A K cation-induced conformational switch within a loop spanning segment of a DNA quadruplex containing G-G-G-C repeats1

Article No. mb982031

J. Mol. Biol. (1998) 282, 637±652

A K Cation-induced Conformational Switch Within a Loop Spanning Segment of a DNA Quadruplex Containing G-G-G-C Repeats Serge Bouaziz, Abdelali Kettani and Dinshaw J. Patel* Cellular Biochemistry and Biophysics Program, Memorial Sloan-Kettering Cancer Center New York, NY 10021, USA

We have identi®ed a unique structural transition (in slow exchange on the NMR time scale) in the tertiary fold of the d(G-G-G-C-T4-G-G-G-C) quadruplex on proceeding from Na‡ to K‡ as counterion in aqueous solution. Both monovalent cation-dependent conformations exhibit certain common structural features, which include head-to-tail dimerization of two symmetry-related stem-hairpin loops, adjacent strands which are antiparallel to each other and adjacent stacked G(syn)
G(anti)
G(syn)
G(anti) tetrads in the central core of the quadruplexes. The Na and K cation stabilized structures of the d(G-G-G-C-T4-G-G-G-C) quadruplexes differ in the conformations of the T-T-T-T loops, the relative alignment of G
C base-pairs positioned opposite each other through their major groove edges and potentially in the number of monovalent cation binding sites. We have identi®ed potential K cation binding cavities within the symmetry-related T-T-T-G segments, suggesting the potential for two additional monovalent cation binding sites in the K cation-stabilized quadruplex relative to its Na cation-stabilized counterpart. Modeling studies suggest that the major groove edges of guanine residues in Watson-Crick G
C base-pairs could potentially be bridged by coordinated K cations in the d(G-G-G-C-T4-G-G-G-C) quadruplex in KCl solution in contrast to formation of G
C
G
C tetrads for the corresponding quadruplex in NaCl solution. Our results de®ning the molecular basis of a Na to K cation-dependent conformational switch in the loop spanning segment of the d(G-G-G-C-T4-G-G-G-C) quadruplex may have relevance to recent observations that speci®c K cation coordinated loop conformations within quadruplexes exhibit inhibitory activity against HIV integrase. # 1998 Academic Press

*Corresponding author

Keywords: G-G-G-C repeat-containing DNA quadruplexes; monovalent cation-dependent conformational transitions; potential K cation coordination sites; potential K cation encapsulation within a hairpin loop; DNA quadruplexes targeted to HIV integrase

Introduction A unique feature of G
G
G
G tetrad containing DNA quadruplexes (Gellert et al., 1962; reviewed by Guschlbauer et al., 1990; Rhodes & Giraldo, 1995; Pilch et al., 1995; Patel et al., 1999) is the Abbreviations used: NOE, nuclear Overhauser enhancement; NOESY, NOE spectroscopy; ppm, parts per million; COSY, correlated spectroscopy; TOCSY, total COSY; rmsd, root-mean-square deviation. E-mail address of the corresponding author: [email protected] 0022±2836/98/380637±16 $30.00/0

absolute requirement for monovalent cations (Pinnavaia et al., 1978; Howard & Miles, 1982; Xu et al., 1993). There is speci®city associated with the binding, since G-quadruplex formation is facilitated by Na and K cations (Howard & Miles, 1982; Sundquist & Klug, 1989; Williamson et al., 1989; Miura et al., 1995) but inhibited by the smaller Li cation (Williamson et al., 1989). The monovalent cations were initially proposed to bind between adjacent G-tetrad planes (Sundquist & Klug, 1989) with coordination to four inwardly pointing guanine oxygen atoms from each G-tetrad plane. This Ê model has received strong support from a 2.5 A # 1998 Academic Press

638 resolution crystallographic study of a K cation sandwiched between G-tetrad planes in the Gquadruplex involving antiparallel alignment of adjacent strands formed by dimerization of Ê resd(G4T4G4) (Kang et al., 1992) and from a 0.9 A olution crystallographic study of Na cations sandwiched between G-tetrad planes in the Gquadruplex involving parallel alignment of adjacent strands formed by tetramerization of d(TG4T) (Phillips et al., 1997). It has been known for some time that more stable G-quadruplexes form in the presence of K cation relative to Na cation (reviewed by Williamson, 1993) and recent studies suggest that this re¯ects the differences in the free energies of monovalent cation dehydration (Hud et al., 1996). The NMR proton chemical shift parameters of the d(G4T4G4) (Schultze et al., 1994) and d(G3T4G3) (Hud et al., 1996) exhibit small differences in Na versus K cation solution but these do not translate into distinct monovalent cation-dependent conformational transitions within the G-quadruplexes. Other studies have suggested the existence of monovalent cationspeci®c conformational transitions between Gquadruplexes and their duplex components (Hardin et al., 1991, 1992) but the structural basis of these transitions at atomic resolution remains unde®ned at the current time. We outline below the molecular basis of a striking structural transition observed for the d(G-G-GC-T4-G-G-G-C) quadruplex on proceeding from Na to K cations in aqueous solution. This system is of importance, since G-G-G-C sequences, either in isolation or in tandem arrays, have been observed in adeno-associated viral DNA and its site-speci®c integration site on chromosome 19 (reviewed by Berns & Linden, 1995). Our structural studies of the d(G-G-G-C-T4-G-G-G-C) quadruplex in K cation solution provide a model for K cation encapsulation within a hairpin loop fold of a quadruplex. Further, the determination of the solution structures of both the Na cation (Kettani et al., 1998, accompanying paper) and K cation (this study) stabilized conformations of the d(G-G-G-C-T4-G-GG-C) quadruplex de®nes for the ®rst time the location, nature and extent of conformational changes at individual residues within the quadruplex and attached loops associated with the monovalent cation-dependent conformational switch. A related structural transition has also been observed for an intra-molecularly folded G-quadruplex which selectively targets HIV-1 integrase in K cation containing aqueous solution (Rando et al., 1995; Bishop et al., 1996; Mazumder et al., 1996; Jing et al., 1997a). No structure has been reported to date for this K cation folded G-quadruplex in either the free or the HIV-1 integrase bound state. Proton chemical shift data suggest the existence of a K cation-induced loop conformational transition within the intra-molecularly folded G-quadruplex (Jing et al., 1997b).

