Conformational Variants of Duplex DNA Correlated with Cytosine-rich Chromosomal Fragile Sites*

We found that several major chromosomal fragile sites in human lymphomas, including the bcl-2 major breakpoint region, bcl-1 major translocation cluster, and c-Myc exon 1-intron 1 boundary, contain distinctive sequences of consecutive cytosines exhibiting a high degree of reactivity with the structure-specific chemical probe bisulfite. To assess the inherent structural variability of duplex DNA in these regions and to determine the range of structures reactive to bisulfite, we have performed bisulfite probing on genomic DNA in vitro and in situ; on duplex DNA in supercoiled and linearized plasmids; and on oligonucleotide DNA/DNA and DNA/2′-O-methyl RNA duplexes. Bisulfite is significantly more reactive at the frayed ends of DNA duplexes, which is expected given that bisulfite is an established probe of single-stranded DNA. We observed that bisulfite also distinguishes between more subtle sequence/structural differences in duplex DNA. Supercoiled plasmids are more reactive than linear DNA; and sequences containing consecutive cytosines, namely GGGCCC, are more reactive than those with alternating guanine and cytosine, namely GCGCGC. Circular dichroism and x-ray crystallography show that the GGGCCC sequence forms an intermediate B/A structure. Molecular dynamics simulations also predict an intermediate B/A structure for this sequence, and probe calculations suggest greater bisulfite accessibility of cytosine bases in the intermediate B/A structure over canonical B- or A-form DNA. Electrostatic calculations reveal that consecutive cytosine bases create electropositive patches in the major groove, predicting enhanced localization of the bisulfite anion at homo-C tracts over alternating G/C sequences. These characteristics of homo-C tracts in duplex DNA may be associated with DNA-protein interactions in vivo that predispose certain genomic regions to chromosomal fragility.

The sequence-specific structural variations of the dsDNA 5 helix are well documented (1)(2)(3)(4). For example, DNA sequences with consecutive A (or T) bases can adopt a variation of the B-form helix with a narrower minor groove, whereas sequences with consecutive G (or C) bases are more prone to adopt the A-form helix under dehydrating conditions (5). The A-form helix observed in RNA and DNA-RNA hybrid duplexes is present during transcription and replication and sometimes when DNA is bound to proteins (6 -10). Most reported DNA crystal structures exhibit either the B-form or A-form helical structure, with a small fraction exhibiting an intermediate structure.
A strong correlation exists between the propensity of a given DNA sequence to adopt a non-B-form structure and the interaction of solvent (water and ions) with the major and minor grooves (5), but neither the natural range of DNA conformations under physiologic solution-phase conditions nor how subtle variations in DNA structure may influence the activity of DNA binding proteins is well understood.
The structure of a particular sequence of dsDNA is often inferred from its pattern of reactivity with structure-specific chemical or enzymatic probes. After treatment with such a probe, the DNA is typically cleaved at reactive sites, and the positions of these cleaved sites are calculated from the fragment lengths observed using gel-based detection methods such as direct labeling, primer extension, and ligation-mediated PCR.
Unlike cleavage-based methods, bisulfite converts singlestranded, unmethylated cytosines into uracils, which, in the context of DNA, can be cloned and sequenced (11)(12)(13)(14)(15). This method is extremely sensitive, highly reproducible, technically simpler to perform, ensures specificity to the sequence of interest, and allows one to examine specific clones and thereby resolve multiple structures. Previous work has demonstrated the utility of bisulfite in analyzing conceptually simple, stable structures such as R-loops and cruciforms (11,12). For these structures, the reasons for the observed pattern of bisulfite reactivity are very clear; they contain completely unpaired cytosines, and those cytosines are the primary ones to react.
However, in genomic regions without demonstrated R-looping or cruciforming ability, the reasons for bisulfite reactivity are less clear, as at the bcl-1 major translocation cluster (MTC) and possibly the bcl-2 major breakpoint region (MBR) (13). Reactivity is largely symmetrical between the two strands, increases with supercoiling, and does not depend on the source of the DNA. Because the structural basis for reactivity in these presumably double-stranded regions is unclear, we performed bisulfite probing on oligonucleotides containing sequences previously characterized by other groups using other methods (16,17). Integration of the genomic in situ and in vitro studies with additional methods as done here (circular dichroism) and elsewhere (x-ray crystallography and NMR) have permitted us to formulate a coherent molecular dynamic view of how regions with consecutive cytosine induce a conformation that is intermediate between B-form and A-form DNA. The presence of this conformation in zones of chromosomal fragility may be relevant to understanding why they are fragile.

