Thymine-methyl/π interaction implicated in the sequence-dependent deformability of DNA

Abstract

The crystal structures of deoxy-oligonucleotides were retrieved from the Nucleic Acid Database and analyzed with the use of our program CHPI. The structure of 5′-ApTpApT-3′ has been shown to be stabilized by the 5-methyl group in the thymidine moiety that favorably interacts with the adenine π-ring preceding it. H2′ of the deoxyribose in adenine also interacts with the thymine ring next to it. Since a 5′-ApT-3′ sequence is accompanied by another 5′-ApT-3′ in the complementary strand, the interaction is duplicated, thus forming a ‘twin A/T-Me interaction’. Coordinates of oligonucleotides with A-T rich sequences were retrieved and analyzed. In almost every case, the thymidine 5-methyl group favorably interacts with an adenine ring in the same strand. The structure of duplexes incorporating A-tracts was also analyzed. The 5-methyl group in the thymidine moiety has been found to interact favorably with the base π-ring before it. Since an A-tract is lined with an oligo-T sequence in the complementary strand, a successive N/T-Me stacking may contribute in making the A-tracts robust and straight. The possible involvement of the  N/T-Me and the twin  A/T-Me motif in the deformability of DNA has been suggested. The role of methyl groups in modified DNA has been discussed on a similar basis.

INTRODUCTION

In 1978, Viswamitra, Kennard and their coworkers (1,2) determined the crystal structure of a deoxytetranucleotide 5′-P-adenylyl-3′,5′-thymidylyl-3′,5′-adenylyl-3′,5′-thymidine (pATAT). An extraordinary feature of their finding was that the conformation of the sugar–phosphate backbone is significantly different from the canonical structure of B-DNA; the conformation of the phosphodiester bridge changes between the A-T and T-A steps. Thus, the first two bases, A and T, pair with the complementary bases from another molecule, but then the phosphate backbone swings away from the helical orientation, to make another pairing of the subsequent A-T sequence with a third molecule. T is stacked with the preceding adenine base, but there is no interaction of T with A next to it. The finding was followed by Klug et al. (3) reporting an unusual conformation of a DNA polymer with repeating A-T sequences (alternating B-DNA). They noted the methyl group of thymine in an A-T step to place itself over the five-member ring of the purine, whereas in the T-A step there was no such interaction. They argued this as the primary cause of the alternating B-DNA structure and suggested the importance of the C⁵ substituent (CH₃) in thymidine; the above points were discussed in detail by Saenger (4) in his textbook.

The CH/π interaction (5) is a weak attractive force working between CH groups and π-systems (a comprehensive literature list regarding the CH/π hydrogen bond is available on the following website: http://www.tim.hi-ho.ne.jp/dionisio). The importance of this molecular force has been recognized in various fields of chemistry (6) and biochemistry (7). Similar to the ordinary hydrogen bond (8) and CH/O interaction (9), this concept is useful in understanding the bases of chemical and biological phenomena. These include crystal packing (10,11), chiral recognition (12), self assembly (13), the specific structure of host/guest compounds (14,15), as well as the specificity of biopolymers. Previously, we reported, by analyzing crystal coordinates in the Brookhaven Protein Data Bank (PDB), that the CH/π interaction plays an important role in the structure of G proteins, SH2 motifs (16), MHC antigens (17) and their complexes with specific substrates. Statistical studies of the PDB demonstrated that CH/π interaction generally works in proteins (18,19). As for the interactions involving DNA, Hirao and coworkers (20) reported an unusually high-field ¹H-NMR shift of H4′ of adenine in a stable hairpin motif GCGAAGC. Nishinaka et al. (21) reported on the importance of CH/π interaction (deoxyribose-CH versus base-aromatic ring) in stabilizing the structure of a protein/DNA complex. Chou and Tseng (22) suggested a role of the CH/π interaction in stabilizing the hairpin structure of human centromeres and telomeres. We studied CH/π interactions in a promoter DNA (TATA-box) complexed with TATA-box binding protein (TBP) (23). A number of CH/π contacts have been disclosed at the border of the DNA and TBP. Here we studied the issue, by a program search of the Nucleic Acid Database (NDB), to know if such a structural feature is generally found in the crystal data of oligonucleotides.