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Results NMR spectra The exchangeable proton (7.0 to 14.0 ppm) NMR spectrum of the d(G1-G2-G3-C4-T4-G9-G10-G11C12) sequence in 100 mM KCl, 2 mM phosphate, H2O at pH 6.6 and 10 C is plotted in Figure 1(b). The spectrum establishes formation of one predominant conformer for the 12-mer in KCl containing solution. A comparison of the exchangeable (8.5 to 14.5 ppm) and non-exchangeable proton (6.5 to 8.5 ppm) spectral regions for the d(G-G-G-C-T4-GG-G-C) in 100 mM NaCl (Figure 1(a)) and 100 mM KCl (Figure 1(b)) establishes distinct chemical shifts for the same protons in Na versus K cation containing solutions. This strongly implies a monovalent cation-dependent conformational switch in the folded structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex on proceeding from Na to K cations in aqueous solution. Exchangeable protons The imino and amino protons are well resolved in the exchangeable proton spectrum of the d(G-GG-C-T4-G-G-G-C) sequence in 100 mM KCl containing H2O buffer (pH 6.6) at 10 C (Figure 1(b)). Two guanine imino protons participating in Watson-Crick pairing resonate at 13.54 and 13.78 ppm while four others participating in G
G
G
G-tetrad formation resonate between 11.0 and 11.8 ppm (Figure 1(b)). The imino protons of the four thymine residues in the sequence resonate between 10.4 and 11.0 ppm. We observe two cytosine amino protons at 8.67 and 8.71 ppm (Figure 1(b)), implying that one amino proton is hydrogenbonded for each of cytosine C4 and C12 in the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) sequence. We also observe four amino protons between 9.4 and 10.4 ppm, which originate in the hydrogen-bonded amino protons of guanine residues involved in Gtetrad formation. These latter protons are observed for the d(G-G-G-C-T4-G-G-G-C) sequence in 100 mM KCl (Figure 1(b)) but are signi®cantly broadened for the same 12-mer in 100 mM NaCl (Figure 1(a)) solution. The imino protons of G1 and G10 have been assigned from the difference spectrum of 15N decoupled and 15N coupled spectra of the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) sequence in 100 mM KCl containing H2O buffer containing 15 N-labeled site speci®cally at the N1, N2 and N7 positions of guanine G1 and G10. The labeling approach has allowed us to de®nitively identify one (G1 at 13.54 ppm) of the two (G1 and G9) imino protons involved in Watson-Crick G
C basepair formation and one (G10 at 11.15 ppm) of the four (G2, G3, G10 and G11) imino protons involved in G-tetrad formation. The availability of the imino proton assignments of G1 and G10 then permitted the assignment of the expanded NOESY (200 ms mixing time) con-

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

639

Figure 1. Proton NMR spectra (6.6 to 14.3 ppm) of the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) quadruplex (4 mM in strand concentration) in (a) 100 mM NaCl and (b) 100 mM KCl containing H2O buffer (2 mM phosphate) at pH 6.6 and 10 C. Imino and amino proton assignments are listed over the spectra.

tour plot (Figure 2(a)) correlating NOEs between imino protons and amino, base and sugar protons in the d(G-G-G-C-T4-G-G-G-C) sequence in 100 mM KCl containing H2O buffer (pH 6.6) at 10 C. The observed NOE cross-peak assignments are listed in the legend of Figure 2(a). The observed NOEs between the imino protons of G2, G3, G10 and G11 to the H-8 protons of G3, G10, G11 and G2, respectively (peaks j, m, i and n, respectively, Figure 2(a)) establish both the formation of a G2
G3
G10
G11 tetrad and the direction of donor to acceptor hydrogen-bonding directionality. The formation of G-tetrads and the observation of a single set of resonances for the predominant conformer of the d(G-G-G-C-T4-G-G-G-C) sequence in KCl solution (Figure 1(b)) requires that this dodecamer sequence form a 2-fold symmetric quadruplex through head-to-tail dimerization of 12-mer hairpins as shown schematically in Figure 3(b). The observed NOEs between the imino proton of G1 and the amino protons of C12 (peaks f and f0 , Figure 2(a)) and between the imino proton of G9 and the amino protons of C4 (peaks e and e0 , Figure 2(a)) establish formation of Watson-Crick G1
C12 and G9
C4 base-pairs in the d(G-G-G-C-

T4-G-G-G-C) quadruplex in 100 mM KCl solution. Two lines of evidence rule out formation of G1
C4
G9
C12 tetrads for the d(G-G-G-C-T4-G-GG-C) quadruplex in 100 mM KCl solution as was reported in the accompanying paper for this quadruplex in 100 mM NaCl solution (Kettani et al., 1998). First, the cytosine amino protons of C4 (8.67 and 6.98 ppm) and C12 (8.71 and 7.51 ppm) exhibit chemical shifts that are characteristic of one amino proton being hydrogen-bonded and the other being exposed for each cytosine, while both amino protons of each cytosine would be expected to be hydrogen-bonded on G
C
G
C tetrad formation (Kettani et al., 1995, 1998). Secondly, NOEs between the G1(H-8) and C4(H-5) protons and between the G9(H-8) and C12(H-5) protons that de®ne the G1
C4
G9
C12 tetrad alignment for the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM NaCl solution (accompanying paper, Kettani et al., 1998) were not observed for the same quadruplex in 100 mM KCl solution (this study). The imino and amino proton chemical shifts for the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution are listed in Table 1. Each guanine participating in G2
G3
G10
G11 tetrad

640

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Figure 2. (a) An expanded NOESY (200 ms mixing time) contour plot correlating NOEs between imino protons and imino, amino and base protons in the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) quadruplex (4 mM in strand concentration) in 100 mM KCl, 2 mM phosphate, H2O at pH 6.6 and 10 C. The cross-peaks a to z and a to e are assigned as follows: a and a0 , G3(NH-1)-G3(NH2-2); b and b0 , G11(NH-1)-G11(NH2-2); c and c0 , G10(NH-1)-G10(NH2-2); d and d0 , G2(NH-1)-G2(NH2-2); e and e0 , G9(NH-1)-C4(NH2-4); f and f0 , G1(NH-1)-C12(NH2-4); g, G11(NH-1)-C12(NH2-4); h, G3(NH-1)-C4(NH2-4); i, G10(NH-1)-G11(H-8); j, G2(NH-1)-G3(H-8); k, G10(NH-1)-G9(H-8); l, G2(NH-1)-G1(H-8); m, G3(NH-1)-G10(H-8); n, G11(NH-1)-G2(H-8); o, G10(NH2-2)-G11(H-8); p, G2(NH2-2)-G3(H-8); q, G10(NH2-2)-G9(H8); r, G2(NH2-2)-G1(H-8); s, G3(NH2-2)-G10(H-8); t, G11(NH2-2)-G2(H-8); u, C12(NH2-4)-C12(H-5); v, C4(NH2-4)C4(H-5); w, G10(NH-1)-G9(H-200 ); w0 , G10(NH2-2)-G9(H-200 ); x, G10(NH-1)-G9(H-20 ); x0 , G10(NH2-2)-G9(H-20 ); y, G2(NH-1)-G1(H-200 ); y0 , G2(NH2-2)-G1(H-200 ); z, G2(NH-1)-G1(H-20 ); z0 , G2(NH2-2)-G1(H-20 ); a, C12(NH2-4)T8(H-10 ); b, C12(NH2-4)-T8(H-30 ); w, C12(NH2-4)-T8(H-40 ); d, C12(NH2-4)-T8(H-200 ); e, C12(NH2-4)-T8(H-20 ). (b) An expanded NOESY (250 ms mixing time) contour plot correlating NOEs between base and sugar H-10 protons in the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) quadruplex (4 mM in strand concentration) in 100 mM KCl, 2 mM phosphate, 2 H2O at pH 6.6 and 20 C. The lines trace the NOE connectivities between the base protons and their own and 50 -¯anking sugar H-10 protons from G1 to C12. The cytidine H-6 ± H-5 NOEs are designated by asterisks. The NOE cross-peaks a to b are assigned as follows: a, G11(H-8)-C12(H-5); b, G3(H-8)-C4(H-5); c, G2(H-8)-G3(H-10 ); d, G10(H-8)-G11(H-10 ). (It should be noted that we also observe additional cross-peaks in this expanded NOESY contour plot, which are unassigned at this time. They are likely to originate in minor conformer(s) of the quadruplex and/or the single-stranded form of the d(G-G-G-C-T4-G-G-G-C) sequence in KCl containing solution. The strongest of these additional cross-peaks are likely to correspond to the strong NOE between H-5 and H-6 protons of cytosine residues in these minor conformers).