EXPERIMENTAL PROCEDURES
Bisulfite Analysis of Duplex Oligonucleotides-As required for bisulfite probing, we lengthened the ends to improve stability at 37°C and lengthened one strand to provide space for PCR primers. As a result, each substrate is composed of a long strand (68 -70 nt) and a short strand (29 -31 nt), but only the long strand can be PCR-amplified and sequenced, and thus only long strand C to U conversions can be detected. The single-stranded arms contain no cytosines to prevent any PCR bias toward lessreacted molecules, and this arrangement also allows us to see reactivity at the ends of the double-stranded portion.
DNA oligos were ordered from Operon Biotechnologies, and oligos containing 2Ј-O-methyl RNA were ordered from IDT. All oligonucleotides were purified using denaturing PAGE. 650 pmol of gel-purified long strand oligo was mixed with 1,300 pmol of gel-purified short strand oligo in 65 l of TE with 100 mM NaCl in a screw-cap tube and annealed by boiling the tube in 1 liter of water for 5 min and then allowing to cool at room temperature overnight. Bisulfite mixture was prepared by mixing 0.5 g of NaHSO 3 (Sigma S-8890) with 0.525 ml of double distilled H 2 O and 0.2626 ml of 2 M NaOH and then mixing 457.5 l of the resulting solution with 12.5 l of 20 mM hydroquinone. 15 l of annealed oligo was mixed with 235 l of bisulfite mixture and put into an air incubator at 37°C for 16 h. After treatment, oligos were serially precipitated with ethanol until the pellet could be dissolved in 20 l of TE, and 5 l was run on 5% native mini-PAGE along with small amounts of component single-stranded oligos and untreated substrate more than one lane away. The gel was stained with 0.25 g/ml ethidium bromide in deionized water for 20 min with light shaking, transferred to a UV transilluminator overlain with clean plastic wrap, and the double-stranded form cut out with a clean razor blade. The gel slice was treated four times with 0.3 M NaOH for 5 min at room temperature, washed four times with TE for 5 min at room temperature, and then crushed and soaked in 200 l of PAGE diffusion buffer (0.5 M ammonium acetate, 10 mM magnesium acetate, 0.1% SDS, 1 mM EDTA, pH 8.0) at 37°C, 250 rpm for Ͼ8 h. The supernatant was precipitated with ethanol and resuspended in 20 l of TE. 0.5 l was used for a 10-l PCR with primers AT131 and AT27 using TaqDNA polymerase (PerkinElmer Life Sciences) using 30 cycles of 94°C for 30s, 55°C for 30s, and 72°C for 30s, with an initial denaturation at 94°C for 2 min and a final extension at 72°C for 2 min. Primers are designed to amplify only the long strand and do not anneal to any cytosines on the long strand, preventing bisulfite conversion PCR bias. PCR products were checked by running on 5% native PAGE and topo TA-cloned (Invitrogen), and individual molecules were sequenced on a Li-Cor IR 2 DNA analyzer using M13 forward primer according to the manufacturers' directions. Molecules with bisulfite conversions at every or all but one cytosine were interpreted as having become single-stranded either before or during bisulfite treatment and were not used for analysis. 16 molecules of data were obtained for each PCR, and two PCRs were done for each bisulfite treatment, and two bisulfite treatments were done for each substrate, giving 64 molecules of data per substrate. Substrates are listed long strand-short strand as follows: GGGCCC, AT160 -AT161; GGCGCC, AT162-AT163; GGCC, AT164 -AT165; GCGC, AT166 -AT167; CGCG, AT168 -AT169; GCGCGC, AT183-AT184; 2Ј-O-methyl-GGGCCC, AT183-AT259; 2Ј-O-methyl-GCGCGC, AT160 -AT264. Oligo sequences are as follows: AT160, gtggggttattgtgggtgtacctgcgttcatgggcccatgcgatccttgaaggaatttggagagaggggt; AT161, tcaaggatcgcatgggcccatgaacgcaggt; AT162, gtggggttattgtgggtgtacctgcgttcatggcgccatgcgatccttgaaggaatttggagagaggggt; AT163, tcaaggatcgcatggcgccatgaacgcaggt; AT164, gtggggttattgtgggtgtacctgcgttcatggccatgcgatccttgaaggaatttggagagaggggt; AT165, tcaaggatcgcatggccatgaacgcaggt; AT166, gtggggttattgtgggtgtacctgcgttcatgcgcatgcgatccttgaaggaatttggagagaggggt; AT167, tcaaggatcgcatgcgcatgaacgcaggt; AT168, gtggggttattgtgggtgtacctgcgttcatcgcgatgcgatccttgaaggaatttggagagaggggt; AT169, tcaaggatcgcatcgcgatgaacgcaggt; AT183, gtggggttattgtgggtgtacctgcgttcatgcgcgcatgcgatccttgaaggaatttggagagaggggt; AT184, tcaaggatcgcatgcgcgcatgaacgcaggt; AT259, tcaaggatcg(caugcgcgcaug) 2Ј-O-methyl RNAaacgcaggt ; AT264, tcaaggatcg(caugggcccaug) 2Ј-O-methyl RNAaacgcaggt ; AT27, acccctctctccaaattcct; and AT131, gtggggttattgtgggtgt.