MATERIALS AND METHODS

The method of exploring CH/π interactions in biopolymers was reported in our previous papers (7,16). Thus, a program (CHPI) was written to examine distances between CHs and π groups in proteins and nucleic acids. To participate in a CH/π interaction, the hydrogen atom should be placed above a π-plane (region 1 in Fig. 1). In order to cover other possibilities (regions 2 and 3), several kinds of CH/C distance and angle parameters were defined. In the present study, the π-system is the aromatic ring of a purine or pyrimidine base. However, crystal data in the NDB do not contain coordinates of hydrogen atoms. Therefore, hydrogens were generated on carbons and their positions were optimized. The CH/π interatomic distance <3.05 Å [2.9 Å (1.2 Å for C-H plus 1.7 Å for a half thickness of the aromatic molecule) × 1.05] (24) with reasonable angle parameters was considered as relevant for the presence of the CH/π interaction. For a methyl group, the position of hydrogen atoms may vary with rotation about the C-CH₃ bond. We therefore restricted the position by putting one of the three methyl C-H hydrogens to point to the closest sp² atom (C or N) of the aromatic ring. We think this procedure is reasonable for the present purpose of surveying CH/π contacts involving a methyl group, since energy minimization often gives inadequate distance by repulsive potentials in the force field. Previous statistical studies on the Cambridge Structural Database and recent high-level ab initio calculations support this. Namely, the CH/π interaction shows a distinct directionality; this is most effective when a C-H bond orients itself to the π-surface (25). Further, the energy of the CH/π interaction, being partly electrostatic in nature (26), does not fall off sharply with distance. However, in view of the limitation to the present methodology we do not discuss the interatomic distances in detail. Only non-complexed duplex B-structures with resolution <2.5 Å (except BDL015) were examined.

Figure 1.

Method for exploring CH/π contacts with a fused aromatic ring of purine. (A) O, center of the π-plane. A¹ and A², nearest and second nearest sp²-atoms, respectively, to the hydrogen H¹. ω, dihedral angle defined by A¹OA² and H¹A¹A² planes. θ, H¹···C^m–A¹ angle. D_pln, perpendicular distance between H¹ and the π-plane (H¹/I). D_atm, H¹A¹ interatomic distance. D_lin, distance between H¹ and the line A¹–A² (H¹/J). (B) Regions to be searched. Region 1, zone where H¹ is above the ring. Regions 2 and 3, zones where H¹ is out of region 1 but may interact with π-orbitals. Unless otherwise noted, the program was run to search for short H¹/π contacts with the following conditions: D_max = 3.05 Å; D_pln < D_max (region 1); D_lin < D_max (region 2); D_atm < D_max (region 3); ω_max = 127.5°, –ω_max < ω < ω_max; θ < 62.2°. D_hpi, CH/π distance (D_pln for region 1, D_lin for region 2, D_atm for region 3).

RESULTS AND DISCUSSION

Tetra-deoxynucleotide pATAT

Table 1 gives an output (edited) of our CHPI analysis of ATAT (NDB code UDD006, resolution 1.04 Å).

Table 1. CHPI analysis of tetra-deoxynucleotide ATAT.

π-Group			CH			Distance (Å)
IDRD	RES	VATM	IDRD	RES	VATM	Dhpi	RG
A1	A	C8	A2	T	H5M	2.86	2
A2	T	C6	A1	A	H2′	2.80	3
A3	A	C8	A4	T	H5M	2.73	1
A4	T	C6	A3	A	H2′	2.71	3
B1	A	C8	B2	T	H5M	2.86	2
B2	T	C6	B1	A	H2′	2.80	3
B3	A	C8	B4	T	H5M	2.73	1
B4	T	C6	B3	A	H2′	2.71	3
C1	A	C8	C2	T	H5M	2.86	2
C2	T	C6	C1	A	H2′	2.80	3
C3	A	C8	C4	T	H5M	2.73	1
C4	T	C6	C3	A	H2′	2.71	3

IDRD, residue ID; RES, residue name; VATM, atom involved.