formation exhibits distinct chemical shifts for hydrogen-bonded and exposed amino protons with a separation ranging from 2.5 ppm for G2 and G10 to 4.0 ppm for G3 and G11. Non-exchangeable protons The non-exchangeable base and sugar protons in the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution have been assigned by following the sequential NOE connectivities in NOESY spectra in 2H2O buffer at pH 6.6 and 20 C

with the tracing supplemented by analysis of the corresponding COSY and TOCSY data sets. The observed NOEs between the base protons to their own and 50 -¯anking sugar H-10 protons are traced from G1 to C12 in an expanded NOESY (250 ms mixing time) contour plot of the d(G-G-G-C-T4-GG-G-C) quadruplex in 100 mM KCl cation solution (pH 6.6) at 20 C in Figure 2(b). The assignments of the H-8 protons of G1 and G10 have been veri®ed independently following analysis of NMR difference spectra (with and without 15N decoupling) on

641

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Figure 3. The folding topology of the d(G1-G2-G3-C4-T4-G9-G10-G11C12) quadruplex in (a) 100 mM NaCl and (b) 100 mM KCl solution. The backbone tracing of the individual hairpins is shown by thick lines and the chain directionality by thick arrows. Hydrogen bonding donor to acceptor directionalities around individual G
G
G
G tetrads are represented by arrows. syn guanine residues (labeled s) are hatched to distinguish them from anti guanine residues (labeled a). The two distinct inter-strand narrow grooves are labeled N1 and N2 while the symmetry-related intra-strand wide grooves are labeled W.

the d(G-G-G-C-T4-G-G-G-C) sequence in 100 mM KCl solution containing 15N-labeled N1, N2, N7guanine residues G1 and G10 in 100 mM KCl solution. We detect strong NOEs between the base and their own sugar H-10 protons for G2 and G10 in a short mixing time (50 ms) stacked NOESY contour plot establishing formation of syn glycosidic torsion angles (Patel et al., 1982) for G2 and G10 around the G2
G3
G10
G11 tetrads in the d(G-GG-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution. The analysis of the entire NOESY and TOCSY data sets has permitted the identi®cation of the non-exchangeable base and sugar proton assignments of the d(G-G-G-C-T4-G-G-G-C) quadruplex in KCl solution with the chemical shifts at 20 C summarized in Table 1.

NOEs between G
G
G
G tetrad and adjacent G
C base-pairs We observe a set of NOEs between protons on the G2
G3
G10
G11 tetrads and protons on the ¯anking Watson-Crick G1
C12 and G9
C4 basepairs that provide critical restraints for de®ning the relative alignments of these elements in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution. These include NOEs between the G3 and G9 imino protons and between the G1 and G11 imino protons, as well as NOEs between the G1 and G3 H-8 protons and between the G9 and G11 H-8 protons (data not shown) that de®ne the relative alignment of these non-adjacent residues in the sequence of the quadruplex. In addition, NOEs are detected between

Table 1. Proton chemical shifts (ppm) for the d(G1-G2-G3-C4-T5-T6-T7-T8-G9-G10-G11-C12) quadruplex in 100 mM KCl, aqueous buffer NH-1 G1 G2 G3 C4 T5 T6 T7 T8 G9 G10 G11 C12

13.54 11.07 11.69

13.78 11.15 11.69

NH2-2/-4 9.58/7.13 10.24/5.94 8.67/6.98

9.71/7.18 10.20/6.43 8.71/7.51

H-8/H-6 7.73 7.50 8.18 7.77 7.75 7.75 7.46 7.12 7.82 7.66 8.15 7.84

H5/CH3-5

5.65 1.96 1.92 1.71 1.13

5.76

H-10

H-20

H-200

H-30

H-40

6.01 5.96 5.71 6.46 6.29 6.14 5.76 6.14 6.02 5.98 5.82 6.59

2.71 3.25 2.47 2.34 2.26 2.21 1.94 1.97 2.66 3.30 2.52 2.37

3.21 2.89 2.68 2.52 2.51 2.43 2.16 2.25 3.27 2.95 2.71 2.37

4.97 5.07 5.04 5.00 4.93 4.65 4.60 4.40 4.97 5.08 5.04 4.71

4.22 4.45 4.50 4.46 4.36 4.23 3.78 3.86 4.30 4.40 4.48 4.21

Exchangeable proton chemical shifts at 10 C. Non-exchangeable proton chemical shifts at 20 C.

