure of stability. Using the experimental values of Tm and the derived values of ∆H, we have calculated ∆G at 298 K (Table 1); the ∆G(298) values are also in the order DD‚R′ > DD‚R > DD‚D. The values of d∆G(298)/dpH are all similar indicating that the difference in stability remains constant at least up to pH 8. This is as expected if the pKa is independent of the nature of the sugar, and pKa′ . pH (cf. eq 5). It is also possible to calculate ∆G(298) for the dissociation of TH2 into D + SH2 from the data in Table 1. We find ∆G(298) is -34, -30, and -26 kJ mol-1 for DD‚R′, DD‚R, and DD‚D, respectively. This implies a stabilization by protonation of at least 34-40 kJ mol-1 (17-20 kJ mol-1 cytosine) compared with the fully unprotonated triplex. It is thought that the hydrogen-bonding capacity of the C2′-OH in RNA adds to the overall thermodynamic stability of RNA versus DNA.26 Although the triplex containing the RNA third strand is more stable than the all-DNA triplex, it is less stable than the O2′-Mecontaining triplex. Hence, the observed order of stability for these triplex is perhaps surprising. We have therefore used highresolution NMR to determine the conformational properties of these triplexes and shed some light on their stability. Conformation of the Three Triplexes. The NMR spectra of the all-DNA triplex DD‚D have been analyzed previously.24,25 Although the spectra of DD‚D and DD‚R show only a single set of resonances, the spectrum of DD‚R′ shows additional peaks, as shown in Figure 2. In addition, at 30 °C the integrals of the U,C H6 and H5 of nonterminal residues in DD‚R′ are (26) Egli, M.; Portmann, S.; Usman, N. Biochemistry 1996, 35, 84898494.
11066 J. Am. Chem. Soc., Vol. 121, No. 48, 1999 only ∼70% of those of the Watson-Crick CH6 resonances. Exchange cross-peaks involving these resonances were observed in ROESY spectra recorded at 40 °C (not shown), and the chemical shifts of the minor species do not correspond to those of the duplex + strand state. Hence, the observed exchange is due to two (or more) different triplex conformations. Furthermore H6 and H5 resonances become broader on increasing the temperature between 15 and 50 °C, showing increasing exchange. To determine the sugar conformations, we have recorded DQF-COSY spectra of the three triplexes. The cross-peaks between H1′ and both H2′, H2′′ indicated that the sugars of the DNA residues in the Watson-Crick strands are primarily in the “S” conformation in all three triplexes (not shown). This was confirmed by detailed analysis of the sums of coupling constants, which gives values of the pseudorotation phase angle Ps in the range 130-170° and fraction of the S state fs of >0.9 in the purine strand and 0.6-0.8 in the pyrimidine strand. There is little difference in the sugar conformation and equilibria in the duplex moieties of the three triplexes, and they are all predominantly S. Furthermore, the sugar conformations determined for the isolated duplex DD are similar to those in the triplex states, indicating that only small changes in sugar conformations are induced by forming the triple helices. This is in contrast to an FT-IR study of triple helices containing an RNA third strand, where the purine strand was reported to contain “N”-type sugars.23 In DD‚D, the sugars in the Hoogsteen strand show a substantial fraction of the N state, especially the two protonated cytosine residues, though the thymines are predominantly S. In contrast, only the terminal Hoogsteen residues of DD‚R and DD‚R′ showed COSY H1′-H2′ crosspeaks. This indicates that the riboses in DD‚R and DD‚R′ are predominantly in the N conformation, except for the terminal residues which may exist as a mixture of N and S states. This conclusion is supported by the observation of strong H2′(i)H6(i+1) NOE cross-peaks in the Hoogsteen strand of DD‚R and DD‚R′, which is characteristic of N-type sugars. In the DD‚D triplex, we have previously observed that the backbone angle γ for the two central A residues is t rather than the more common g+.14,24 TOCSY and ROESY spectra of the three triplexes were compared to determine γ for the residues in DD‚R and DD‚R′. The downfield shifted H4′ resonance, and the splitting of the H4′ resonance characteristic of the trans rotamer were evident for A3 in DD‚R, but not for A4. All of the other H3′-H4′ cross-peaks for the deoxyriboses appeared as narrow singlets. In contrast, none of the deoxyribose H4′ of DD‚R′ were unusually downfield-shifted, and these resonances appeared as narrow singlets (not shown). The apparent ∑4′ values were estimated from the width at half-height for the singlets (which is an upper limit to ∑4′), or from the splitting in NOESY or ROESY spectra (which gives a lower limit to ∑4′) (Table 2). In addition to the coupling to H5′ and H5′′, the observed ∑4′ includes contributions from 3J3′4′, the small 4J4′-P, and the line width. As the deoxypurine residues are predominantly in the S state, 3J3′4′ < 2-3 Hz. In the g+ rotamer, both 3J4′5′ and 3J4′5" are small (