D_hpi, CH/π distance (Å) (D_pln for region 1, D_lin for region 2, D_atm for region 3). RG, region.

Here we see that a thymine-methyl group is in contact with the adenine base preceding it. One of the three hydrogens of the T-methyl (H⁵M) is close to C⁸ of the purine ring. Figure 2 illustrates this. An A-T sequence will become stickier than other steps by the two CH/π interactions. Hereafter, we refer this as the ‘twin A/T-Me interaction’. H2′ of the deoxyribose in adenine also interacts with the thymine ring next to it. On the contrary, such a situation is absent in the T-A step. Since an A-T sequence is lined with another A-T in the complementary strand, this makes up two sets of such interactions. We suggest this to be the chemical basis of the extraordinary structure of ATAT and the alternating B-DNA. Hunter (27) pointed out the possibility that intra-strand steric clashes occur between the T-methyl and the 5′-neighboring sugar in the A-T steps. π/π Stacking at purine–pyrimidine steps may be greater than at pyrimidine–purine steps. However, we do not discuss these possibilities since the purpose of the present paper is to investigate whether the phenomenon is understood in terms of the CH/π interaction.

Figure 2.

Schematic illustration of the CH/π network in ATAT (UDD006), disclosed by the program CHPI. Twin A/T-Me interactions are shown in red boxes.

Nucleotides with more than two A-T steps

We then analyzed, systematically, the structure of B-DNA duplexes, deposited in the NDB, bearing more than two A-T steps in the sequence. Drug complexes and mismatch DNAs were excluded. When there are more than two entries of identical sequence, we chose crystal structures of the best resolution. The oligonucleotides analyzed were CGATATATCG (BDJ036) (28), CGCATATATGCG (BDL007) (29), CGCGATATCGCG (BDL078) (30), CATGGGCCCATG (BD0026) (31), CATGGCCATG (BDJ051) (32), CGATTAATCG (BDJ031) (33), CCATTAATGG (BDJ055) (34) and CGATCGATCG (BDJ025) (35).

Table 2 gives the result for a decanucleotide CGATATATCG. Figure 3 illustrates the result. Here we note many CH/π interactions between a T-methyl and the base preceding it. Thus, an N-T sequence becomes stickier if N = A, by the duplication of the π/T-Me interaction (twin A/T-Me interaction).

Table 2. CH/π interactions in CGATATATCG.

π-Group			CH			Distance (Å)
IDRD	RES	VATM	IDRD	RES	VATM	Dhpi	RG
A2	G	C8	A1	C	H2′	2.27	2
A3	A	C8	A4	T	H5M	2.84	2
A5	A	C8	A6	T	H5M	2.72	3
A7	A	C8	A6	T	H1′	2.99	2
A7	A	C8	A8	T	H5M	3.01	3
A10	G	C8	A9	C	H2′	2.66	3
B2	G	C8	B1	C	H1′	2.80	2
B3	A	C8	B4	T	H5M	2.66	3
B5	A	C8	B4	T	H2′	2.66	3
B5	A	C8	B6	T	H5M	2.50	3
B7	A	C8	B6	T	H1′	2.79	3
B7	A	C8	B8	T	H5M	2.94	1
B8	T	C6	B7	A	H2′	2.45	3

IDRD, residue ID; RES, residue name; VATM, atom involved.

D_hpi, CH/π distance (Å) (D_pln for region 1, D_lin for region 2, D_atm for region 3). RG, region.

Figure 3.

Crystal structure of CGATATATCG (BDJ036) analyzed by the program CHPI. (A) CH/π network. Thymine in white, adenine in yellow and G/C pairs are in green. Red sticks indicate CH/π contacts. (B) Schematic illustration of the network. Twin A/T-Me interactions are shown in red boxes.