642 the imino proton of G10 and the H-8 proton of G9 (peak k, Figure 2(a)) and between the imino proton of G2 and the H-8 proton of G1 (peak l, Figure 2(a)) that de®nes the relative alignment of these adjacent residues in the sequence of the quadruplex. Further, the imino and hydrogen-bonded amino protons of G2 and G10 exhibit NOEs to the sugar H-20 and H-200 protons of their preceding G1 (peaks y, z, y0 and z0 , Figure 2(a)) and G9 (peaks w, x, w0 and x0 , Figure 2(a)) residues, respectively. NOEs between T-T-T-T loop segment and adjacent G
C base-pairs We have identi®ed a large set of NOEs that are critical towards de®ning the alignment of the T8 residue in the T5-T6-T7-T8 loop relative to the Watson-Crick G1
C12 base-pair in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution. These include NOEs between the T8 methyl protons and the imino and H-8 protons of G1 (data not shown), between the T8(H-10 ) and C12(H-5) protons (data not shown), and between the T8(H-10 , H-20 , H-200 , H-30 and H40 ) protons and the hydrogen-bonded amino proton of C12 (peaks a, e, d, b and g, respectively, Figure 2(a)). Further, the observation of NOEs between protons on the T8 residue and protons on the G1
C12 base-pair is consistent with formation of the d(G-G-G-C-T4-G-G-G-C) quadruplex through a head-to-tail alignment (Figure 3(b)) in KCl solution. The symmetry-related T5-T6-T7-T8 loops can be either of the lateral or diagonal type in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution. The observed pattern of G2(syn)
G3(anti)
G10(syn)
G11(anti) alignment around the G-tetrads (Figure 3(b)) is indicative of antiparallel alignment of adjacent strands around the quadruplex. This result and the formation of G1
C12 base-pairs can only be consistent with formation of lateral loops in this quadruplex. Input restraints and structure calculations The input distance restraints were deduced from measurement of NOE intensities in NOESY data sets of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl containing H2O (two mixing times) and 2H2O (®ve mixing times) solution. The input restraints statistics are listed in Table 2 with the restraints distribution by residue plotted in Figure 4(a) and (b). The structure calculations started from metric matrix distance geometry computations followed initially by distance restrained and subsequently by intensity restrained molecular dynamics computations. All experimentally observed distance restraints were speci®ed ambiguously with ``SUM'' averaging during the computations. The re®nement protocols are discussed brie¯y in Materials and Methods, with a detailed description available in the accompanying paper (Kettani et al., 1998). The quality of the struc-

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex Table 2. NMR restraints statistics for d(G-G-G-C-T-TT-T-G-G-G-C) quadruplex structures in 100 mM KCl solution A. Stage 1: Distance geometry and simulated annealing Non-exchangeable protons Number of NOE derived distance 532 restraints Intra-residue 360 Sequential (i, i ‡ 1) 144 Long range 5(i, i ‡ 2) 28 B. Stage 2: Distance restrained molecular dynamicsa Number of NOE derived distance restraints on non-exchangeable protons Intra-residue Sequential (i, i ‡ 1) Long range 5(i, i ‡ 2) Number of NOE derived distance restraints on exchangeable protons Intra-residue Sequential (i, i ‡ 1) Long range 5(i, i ‡ 2)

266 180 72 14 57 0 28 29

C. Stage 3: Relaxation matrix based NOE intensity refinement 1330 Number of NOE intensity restraints on non-exchangeable protonsb 57 Number of NOE derived distance restraints on exchangeable protonsa D. Additional restraints for all stages Hydrogen bond distance restraints Non-crystallographic symmetry restraints

56 On all heavy atoms

a NOE distance restraints were ``SUM'' averaged during these stages, with the number of monomers set to 2. b NOE intensity restraints were speci®ed ambiguously with multiple atom selection. The number is for ®ve NOESY data sets collected as a function of mixing time.

tures as monitored by pairwise root-mean-squared deviation (rmsd) values and R factors of the d(GG-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution following each stage of the re®nement are listed in Table 3. The ®nal intensity re®ned structures of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution exhibit low R factors and no Ê (Table 3). distance violations 50.2 A Structure analysis The pairwise rmsd values at individual positions between the eight intensity re®ned structures of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution are plotted in Figure 4(c). Stereo views of the eight intensity re®ned structures of the quadruplex looking normal to the helix axis and looking down the helix axis (excludes loop segments) are shown in Figure 5(a) and (b), respectively. The conformation of the T5-T6-T7-T8 loop bridging the G1
C12 and G9
C4 base-pairs together with the alignment of the Watson-Crick G1
C12 and G9
C4 base-pairs along their major groove edges in a representative intensity re®ned structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl are plotted in Figure 6(c) and (d), respectively. These alignments can be compared

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

643 with the conformation of the T5-T6-T7-T8 loop together with the alignment of the G1
C4
G9
C12 tetrad in a representative intensity re®ned structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM NaCl as shown in Figure 6(a) and (b), respectively. The stacking pattern between adjacent symmetry-related G2
G3
G10
G11 tetrads in a representative intensity re®ned structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution is shown as a stereo pair in Figure 7(a). The corresponding stacking pattern between the G2
G3
G10
G11 tetrad and adjacent Watson-Crick G1
C12 and G9
C4 basepairs is shown in stereo in Figure 7(b). There are two distinct narrow grooves (designated N1 and N2) and two symmetrical wide (W) grooves in the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution as labeled in Figures 7(a) and (b).

Models of the quadruplex that include bound K cations

Figure 4. Distribution of restraints by residue in the structure calculations of the d(G1-G2-G3-C4-T5-T6-T7T8-G9-G10-G11-C12) quadruplex formed in 100 mM KCl containing solution. (a) The restraints have been classi®ed as follows: non-exchangeable proton restraints (®lled rectangles), exchangeable proton restraints (open rectangles) and hydrogen bond restraints (hatched rectangles). Each intra-residue restraint is counted as one for that residue. Each inter-residue restraint is counted as half for that residue and half for its partner in the interaction. (b) Distribution by residue position of intraresidue (hatched rectangles) and inter-residue (®lled rectangles) experimentally observed exchangeable proton and non-exchangeable proton restraints. (c) rmsd values Ê ) in the atomic coordinate positions calculated by resi(A due for the eight relaxation matrix re®ned structures and the average of these structures. The average values (®lled circles) are shown along with the standard deviation (vertical bars).

We have used the GRASP program (Nicholls et al., 1991) to identify ®ve cavities that can potentially accommodate K cations within the re®ned structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution. One of these potential K cation-accommodating cavities was located in the center of the quadruplex between G-tetrad planes, one potential cavity each (related by symmetry) was located between the G-tetrad planes and the pair of Watson-Crick G
C base-pairs aligned along their major groove edges and one potential cavity each (related by symmetry) within the T6-T7-T8-G9 segments positioned towards either end of the quadruplex. An averaged structure was computed from the eight distance-re®ned structures of the d(G-G-G-CT4-G-G-G-C) quadruplex in 100 mM KCl without incorporation of ions, and this averaged structure subjected to minimization and dynamics protocols with all restraints. Two views (similar in orientation to those in Figure 5) of this averaged minimized structure are shown in Figure 8(a). A nonhydrated K cation was next inserted into each of the ®ve cavities in the averaged non-minimized structure and molecular dynamics re®nements undertaken with the full set of distance restraints together with additional restraints de®ning K‡ to coordinating oxygen and K‡ to K‡ distances (with bounds of 20%). The protocols for the re®nement computations are outlined in Materials and Methods. Two views of the averaged minimized model of the d(G-G-G-C-T4-G-G-G-C) quadruplex with ®ve bound K cations are shown in Figure 8(b). The coordinating oxygen atoms are shown as red balls in Figure 8(a) and (b) while the bound K cations are shown as yellow balls in Figure 8(b). Two views of the model of the K cation encapsulated within each of the symmetry-related T6-T7-