Molecular Dynamics Simulations-Trajectories of 10 ns were recorded for the 14-bp duplexes (TCATGGGCCCATG-C)⅐d(GCATGGGCCCATGA) and d(TCATGCGCGCATGC)⅐d(GCATGCGCGCATGA), starting from a canonical B-DNA conformation. Each duplex was solvated in a periodic box containing TIP3P water molecules, with the box extending 12 Å from the extremes of the solute, giving about 6000 water molecules. The duplexes were neutralized by addition of 26 sodium ions. Following preliminary minimization, the simulations were run with a time step of 0.002 ps, a residue-based cutoff of 8 Å, the particle-mesh Ewald method for electrostatic interactions, and application of SHAKE to all hydrogen atoms (18). Equilibration of the solvent was performed in an initial simulation of 100 ps with position restraint of the DNA and with a linear temperature increase from 0 to 298 K over the first 50 ps. The restraint on the solute atoms was removed, and a further 100 ps of equilibration was performed at 298 K. This was followed by a 10-ns simulation for data collection, during which structures were collected every 0.2 ps, giving 50,000 structures over the trajectory.
To model bisulfite reactivity, theoretical bisulfite probes (SO 3 with a tetrahedral geometry) were docked against the computed structures at variable distances (2.0 to 3.0 Å, step 0.2 Å) from S (bisulfite) to C-5 of each cytosine, angles (30 -150°, step 10°) S-C-5-C-2, and torsions S-C-5-C-2-N-3 (0 -350°, step 10°). The bisulfite location was considered acceptable if no probe atom was within 2 Å of any DNA atom. The bisulfite accessibility was calculated as the number of different allowed angle and torsion combinations and, essentially, the number of different allowed pathways of approach of the bisulfite to cytosine C-5 to a distance of 3 Å or less.
Calculation of Electrostatic Potentials-Coordinates for (d(CATGGGCCCATG)) 2 and (d(CATGCGCGCATG)) 2 in the canonical B-form helix were generated using NUCGEN in the AMBER8 suite of programs (19). These coordinates were converted to PQR format with AMBER charges using PDB2PQR (20). The electrostatic potential of these structures was calculated using the nonlinear Poisson-Boltzmann equation as implemented in APBS (21). The bisulfite concentration and temperature were those used in the bisulfite probing experiments. The sodium ion was assigned a radius of 2 Å, whereas the bisulfite and sulfite ion radii were assigned 4 Å. Electrostatic potentials were converted into bisulfite concentration maps using the relationship C(x,y,z) ϭ C b e (Ϫ(Z*⌽(x,y,z))) , where C b is the bulk ion concentration, Z is the ion valence, and ⌽(x,y,z) is the electrostatic potential (22). These maps were visualized using the VMD software package using a spherical probe of radius 4 Å (23).
Circular Dichroism Spectra-DNA and RNA oligonucleotides from IDT (Coralville, IA) were used as received. Circular dichroism was performed on 10 M total oligonucleotide concentration (all sequences are palindromic) in a 1:1 mixture of disodium phosphate and monosodium phosphate in 5-mm rectangular cells on a Jasco J-810 at 5°C.