Table 3 summarizes the results. Twin A/T-Me interactions are italic. N/T-Me interactions disclosed in the corresponding position of the strands are indicated in bold. Underlined steps indicate N/T-Me interactions disclosed in only one strand. Note that A/T-Me interactions are found in every oligonucleotide. The number of A-T steps is 50, including those given in Tables 3 and 4, and BDL042 (CGTAGATCTACG) (36). We found 47 A/T-Me short (<3.05 Å) contacts in these steps.

Table 3. CH₃/π interactions in AT-rich DNA.

Sequence	NDB code	Resolution (Å)	sp2-Atoma
CGAT AT ATCG	BDJ036	1.70	C8, C8, C8 // C8, C8, C8 (6/6)
CGCAT ATATGCG	BDL007	2.2	C8, C8, 3.4 // 3.3, C8, C8 (4/6)
CGCGAT ATCGCG	BDL078	2.2	C8, C8 // C8, C8 (4/4)
CATGGGCCCATG	BD0026	1.30	N7, N7 // N7, N7 (4/4)
CATGGCCATG	BDJ051	2.0	C8, C8 // C8, C8 (4/4)
CGATTAATCG	BDJ031	1.50	C8, C6, C8 // C8, C6, C8 (6/6)
CCATTAATGG	BDJ055	2.30	C8, C6, C8 // C8, 3.2, C8 (5/6)
CGATCGATCG	BDJ025	1.50	C8, 3.1 // C8, C8 (3/4)

^aAtoms interacting with the methyl hydrogen are shown [from left to right: 5′–3′, then 5′–3′ in the complementary strand. N⁷ and C⁸, purine numbering; C⁶, pyrimidine numbering. Numbers refer to the distance (Å) longer than the cut-off value; a double slash separates the complementary strands]. Numbers precedent to the slash refer to the number of N/T-Me contacts <3.05 Å, while those following the slash indicate the number of N-T steps in the oligonucleotide.

Table 4. CH₃/π interactions disclosed in adenine-rich oligonucleotides.

Sequence	NDB code	Resolution (Å)	sp2-Atoma
CTTTT CTTTG	BDJ081	1.85	C6, C6, C6, C6, C6, C6, C6 (7/7)
CGCTTTTTTGCG	BDL006	2.50	C6, C6, C6, C6, C6, C6 (6/6)
CGCTTTTTAGCG	BDL015	2.60	C8, C6, C6, C6, C6 // C8 (6/6)
CGT TTTTTCGCG	BDL047	2.30	C8, 3.4, C6, C6, C6, C6 (5/6)b
CGCAAATTTGCG	BDL038	2.2	C8, C6, 3.2 // C8, C6, C6 (5/6)
CGCGAATTCGCG	BD0041	1.20	C8, C6 // C8, C6 (4/4)
GCGAATTCGCG	BD0018	1.30	C8, C6 // C8, C6 (4/4)
CGCAATTGCG	BDJ069	2.30	C8, C6 // C8, C6 (4/4)
GCGAATTCG	UDI030	2.05	C8, C6 // C8, C6 (4/4)
CGTGAATTCACG	BDL029	2.5	C8, C8, C6 // 3.5, C8, C6 (5/6)
CGCGTTAACGCG	BDL059	2.3	C8, 3.4 // 3.3, C6 (2/4)

^aAtoms interacting with the methyl hydrogen are shown [from left to right: 5′–3′, then 5′–3′ in the complementary strand. C⁸, purine numbering; C⁶, pyrimidine numbering. Numbers refer to the distance (Å) longer than the cut-off value; a double slash separates the complementary strands]. Numbers precedent to the slash refer to the number of N/T-Me contacts <3.05 Å, while those following the slash indicate the number of N–T steps in the oligonucleotide.

^bThree independent molecules are present: A/B, C/D and E/F pairs. B-chain (5/6), D-chain (4/6), F-chain (4/6). Total: (13/18).