644

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Table 3. Quality of the d(G-G-G-C-T-T-T-T-G-G-G-C) quadruplex structures in 100 mM KCl solution Parameter

DG and SA structures

Distance restrained MD structures

Relaxation matrix MD structures

A. Convergence Ê) Pairwise rmsd values (A For all residues For residues 1 to 4, 9 to 12 Ê) Average rmsd values (A For all residues For residues 1 to 4, 9 to 12

1.74  0.36 1.59  0.44

1.46  0.37 1.22  0.17

1.22  0.19 1.07  0.21

1.16  0.21 1.06  0.27

1.15  0.21 1.05  0.27

0.81  0.12 0.64  0.11

B. R-factors R1/6 Unweighted R1 Weighted R1

0.068-0.077 0.292-0.342 0.792-0.902

0.067-0.072 0.263-0.302 0.760-0.940

0.046-0.050 0.158-0.177 0.482-0.534

17-29 3-6

0 0

0a 0a

0.078-0.102

0.031-0.039

0.029-0.033

0.009-0.011 1.495-1.640 0.552-0.838

0.009-0.010 2.515-2.953 0.156-0.282

0.010-0.012 2.782-3.089 0.310-0.358

C. Number distance violations Ê) (>0.2 A Ê) (>0.5 and <1.0 A D. rmsd values of distance restraints (AÊ) E. rmsd values from ideal covalent geometry Ê) Bond length (A Bond angles ( ) Impropers ( )

a The number of violations and the rmsd values for distance restraints during relaxation matrix MD re®nement do not include restraints involving non-exchangeable protons for which the force constant was scaled to zero during this stage of the calculation.

Figure 5. Stereo views of the eight intensity re®ned structures of the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) quadruplex formed in 100 mM KCl solution. The G1-G2-G3-C4 segments are shown in green and yellow, the T5-T6-T7-T8 segments are shown in white and the G9-G10-G11-C12 segments are shown in magenta and cyan. (a) View normal to the helix axis looking into the narrow groove formed by the two G9-G10-G11-C12 segments aligned in an antiparallel orientation. (b) View looking down the helix axis showing only the tetrad segments.

645

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Figure 6. (a) and (c) A comparison of the conformation of the T5-T6-T7-T8 loop for the intensity re®ned structures of the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) quadruplex formed in (a) 100 mM NaCl and (c) 100 mM KCl solution. Note the very different stacking patterns of the thymine bases for the 12-mer in (a) 100 mM NaCl and (c) 100 mM KCl solution. (b) and (d) A comparison of the alignments of the coplanar Watson-Crick G1
C12 and G9
C4 basepairs for the intensity re®ned structures of the d(G1-G2-G3-C4-T4-G9-G10-G11-C12) quadruplex formed in (b) 100 mM NaCl and (d) 100 mM KCl solution. These pairs align in (b) through inter-strand hydrogen bonding to form a G1(anti)
C12(anti)
G9(anti)
C4(anti) G
C
G
C tetrad in 100 mM NaCl solution. The hydrogen bonds are indicated by broken white lines. Such inter-strand hydrogen bonding alignments are not seen in (d) where the G1
C12 and G9
C4 base-pairs are sheared along their hydrogen bonding axis in 100 mM KCl solution.

T8-G9 segments are shown in Figure 9(a) and (b), while two views of the model of the K cation that penetrates into the plane and between the major groove edges of the Watson-Crick G1
C12 and G9
C4 base-pairs (at symmetry-related sites) are shown in Figures 9(c) and (d). The NMR restraint statistics and the quality of the averaged re®ned structures (without K cations) and averaged re®ned models (with K cations) of the d(G-G-G-CT4-G-G-G-C) quadruplex are listed in Table 4.

Discussion The present study has focused on the structural characterization of the predominant conformer of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution and a comparison with the corresponding structure of the predominant conformer of the d(G-G-G-C-T4-G-G-G-C) quadruplex

in 100 mM NaCl solution presented in the accompanying paper (Kettani et al., 1998). Conformational switch in slow exchange The distinct spectral signatures of the Na-cation stabilized d(G-G-G-C-T4-G-G-G-C) quadruplex (Figure 1(a)) and its K-cation stabilized counterpart (Figure 1(b)) have permitted us to monitor the conformational switch on proceeding from the Na cation to K cation conformations of the quadruplex. This transition is slow on the NMR time scale (observation of resonances from both conformers with no detectable line broadening due to exchange) when KCl is gradually added to a solution of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 50 mM NaCl solution at 10 C. The midpoint of the conformational switch in the quadruplex is observed on addition of 7.5 mM

646

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Figure 7. Two stereo views of overlap geometries in the representative intensity re®ned structure of the d(G1-G2G3-C4-T4-G9-G10-G11-C12) quadruplex formed in 100 mM KCl solution. Narrow and wide grooves are labeled by 0 N10 , 0 N20 and `W', respectively. (a) View down the helix axis with the symmetry-related adjacent G2
G3
G10
G11 tetrads (in cyan and magenta) stacked on each other. (b) View down the helix axis with the G2
G3
G10
G11 tetrad (magenta) stacked on top of the Watson-Crick G1
C12 and G9
C4 base-pairs (yellow). All hydrogen atoms and phosphate oxygen atoms in the backbone have been deleted for clarity.

KCl to a 50 mM NaCl containing solution with the transition being essentially complete on addition of 25 mM KCl. Folding topology The folding topology of the d(G-G-G-C-T4-G-GG-C) quadruplex in 100 mM KCl solution (shown schematically in Figure 3(b)) has many features that are similar to its counterpart in 100 mM NaCl solution (shown schematically in Figure 3(a)); Kettani et al., 1998). Speci®cally, both Na cationcoordinated and K cation-coordinated quadruplexes form through head-to-tail dimerization of a pair of hairpins such that adjacent strands are antiparallel to each other around the quadruplex and the connecting T-T-T-T loops are of the lateral type. The dimerization of the hairpins involves the mutual alignments of the major groove edges of the Hoogsteen G2
G11 and G10
G3 mismatch pairs to form the internal G2(syn)
G3(anti)
G10(syn)
G11(anti) tetrads for both quadruplexes.