RESULTS AND DISCUSSION
Bisulfite Reactivity at the bcl-1 MTC and bcl-2 MBR-Bisulfite catalyzes the deamination of single-stranded, unmethylated cytosines to uracils (11). First, the bisulfite anion adds to the 5-6 double bond of cytosine, breaking aromaticity and resulting in cytosine sulfonate (24,25). When protonated, this modified cytosine can undergo a rate-limiting hydrolytic deamination to uracil sulfonate. Later, mild base and dilution can eliminate the sulfonate group, yielding uracil. Stacking within duplex DNA is thought to protect cytosines from bisulfite attack, providing the basis for structural specificity (i.e. sensitivity of single-stranded regions) (11)(12)(13). Additionally, the presence of two polyanionic backbones makes the close approach of the bisulfite anion electrostatically unfavorable.
Several chromosomal fragile sites are in close proximity to bisulfite-reactive regions rich in stretches of consecutive cytosines ( Fig.  1 and supplemental Fig. S1) (13). An abnormally high proportion of breakpoints from t(14;18)(q32;q21) translocations occur in a region termed the bcl-2 MBR, and bisulfite reactivity at the MBR was previously reported by us (13). The sum of reactivity on the FIGURE 1. Correlation of bcl-2 MBR and bcl-1 MTC bisulfite reactivity with sequence features. A, bisulfite reactivity of the bcl-2 MBR has been published and is plotted in the top panel (13). The x axis denotes every base position along the 528-bp fragment (which includes A and T in addition to C and G). C3 T and G3 A conversion percentages are plotted, calculated as the number of molecules converted at that particular position, and divided by the total number of molecules sequenced for that strand. Bases in the 2nd panel are aligned with those in the top panel, and plot the C or G "string" length, or the length of consecutive Cs or Gs within which a particular C or G position is located. For instance, ACGGGGCCGGT would have values of 0, 1, 4, 4, 4, 4, 2, 2, 2, 2, 0 for each base position, from left to right. The bottom two graphs show the C, G, CϩG, A, T, and AϩT densities calculated using a moving window of 21 bp. Although there is some correlation between C and CϩG density and bisulfite reactivity, the fine peak structure of reactivity matches best with the string length. B, bisulfite reactivity at the bcl-1 MTC. Reactivity and sequence features for the bcl-1 MTC are plotted similar to A, but using the data from supplemental Fig. S1 for the three bisulfite plots.
top and bottom strands is plotted in Fig. 1A, top graph. One can see the correlation with consecutive Cs (C-string length) in the second plot, and this correlation is better than with any other sequence composition (lower two plots).
Because breakpoints also occur frequently in the bcl-1 MTC from t(11;14)(q13;q32) translocations, we decided to probe this region for bisulfite as well (26). Bisulfite reactivity occurs primarily at the downstream edge of the bcl-1 MTC (Fig. 1B, top plot, and supplemental Fig. S1, all plots). Two peaks of reactivity are observed. The first and smaller peak occurs at the semi-palindromic sequence cgaggggaagcccctcc. The second, broader, and larger peak occurs at the sequence ccctctaagccccctctccccgtcacatccccccgaccctgcc. Note the stretches of consecutive Cs within both sequences.
The pattern of reactivity is very similar between the different substrates, extracted genomic DNA, supercoiled plasmid, or linearized plasmid. However, the overall level of bisulfite reactivity is higher on supercoiled plasmid than on linearized plasmid or genomic DNA (supplemental Fig. S1). Supercoiling destabilizes the duplex, accelerating DNA breathing and allowing it to more easily form non-B-DNA structures.
Because the top and bottom strands of the original bisulfiteprobed DNA molecule are separated during PCR and separately cloned, each sequence obtained gives information for deamination events on only one of the two original strands. When the reactivity is rescaled to balance the top and bottom strand representation, it is roughly symmetrical between the two strands (Fig. 1B, top plot; compare with supplemental Fig. S1, which displays top strand in black and bottom strand in red).
To rule out our primers biasing against binding to converted regions, we used a second set of outer primers to confirm that few conversions occur at the binding sites of the first set. Another strategy, designing two sets of primers with one set insensitive to top strand conversions and the other insensitive to those on the bottom, was also performed. These show no differences in pattern (data not shown).