The T-methyl group is always close to C⁸ (purine) or C⁶ (pyrimidine). CATGGGCCCATG (BD0026) is an exception; in this case, T-methyl is close to N⁷ of the adenine ring. The reason remains unclear but this dodecamer is a stable intermediate in the A-DNA↔B-DNA helix transition. Therefore, we analyzed the crystal structure of A-DNA octamers GGGATCCC (ADH007) (37) and GGCATGCC (ADH076) (38). In ADH007, T-methyl is close to N⁷ and in ADH076 T-methyl is almost at a similar distance from N⁷ and C⁸. The A-T step also forms a twin A/T-Me interaction (italic) in these A-DNAs. It may be concluded from the above facts that the twin A/T-Me interaction is a general motif for DNA duplexes incorporating A-T steps.

Adenine-rich sequences (A-tracts)

The homopolymer A_n has many unusual properties distinct from random sequence B-DNA. For instance, fibers of A_n are not affected by the environmental changes such as humidity or salt concentration. Adenine-rich sequences (A-tracts) fail to wind around nucleosome cores. Further, the A-rich sequence has been known to be a source of the DNA curvature; at least four continuous A-A steps are necessary for the overall bending of the DNA (39).

To address this question, Nelson et al. (40) solved the crystal structure of an A-tract, CGCAAAAAAGCG/CGCTTTTTTGCG. They found unusual conformational features differentiating the dodecamer from other B-DNA sequences. For instance, the helical repeat is 10 bp per turn, whereas in random DNA it is normally 10.6. The average axial rise in the A-tract is 3.1 Å, in contrast to 3.4 Å for orthodox B-DNA. The A:T base pairs of the central region of the dodecamer show a large propeller twist. The minor groove in the A-tracts is narrowed, consequently. They attributed the result to the three-center hydrogen-bonding network between the opposing strands. The stacking of adenine rings and the van der Waals contact of the thymine-methyl group to the adjacent base were argued as being also relevant. Dickerson and colleagues (29) determined the crystal structure of an alternating A-T sequence CGCATATATGCG. They compared the crystallographic data of the dodecamer with those of four A-tracts and argued that the cross-strand hydrogen bonds were the principal source of the large propeller twist of the A-tracts. Other workers suggested that the minor groove spine of hydration (41) or cations (42) is responsible.

We analyzed the crystal structure of A-tracts. Every sequence incorporating four to seven A:T (or T:A) base pairs was examined. Drug complexes and mismatch DNAs were excluded. Coordinates of the best resolution were used when we found two or more DNAs of equal sequence (e.g. there are nine sets of coordinates for CGCGAATTCGCG in the present version of the NDB). The oligonucleotides analyzed were CAAAGAAAAG (BDJ081) (43), CGCAAAAAAGCG (BDL006) (40), GCGAAAAATCGC (BDL015) (44), CGCGAAAAAACG (BDL047) (41), CGCAAATTTGCG (BDL038) (45), CGCGAATTCGCG (BD0041) (46), GCGAATTCGCG (BD0018) (47), CGCAATTGCG (BDJ069) (48), GCGAATTCG (UDI030) (49), CGTGAATTCACG (BDL029) (50) and CGCGTTAACGCG (BDL059) (51).

Table 5 lists interactions disclosed in CAAAGAAAAG/CTTTTCTTTG (BDJ081). Figure 4 illustrates the result. We see many CH/π interactions between T-methyl and the base before it. The stacking specifically involves a H⁵M/π interaction with C⁶ of a pyrimidine ring.

Table 5. CH/π interactions disclosed in CAAAGAAAAG/CTTTTCTTTG.