Such structures exhibit similar strand symmetry and contain two distinct narrow and two symmetry-related wide grooves. There are, however, several distinct conformational features that differentiate the folding topology of the K cation-coordinated d(G-G-G-C-T4-GG-G-C) quadruplex from its Na cation-coordinated counterpart. These include the conformation of the T5-T6-T7-T8 loop (compare Figure 6(a) (Na‡ coordination) and (c) (K‡ coordination)), the relative alignment of the major groove edges of the Watson-Crick G1
C12 and G9
C4 base-pairs (compare Figure 6(b) (Na‡ coordination) and (d) (K‡ coordination)) and the consequences associated with the potential binding of one additional K cation within the coordination cavity of each of the two symmetry-related T-T-T-T loops of the quadruplex (Figure 9(a) and (b)). We shall discuss comparatively each of these distinct features associated with the solution structures of the Na cation-coordinated and K cation-coordinated d(G-G-G-C-T4G-G-G-C) quadruplexes and their functional conse-

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

647

Figure 8. (a) View of an averaged distance re®ned structure of the d(G1-G2-G3-C4-T4-G9-G10G11-C12) quadruplex formed in 100 mM KCl solution. (b) View of an averaged distance re®ned model of the d(G1-G2-G3-C4-T4-G9-G10G11-C12) quadruplex formed in 100 mM KCl solution following computations including ®ve K cations positioned within cavities identi®ed in the intensity re®ned structures of the quadruplex. See Materials and Methods for protocols of computations including K cations. The mono views and coloring codes are similar to their counterpart stereo views of the re®ned structures of Figure 5. The bound K cations in the model structure of the quadruplex in (b) are represented by yellow balls while the coordinating oxygen atoms are shown by red balls.

quences in some detail under separate sections below. Helical parameters The helical parameters of the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution have been estimated using the CURVES 5.2 program (Lavery & Sklenar, 1989). The adjacent G2
G3
G10
G11 tetrads stack primarily through overlap of their purine ®ve-

membered rings (Figure 7(a)) with the tetrad planes related by a right-handed twist of 15.0 . The Watson-Crick G1
C12 and G9
C4 base-pairs stack with the adjacent G2
G3
G10
G11 tetrad primarily through overlap of the purine rings of G1 and G2 and the purine rings of G9 and G10 (Figure 7(b)). We also observe three distinct grooves with unique dimensions for the solution structure of the 2-fold symmetric d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution (Figure 7(a) and (b)). Symmetry-related wide grooves are formed by the

Figure 9. (a) and (b) Two views of the model showing the potential K cation coordinated to acceptor atoms in the T6-T7-T8-G9 segment, and (c) and (d) two views of the model showing the potential K cation coordinated to acceptor atoms in the G1
C12 and G9
C4 pairs in the average distance re®ned model of the d(G1-G2-G3-C4-T4-G9-G10G11-C12) quadruplex in 100 mM KCl solution computed in the presence of K cations. The bound K cation in the model of the quadruplex is shown in yellow, the coordinating oxygen atoms in red, the T6-T7-T8 in white and G9 in cyan.

648

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Table 4. NMR restraint statistics and quality of d(G-G-G-C-T-T-T-T-G-G-G-C) quadruplex averaged structures without K ions and averaged models with ®ve K ions A. NMR restraints statistics NOE derived distance restraints on non-exchangeable protons NOE derived distance restraints on exchangeable protons Hydrogen bond distance restraints Non-crystallographic symmetry restraints

266 57 56 On all heavy atoms

B. Quality of the structure Ê ) between average structures without K cations and average models with five K cations (1) Convergence pairwise rmsd values (A For all residues 0.64 For residues 1 to 4, 9 to 12 0.54 For residues 5 to 8 0.75 Average structure Average model without K ion with 5 K ions (2) Distance violations Ê) (>0.2 A 0 0 Ê) rmsd values of distance restraints (A 0.034 0.035 (3) rmsd values from ideal covalent geometry Ê) Bond length (A 0.008 0.009 2.651 2.748 Bond angles ( ) 0.176 0.231 Impropers ( )

antiparallel alignment of adjacent G1-G2-G3-C4 and G9-G10-G11-C12 strands of each hairpin with Ê (measured as the shortest a groove width of 13.2 A 40 40 C -C distance across the groove for the segment spanning the central adjacent G2
G3
G10
G11 tetrads). Narrow grooves of somewhat different dimensions are formed by the antiparallel alignment of adjacent G1-G2-G3-C4 strands (designated Ê and adjacent G9N1) with a groove width of 9.9 A G10-G11-C12 strands (designated N2) with a Ê . The larger width of the groove width of 10.5 A symmetry-related wide grooves in the d(G-G-G-CT4-G-G-G-C) quadruplex in 100 mM KCl solution Ê ) compared to its value (11.8 A Ê ) in the (13.2 A 100 mM NaCl solution affects the shape of the radial projection of the quadruplex when viewed down the helix axis. This projection for the K cation-stabilized quadruplex (Figure 5(b)) is more oblong when compared with its counterpart for the Na cation-stabilized quadruplex (Figure 6(b); accompanying paper, Kettani et al., 1998). Modeling of K‡ coordination site between G
G
G
G tetrads One potential K‡ coordination site may be located between the planes of the symmetryrelated G1
C4
G9
C12 tetrads in the center of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution (Figure 8(b)). The K cation in this model is coordinated to four inwardly directed guanine O6 carbonyl oxygen atoms from each of the two G
G
G
G tetrads, similar to what has been observed in the X-ray structure of the quadruplex formed through dimerization of two d(G4T4G4) hairpins (Kang et al., 1992). This monovalent cation site is common to the solution structures of both the Na‡-coordinated (Kettani et al., 1998) and K‡-coordinated (this study) d(G-G-G-CT4-G-G-G-C) quadruplexes.