The pattern does not change if bisulfite treatment is done on intact Reh cells embedded in 1% low melt agarose or intact Escherichia coli embedded in agarose after transformation with plasmid containing the bcl-1 sequence. Any proteins that might bind to the DNA are likely denatured by the high salt concentration of the bisulfite solution. The pattern also does not change if the bisulfite solution is adjusted to pH 6, although overall reactivity is decreased about 3-fold, consistent with the 3-fold lower efficiency of the bisulfite-catalyzed deamination reaction itself at pH 6 compared with pH 5.2 (25). Reactivity increases proportionately if DNA is treated for 24 h rather than the usual 16 h, but again the pattern of reactivity is still the same. Neither does it change if the   ). Therefore, the specific pattern of bisulfite reactivity appears dependent primarily upon the DNA sequence. Analysis of Defined DNA Duplexes Using Bisulfite-To determine the basis for the full range of reactivity of bisulfite with DNA, we probed several well characterized and structurally divergent oligonucleotide duplexes with bisulfite as follows: d(…catGGGCCCatg…) 2 , d(…catGGCGCCatg…) 2 , and d(…catGCGCGCatg…) 2 ( Fig. 2 and tabulated in Table 1, top 3 rows). Consistent with its single strand specificity, bisulfite virtually always detected the frayed edges of the duplex. At internal portions of the duplex, typically none or only one of the three central Cs in any given molecule reacted with bisulfite (supplemental Fig. S2). Interestingly, we found that the GGGCCC duplex was 5-fold more reactive than GCGCGC, with GGCGCC exhibiting intermediate reactivity.
A similar but less dramatic effect was observed for GGCC over GCGC or CGCG (Fig. 3 and Table 1, middle rows).
Clearly, bisulfite detects one or more subtle structural differences between these duplexes. Circular dichroism showed the series d(catGCGCGCatg) 2 (B-form DNA), d(catGGCGCCatg) 2 , d(catGGGCCCatg) 2 , and r(cauGCGCG-Caug) 2 (A-form RNA) exhibits increasing A-form character (Fig. 4). This effect was present at two markedly different salt concentrations (supplemental Fig. S3). Consistent with these data, a B/A-intermediate structure of d(catGGGCCCatg) 2 (PDB code 1DC0) was previously observed by x-ray crystallography (16). The bisulfite reactivity of a corresponding A-form DNA/2Ј-O-methyl RNA heteroduplex was lower than that for B-form DNA (Fig. 5 and Table 1 2 and d(catGCGCGCatg) 2 sequences using AMBER8 (19). The structures generated in this simulation were  subjected to theoretical bisulfite probing using a dedicated inhouse algorithm. Details of the simulations and the probing algorithm are given under the "Experimental Procedures" and supplemental material.
Helical parameters of the central d(GGGCCC) 2 region, averaged over the simulation, indicate that this sequence adopts a B/A-intermediate conformation between that of the 1DC0 x-ray structure and canonical B-form DNA (Table 2 and Fig. 6) (27). Theoretical bisulfite probing of the d(GGGCCC) 2 region in four different conformations showed that the C-5 atoms of cytosines in the average MD structure are more accessible to bisulfite than those in the 1DC0 x-ray structure and canonical B-and A-DNA (supplemental Table S1). This may be due to the similarity of the shift in the MD structure to that of B-DNA (Table 2), which indicates that the cytosine bases are less buried in the major groove (Fig. 6), in conjunction with a positive roll similar to that of A-DNA, which may orient the C-5 atoms more favorably for approach of the bisulfite anion from the 3Ј face of cytosine (supplemental Fig. S4). The positive roll angle in the MD structure is also associated with widening of the minor groove, and in this context we note that hydroxyl radicals also have enhanced reactivity at consecutive cytosines (28). These radicals are thought to react more readily when the minor groove is more accessible, consistent with increased A-form character at consecutive cytosines. Interestingly, we observe the least accessibility of cytosine C-5 atoms by bisulfite in A-form DNA (Fig. 6).
Given that the sequence-specific electrostatic potential of the major groove may also enhance the bisulfite reactivity of consecutive cytosine bases, we used the nonlinear Poisson-Boltzmann equation (22) to estimate the local occupancy of the major groove with bisulfite for the duplexes d(catGGGC-CCatg) 2 and d(catGCGCGCatg) 2 in the canonical B-form helix (see supplemental "Results" and "Experimental Procedures"). We found a contiguous electropositive "patch" in the major groove of d(catGGGCCCatg) 2 not present in d(catGCGCG-Catg) 2 (Fig. 7). Such electropositive patches, composed of consecutive cytosines, could localize the bisulfite anion and thereby increase its reactivity with those cytosines. This proposed sequence-specific localization of bisulfite ion at an electropositive region of DNA is analogous to the well documented sequence-specific localization of cations along electronegative regions of DNA formed by consecutive guanines (in the major groove) or A/T base pairs (in the minor groove) (5).