π-Group			CH			Distance (Å)
IDRD	RES	VATM	IDRD	RES	VATM	Dhpi	RG
A2	A	N7	A1	C	H2′	2.28	2
A2	A	C4	A1	C	H1′	2.95	1
A10	G	C8	A9	A	H2′	2.98	2
A10	G	C8	A9	A	H2′	2.92	3
B1	C	C6	B2	T	H5M	2.57	2
B2	T	C6	B1	C	H2′	2.29	3
B2	T	C6	B3	T	H5M	2.87	2
B3	T	C6	B4	T	H5M	2.88	3
B4	T	C6	B5	T	H5M	2.50	2
B6	C	C6	B7	T	H5M	2.43	1
B7	T	C6	B8	T	H5M	2.95	3
B8	T	C6	B9	T	H5M	2.96	3
B9	T	C6	B8	T	H2′	2.87	3
B10	G	N7	B9	T	H2′	2.37	3
B10	G	N7	B9	T	H1′	3.01	1

IDRD, residue ID; RES, residue name; VATM, atom involved.

D_hpi, CH/π distance (Å) (D_pln for region 1, D_lin for region 2, D_atm for region 3). RG, region.

Figure 4.

Crystal structure of CAAAGAAAAG/CTTTTCTTTG (BDJ081) analyzed by the program CHPI. (A) CH/π network. Thymine in white, adenine in yellow, G/C pairs in green. Red sticks indicate CH/π contacts. (B) Schematic illustration of the network.

Table 4 summarizes the results (the complementary strand is shown for BDJ081, BDL006, BDL015 and BDL047). The underlined steps indicate N/T-Me interactions. N/T-Me interactions disclosed in the corresponding part of the strands are in bold. Twin A/T-Me interactions are italic.

Note that, in almost every step involving thymine, CH/π interactions are found between T-methyl and the base before it. Since an oligo-A sequence is lined with an oligo-T sequence in the complementary strand, a successive T/T-Me sequence obtains in the duplexes. A successive N/T-Me stacking such as N/T/T/T/T/T/T (N = C, G or T) may contribute in making the conformation of A-tracts stable. Further, a twin A/T-Me interaction has been found in every oligonucleotide with an AATT or AAATTT sequence. Five DNA duplexes (BD0041, BD0018, BDJ069, UDI030 and BDL029) bear a common sequence AATT but this is flanked by a different kind (or number) of G:C pairs. These oligonucleotides show the same patterns, nevertheless, in terms of the CH/π interaction. In GCGAATTCGCG (BD0018) and GCGAATTCG (UDI030), where the central AATT sequence is flanked by a differing number of G:C pairs, the opposing two strands are offset to each other, thus keeping a twin A/T-Me interaction in the central part of the duplex. Figure 5 illustrates this. We think it likely that the twin A/T-Me interaction is strong enough to make such offset pairings for the unsymmetrical oligonucleotides. It is tempting to speculate that an A-T part seeks another A-T in the process of double-helix formation, thus forming a stable AT core. Pairing of the other parts may occur, subsequently.

Figure 5.

Schematic illustrations of CH/π interactions disclosed in (A) GCGAATTCGCG (BD0018) and (B) GCGAATTCG (UDI030). Twin A/T-Me interactions are shown in red boxes.

Figure 6 compares CH/π interactions in four dodecamers CGCAAAAAAGCG (40), CGCAAATTTGCG (45), CGCGAATTCGCG (46) and CGCATATATGCG (29).

Figure 6.

Schematic illustrations of the networks (cross-strand hydrogen bonds and CH/π interactions; twin A/T-Me interactions are in red boxes) in (A) CGCAAAAAAGCG (BDL015), (B) CGCAAATTTGCG (BDL038), (C) CGCGAATTCGCG (BD0041) and (D) CGCATATATGCG (BDL007). The plausible twin A/T-Me interaction is indicated by a dotted box.