Modeling of K‡ coordination site within the T6-T7-T8-G9 segment The fold of the T5-T6-T7-T8 hairpin loops are strikingly different in the solution structures of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM NaCl (Figure 6(a)) and 100 mM KCl solution (Figure 6(b)). All four loop thymine bases adopt different orientations between the Na cation-coordinated (Figures 6(a)) and K cation-coordinated (Figure 6(b)) conformations of the quadruplex. No coordination site for a Na cation could be identi®ed within the T-T-T-T loop in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM NaCl solution (Kettani et al., 1998). By contrast, modeling suggests a potential coordination site for a K cation within the fold of the T6T7-T8-G9 segment in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution. The potential K cation could coordinate to the 0 O2 oxygen atoms of T6 and T8, atoms of T7, T8 and G9, the sugar ring O4 oxygen 0 and the backbone O5 oxygen atoms linking the T7T8 and T8-G9 steps (shown in red in Figure 9(a) and (b)). The modeling suggests two such symmetry-related K‡ coordination sites in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution and correspond to the outer K‡ coordination sites in Figure 8(b) (top view). This novel potential K cation coordination site could immobilize the T6-T7-T8-G9 segment into a speci®c fold, which in turn orients functional groups on its external surface in a de®ned alignment for recognition of potential targets. Thus, the structural transition observed for the d(G-G-G-CT4-G-G-G-C) quadruplex on proceeding from Na to K cations could re¯ect, in part, a potential K cation-induced conformational transition in the fold of the T6-T7-T8-G9 segment. These results have the potential for considerably extending our

649

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

understanding of monovalent coordination sites on folded nucleic acids. Thus, monovalent cations can bind not only to the backbone phosphates (Saenger, 1984) and between G-tetrad planes (Sundquist & Klug, 1989), but can also potentially bind within cavities generated by a unique fold of T-T-T-T hairpin loops. Further, there appears to be an exquisite selectivity that allows potential discrimination between monovalent Na (van der Ê ) and K (van der Waals Waals radius of 3.0 A Ê ) cations in the case of the d(G-G-Gradius of 4.7 A C-T4-G-G-G-C) quadruplex. It should be noted that the d(G-G-G-C-T3-G-GG-C) sequence containing one less thymine in the loop does not form a quadruplex in 100 mM KCl solution, even though it does so in 100 mM NaCl solution (accompanying paper, Kettani et al., 1998). This implies that a potential K cation binding cavity cannot form within a T-T-T loop and that the potential K‡ encapsulation within the T-T-T-T loop may make a critical contribution to the formation of the K cation-stabilized structure of the d(G-G-GC-T4-G-G-G-C) quadruplex. Modeling of K‡ coordination site aligning major groove edges of a pair of Watson-Crick G
C base-pairs The O6 and N7 acceptor atoms along the major groove edges of guanines G1 and G9 of the Watson-Crick G1
C12 and G9
C4 base-pairs are directed towards each other in the solution structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution (Figure 6(d)). The modeling suggests that a K cation could partially penetrate into the plane between these four acceptor atoms from opposing guanine G1 and G9 (Figure 9(c) and (d)) and coordinate to them (Figure 9(c) and (d)). In addition, this potential K cation could also coordinate to the inwardly pointing O6 oxygen atoms of the four guanine residues of the adjacent G2
G3
G10
G11 tetrad and the O2 oxygen of T8 loop residue. There are two such potential symmetry-related K‡ coordination sites in the proposed model of the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution and are ¯anked in this model by the central and outer proposed K‡ coordination sites in Figure 8(b) (top view). Potential monovalent cation binding sites in the d(G-G-G-C-T4-G-G-G-C) quadruplex A key insight emerges following comparison of the position of the three potential Na cation binding sites (Figure 8(b), top view; accompanying paper, Kettani et al., 1998) and the corresponding ®ve potential K cation binding sites (Figure 8(b), top view; this study) in their respective models of the d(G-G-G-C-T4-G-G-G-C) quadruplex. The potential Na cation binding sites may be posi-

tioned either between the adjacent central G
G
G
G tetrads (central binding site) or between adjacent G
G
G
G and G
C
G
C tetrads (symmetry-related ¯anking binding sites) as shown in Figure 8(b) (top view) (accompanying paper, Kettani et al., 1998). The larger K cation may also be positioned between the adjacent central G
G
G
G tetrads (central binding site) but the symmetry-related ¯anking potential K‡ sites could penetrate into the approximate plane of adjacent G
C base-pairs (Figure 8(b), top view; this study). Indeed, such a potential penetration of monovalent cation sites into the plane of the coordinating atoms has been observed in the high resolution X-ray structure of the parallelstranded d(T-G-G-G-G-T) quadruplex, where the Na‡ coordination sites shift gradually from being positioned between G-tetrad planes to being positioned within the G-tetrad plane on moving from the center towards either end of a pair of stacked G-quadruplexes (Laughlan et al., 1994; Phillips et al., 1997).

Biological significance The role of monovalent cations in stabilizing the architecture of G-quadruplexes (Howard & Miles, 1982) has been a matter of keen interest, beginning with the ®rst demonstration of G-tetrad formation (Gellert et al., 1962), and in the subsequent reinterest in this multi-stranded architecture due to the potential role of G-quadruplexes in telomere function, chromosome condensation and immunoglobulin switch regions (Sen & Gilbert, 1988; Williamson et al., 1989; Sundquist & Klug, 1989; reviewed by Rhodes & Giraldo, 1995). A potential role for monovalent cations emerged when it was determined that d(G4T4G4) folds into a quadruplex composed of four G-tetrads linked through lateral T4 loops in the crystal structure with K‡ as cation (Kang et al., 1992) while they were linked through diagonal T4 loops in the NMR based solution structure with Na‡ as cation (Smith & Feigon, 1992). However, more recent NMR studies have shown that the d(G4T4G4) quadruplex in solution contains diagonal loops independent of Na‡ or K‡ as counterion (Hud et al., 1996). This group found that speci®c proton markers associated with the diagonal loop-linked d(G4T4G4) quadruplex underwent small shifts as average resonances on proceeding from Na to K cation containing medium (Hud et al., 1996). Our results on the other hand de®ne a unique monovalent cation-induced conformational transition in the structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex with the two conformers in slow exchange on the NMR time scale. The key observation of the present study relates to the modeling of unique K cation coordination sites positioned within the T6-T7-T8-G9 segments at either end of the solution structure of the d(G-GG-C-T4-G-G-G-C) quadruplex in 100 mM KCl

650 solution. The biological signi®cance of this result is related to the potential role of G-quadruplexes as inhibitors of human immunode®ciency virus activity (Marshall et al., 1992; Wyatt et al., 1994). Indeed, recent research has established that the K cation selective folding of loop domains within intra-molecular G-quadruplexes results in the formation of potent anti-HIV oligonucleotides that constitute the ®rst drugs undergoing clinical trials as potential inhibitors of HIV integrase (Rando et al., 1995; Bishop et al., 1996; Mazumder et al., 1996; Jing et al., 1997a,b). The current premise is that loop segments of the quadruplex could adopt a unique fold de®ned by the coordinated K cation and that such segments are responsible for targeting the HIV-1 integrase. This premise receives support from the reported X-ray structure of a thrombin-DNA aptamer complex where the loop segments of the intramolecularly folded DNA aptamer quadruplex target the active site of thrombin (Padmanabhan et al., 1993). Our solution structure of K cation-coordinated d(G-G-G-C-T4-G-G-G-C) quadruplex de®nes for the ®rst time the principles involved in potential K cation coordination within a T-T-T-G segment, resulting in a de®ned loop architecture whose outwardly pointing functional groups provide a unique folded topology that can target potential receptor sites.