The sequence-dependent dynamics of base pairing may also contribute to the sequence-dependent bisulfite reactivity reported here for duplex DNA. Imino proton exchange NMR shows that the central 6 bp in d(catGGGCCCatg) 2 are overall more dynamic (i.e. prone to base pair opening) than those in d(catGGCGCCatg) 2 , and the central four in d(catGGCCatg) 2 are more unpaired than those in d(catGCGCatg) 2 or d(catCGCGatg) 2 (17). Dornberger et al. (17) have suggested that sequences with consecutive cytosines adopt a variation of the B-form structure that causes the base pairs in these sequences, when flanked by canonical B-form helices, to breathe more than sequences alternating between guanines and   cytosines. This conclusion is consistent with our observation that d(catGGGCCCatg) 2 exhibits a CD spectrum that is indicative of a B/A-form structural intermediate, as this CD could result from the GGGCCC region adopting a more A-form structure, whereas the flanking 3 bp on each side maintain a B-form helix. Finally, supercoiling is known to destabilize duplex DNA and can induce structures that underwind DNA to mitigate the supercoiling. With bcl-2 MBR and bcl-1 MTC sequences, supercoiling enhances overall reactivity, including that at stretches of consecutive cytosines ( Fig. 1 and supplemental  Fig. S1). 6 Concluding Comments-We have previously shown that the bcl-2 MBR has a relatively high degree of bisulfite reactivity, which correlates with its recombination efficiency in vivo (as measured using minichromosomal substrates and recapitulates many aspects of the bcl-2 translocation, t(14;18) (13). A 3-base mutation disrupting a set of consecutive cytosines reduced both the bisulfite reactivity and the recombination efficiency, and these recombination events were also shown to be dependent on the RAG endonuclease (13). Here, we report that strings of consecutive cytosines are a prominent feature at the bcl-1 MTC, and we find similar features at the c-myc fragile region at its intron 1-exon 1 boundary (data not shown).
This study examines the precise structural conformations reflected by bisulfite reactivity. Bisulfite appears to be exquisitely sensitive to very subtle perturbations of DNA structure that increase the steric accessibility of cytosine and alter the electrostatic potential of the major groove. Although some effects on bisulfite reactivity were expected based on duplex DNA end fraying and supercoiling, the dramatic effect observed simply by switching 2 bp, from GGGCCC to GCGCGC, was entirely unexpected and is indicative of significant sequence-dependent structural alterations. Bisulfite reactivity at consecutive cytosines appears to be a consistent feature regardless of whether the DNA segment is located within large duplexes (e.g. plasmids and genomic DNA) or short oligonucleotide substrates.
In light of these results, we suggest the possibility that consecutive cytosines induce structural and electrostatic changes that predispose to nucleases, such as the RAG complex, which is essential for recombination in vivo. Although the precise structural mechanisms remain speculative, both the RAG complex and the Artemis-DNA-PK CS complex are known to cleave at bubble structures (single-stranded regions) as small as 1-3 nt (29 -31). 6 A study of c-myc chromosomal translocations in sporadic Burkitt lymphomas, known to depend on a singlestranded cytosine deaminase, AID, identified a consensus motif of 5Ј-CTCCTCCCC-3Ј for translocation breakpoints (32), consistent with translocation zones being GC-rich and rich in consecutive cytosines (33). To the extent that bisulfite reactivity reflects a tendency toward B/A-intermediate structure or single-stranded character, bisulfite-reactive regions with stretches of consecutive cytosines, such as these translocation cluster regions, appear to be targets for such DNA double strand break mechanisms.  Table 2 shows that this structure has parameters that are similar to those averaged over 10 ns. All structures are shown as 12-bp duplexes (for the simulated structure, the terminal base pairs were deleted). The central region is numbered with reference to the six cytosine bases, as d(G 6 G 5 G 4 C 1 C 2 C 3 )⅐d(G 3 G 2 G 1 C 4 C 5 C 6 ).