According to Dickerson and colleagues (29), the number of cross-strand hydrogen bonds decreases in the order: CGCAAAAAAGCG > CGCAAATTTGCG > CGCGAATTCGCG > CCGATATATGCG (the cross-strand hydrogen bond can be formed only at the A-A step). They argued this as the primary cause of the high propeller twist and robustness of the A-tracts. However, note that the number of N/T-Me interactions (including potential ones) is six in CGCAAAAAAGCG (Fig. 6A), CGCAAATTTGCG (Fig. 6B) and CCGATATATGCG (Fig. 6D). CGCGAATTCGCG (Fig. 6C) is an exception since this bears only four A:T pairs. In CGCAAAAAAGCG, the twin A/T-Me interaction is absent but in CGCAAATTTGCG and CGCGAATTCGCG, we see a twin A/T-Me interaction at the central region. In the alternating AT tract CCGATATATGCG, three twin A/T-Me interactions are possible but they are disconnected at every T-A step. Therefore, our suggestion is that the consecutive N/T-Me interactions and the twin A/T-Me interaction are both responsible for making the structure of A-tracts compact.

The dominance of A-tracts as the primary source of the global DNA curvature is well known (52). However, a firm structural basis at the molecular level has not been presented. The intra-strand N/T-Me interaction and the twin A/T-Me interaction seem to be important contributors to the robustness of the A-tracts, which, in turn, is responsible for the overall bending of DNA in physiological conditions. In this regard, it may be pointed out that the CH/π interaction, being largely dispersive in nature (25,26), functions well in protic media. On the contrary, the ordinary hydrogen bond and Coulomb force do not effectively work in water.

Hagermann (53) reported that oligonucleotide duplexes (GAAAATTTTC)_n or (CAAAATTTTG)_n showed abnormally small mobility in gel electrophoresis while oligonucleotides (GTTTT*AAAAC)_n or (CTTTT*AAAAG)_n moved much faster. The results seem reasonable since in the former a twin A/T-Me interaction (italic) brings the entire sequence (bold) more straight, like a double-arm lock. On the contrary, in the latter, such an effect cannot be expected to operate; the T-A step (starred) behaves as a hinge between the rods. Crothers and colleagues. (54) compared the mobility of (GACTAAAAATGACTAAAAATGACTAAAAAAT)_n with that of (TCGGAAAAAGTCGGAAAAAGTCGGAAAAAAG)_n. In the former there are twin A/T-Me interactions extending the rods, while in the latter such an effect is absent. A slower mobility may be anticipated for the former; this is what Crothers and colleagues found in the electrophoresis experiment. From the electrophoretic data and crystallographic evidence, Dickerson and colleagues (34) concluded that the ‘non-A-tract bending model’ of Calladine et al. (55) is appropriate. We concur with this view.

Modified DNA

Methylation at position C⁵ of cytosine or N⁶ of adenine is interesting with regard to various biochemical events such as replication, transcription and resistance to restriction enzymes. Such a modification may result in the change in the robustness of DNA (56). Therefore, we examined CH/π interactions in methylated DNA in the NDB. The structure of those bearing inosine or uridine was excluded.

Table 6 lists the CH₃/π-base interactions disclosed in oligonucleotides CGCGAATTC⁺GCG [BDLB73: B.L.Partridge and S.A.Salisbury, unpublished work], CGCGAA⁺TTCGCG (BDLB13) (57), CGCGA^%ATTCGCG (BD0031) (58), CGATCGA⁺TCG (BDJB48) (59) and CCAGGCC⁺TGG (BDJB50) (60) (C⁺, 5-methylcytosine; A⁺, N⁶-methyladenine; A^%, N⁶-methoxyadenine).

Table 6. CH₃/π interactions disclosed in methylated DNA.

Sequence	NDB code	Resolution (Å)	sp2-Atoma
CGCGAATTC+GCG	BDLB73	2.03	C8, C6, C6 // C8, C6, C6 (6/6)
CGCGAA+TTCGCG	BDLB13	2.0	N7, C8, C6 // N7, C8, C6 (6/6)
CGCGA%ATTCGCG	BD0031	1.6	N7, C8, C6 // N7, C8, C6 (6/6)
CGATCGA+TCG	BDJB48	2.0	C8, N7, C8 // C8, N7, C8 (6/6)
CCAGGCC+TGG	BDJB50	1.70	C6, C6 // C6, C6 (4/4)

C⁺, C⁵-Me; A⁺, N⁶-Me; A^%, N⁶-OMe.

^aAtoms interacting with the methyl hydrogen are shown (from left to right: 5′–3′, then 5′–3′ in the complementary strand. C⁸ and N⁷, purine numbering; C⁶, pyrimidine numbering. A double slash separates the complementary strands). π-Atoms interacting with the methyl group are italicized.

Figure 7 illustrates the result for a representative case of CGCGAATTC⁺GCG (BDLB73). Figure 8 compares the results for BDLB73, BDLB13 and BD0031.

Figure 7.

A part of the crystal structure of CGCGAATTC⁺GCG (BDLB73). Thymine in white, adenine in yellow, G/C pairs in green. Methyl group in 5-methylcytosine in purple. Red sticks indicate the CH₃/π contacts.

Figure 8.

Schematic illustrations of CH/π interactions disclosed in methylated DNA. (A) CGCGAATTC⁺GCG (BDLB73; C⁺, 5-methylcytosine), (B) CGCGAA⁺TTCGCG (BDLB13; A⁺, N⁶-methyladenine) and (C) CGCGA^%ATTCGCG (BD0031; A^%, N⁶-methoxyadenine).

It is remarkable that, in every case, CH₃ groups in the methylated DNA form additional interactions with the π-base preceding it. BDLB13 (Fig. 8B) is an extraordinary case, where three consecutive N-Me/π interactions are involved in both the strands. It has been noted, by crystallography, that methylation does not significantly alter the conformation of DNA (57,60). It is likely that methylation modulates the recognition of DNA by enzymes such as restriction endonucleases. However, we think it also possible that the modification brings about an appreciable change in the stiffness of duplexes and thus affects the overall bending of DNA (56).

CONCLUSION

To summarize, an N-T step has been shown to often accompany interaction of the thymine-methyl group. The N/T-Me interaction is duplicated, if N = A, by another A-T in the complementary strand to make up a ‘twin A/T-Me interaction’. An A-T step becomes stiffer than other N-T sequences. Since an A-tract is lined with an oligo-T in the complementary strand, a consecutive N/T-Me stacking such as N/T/T/T/T/T/T may contribute in making the A-rich steps straight and robust. In view of this, the term ‘T-tract’ may be more adequate. The issue of global DNA-bending can be interpreted on these grounds. Consequences of the methylation of cytosine or adenine to various biochemical events may also be understood on a similar basis.

In support of the CH/π concept (soft acid/soft base interaction), a number of high-level ab initio calculations have recently appeared. According to theory (25,26,61–63) and experimental data (64–67), enthalpy of a single CH/π bond is estimated to be around 1 kcal mol^–1. A noteworthy feature of this attractive force, albeit weak, is that many CHs and π groups can cooperatively participate in an interaction (6,7,68). Further, this type of interaction is entropically advantageous. A CH₃ group has the C₃ symmetry and thus a triple chance to interact with a π-ring. The DNA base has a wide π-surface capable of interacting with many CH groups. Another remarkable feature of the CH/π interaction is that this is effective either in polar protic solvents such as water unlike the conventional hydrogen bond (hard acid/hard base combination), or in non-polar environments unlike the so-called ‘hydrophobic effect’. This point is crucial in considering the role of the CH/π interaction in biochemical processes.

There are two fundamental differences between the structure of DNA and RNA. First, the existence of 2-deoxyribose in the former and ribose in the latter. The use of deoxyribose by DNA may, at least partly, be understood in terms of the presence of an extra hydrogen at the C2′ position, thus enabling more effective CH/π interaction. The second issue is the presence of T in DNA whereas in RNA uracil (U) is found in place of thymine. The only structural variance is the existence of a methyl group in the 5-position of the pyrimidine ring in T, while in U this is absent. Deformation of DNA curvature is essential in initiating important biochemical events such as replication, transcription or binding to nucleosome cores. It is tempting to speculate that these processes involve interactions of thymine, more specifically its methyl group.