Materials and Methods Oligonucleotide synthesis and purification The d(G-G-G-C-Tn-G-G-G-C) oligodeoxyribonucleotides (n ˆ 3 and 4) and selectively labeled dI and dU analogs were synthesized as described in the accompanying paper (Kettani et al., 1998).

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex Distance geometry and molecular dynamics computations The distance geometry, distance restrained molecular dynamics and intensity restrained molecular dynamics computations (BruÈnger, 1992) for the d(G-G-G-C-T4-G-GG-C) quadruplex in 100 mM KCl solution used the same protocols as described in the accompanying paper (Kettani et al., 1998) for the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM NaCl solution. We also used the ``SUM'' averaging option in X-PLOR (BruÈnger, 1992; Nilges, 1995), which allows for ambiguity in the atom speci®cation in input distance restraints. The number of NOE intensities were 266 for each of ®ve mixing times during the intensity re®ned molecular dynamics computations.

Distance restrained molecular dynamics with K cations The GRASP program (Nicholls et al., 1991) was used to identify potential cavities capable of accommodating Ê ) within the K cations (van der Waals radius 4.70 A fold of the structure of the d(G-G-G-C-T4-G-G-G-C) quadruplex re®ned in the absence of K cations. Five such cavities lined by oxygen atoms were identi®ed within the structure of the quadruplex. The distance between the K cation and the coordinated guanine oxygen atoms and between adjacent internal bound K Ê and 4.85 A Ê , respectcations were given values of 3.0 A ively. These distances with bounds of 20% were used to position the K cations within the cavities during the computations. The protocol for distance restrained molecular dynamics calculations for the d(G-G-G-C-T4-G-G-G-C) quadruplex with K cations used the same protocols as described in the accompanying paper (Kettani et al., 1998) for the d(G-G-G-C-T4-G-G-G-C) quadruplex with Na cations. No planarity restraints were used during these computations.

NMR sample preparation and experiments NMR samples were prepared by dissolving 300 A260 units of d(G-G-G-C-T4-G-G-G-C) in 0.6 ml aqueous buffer (5 mM in single strands) containing 100 mM KCl. Details of the NMR experiments undertaken on the Varian Unity Inova 600 MHz spectrometers and the data processing protocols are as described in the accompanying paper (Kettani et al., 1998).

Restraints set The procedure for obtaining distance and intensity restraints for the d(G-G-G-C-T4-G-G-G-C) quadruplex in 100 mM KCl solution are the same as described in the accompanying paper (Kettani et al., 1998) for the d(G-GG-C-T4-G-G-G-C) quadruplex in 100 mM NaCl solution. Experimentally de®ned hydrogen bonds were restrained and non-crystallographic symmetry constraints were maintained throughout the computations. Planarity restraints were not used during the later stages of the computations and were not incorporated during the intensity restrained dynamics runs.

Graphics programs INSIGHT II (Molecular Simulations, Inc) and GRASP programs were used to display structures. Helical torsion angles were measured using the program CURVES 5.2 (Lavery & Sklenar, 1989).

Coordinates deposition Coordinates have been deposited in the Brookhaven Protein Data Bank, accession number: 1a8w.

Acknowledgments S.B. and A.K. contributed equally to this work. This research was supported by NIH grants GM 34504 to D.J.P. We thank Drs H. Zhao and R. Jones for a speci®c [15N]guanine-labeled sample of the dodecamer.

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

References Berns, K. I. & Linden, R. M. (1995). The cryptic life cycle of adeno-associated virus. BioEssays, 17, 237 ± 245. Bishop, J. S., Guy-Caffey, J. K., Ojwang, J. O., Smith, S. R., Hogan, M. E., Cossum, P. A., Rando, R. F. & Chaudhary, M. (1996). Intramolecular G-quartet motifs confer nuclease resistance to a potent antiHIV oligonucleotide. J. Biol. Chem. 271, 5698± 5703. BruÈnger, A. T. (1992). X-PLOR. A System for X-ray Crystallography and NMR, Yale University Press, New Haven, CT. Gellert, M., Lipsett, M. N. & Davies, D. R. (1962). Helix formation by guanylic acid. Proc. Natl Acad. Sci. USA, 48, 2013± 2018. Guschlbauer, W., Chantof, J. F. & Thiele, D. (1990). Four-stranded nucleic acid structures 25 years later: from guanosine gels to telomeric DNA. J. Biomol. Struct. Dynam. 8, 491 ± 511. Hardin, C. C., Henderson, E., Watson, T. & Prosser, J. K. (1991). Monovalent cation induced structural transitions in telomeric DNAs: G-DNA folding intermediates. Biochemistry, 30, 4460± 4472. Hardin, C. C., Watson, T., Corregan, M. & Bailey, C. (1992). Cation-dependent transition between the quadruplex and Watson-Crick hairpin forms of d(CGCG3GCG). Biochemistry, 32, 833 ± 841. Howard, F. B. & Miles, H. T. (1982). Poly (inosinic acid) helices: essential chelation of alkali metal ions in the axial channel. Biochemistry, 21, 6736± 6745. Hud, N. V., Smith, F. W., Anet, F. A. L. & Feigon, J. (1996). The selectivity of K versus Na in DNA quadruplexes is dominated by relative free energies of hydration: a thermodynamic analysis by 1H NMR. Biochemistry, 35, 15383± 15390. Jing, N., Rando, R. F., Pommier, Y. & Hogan, M. E. (1997a). Ion selective folding of loop domains in a potent anti-HIV oligonucleotide. Biochemistry, 36, 12498± 12505. Jing, N., Gao, X., Rando, R. F. & Hogan, M. E. (1997b). Potassium-induced loop conformational transition of a potent anti-HIV oligonucleotide. J. Biomol. Struct. Dynam. 15, 573 ± 585. Kang, C., Zhang, X., Ratliff, R., Moyzis, R. & Rich, A. (1992). Crystal structure of a four-stranded Oxytricha telomeric DNA. Nature, 356, 126 ± 131. Kettani, A., Bouaziz, S., Zhao, H. & Jones, R. A. (1998). Solution structure of a Na cation stabilized DNA quadruplex containing G
G
G
G and G
C
G
C tetrads formed by G-G-G-C repeats observed in adeno-associated viral DNA. J. Mol. Biol. 282, 619± 636. Laughlan, G., Murchie, A. I., Norman, D. G., Moore, M. H., Moody, P. C., Lilley, D. M. & Luisi, B. (1994). The high resolution crystal structure of a parallel-stranded guanine tetraplex. Science, 265, 520 ± 524. Lavery, R. & Sklenar, H. (1989). De®ning the structure of regular nucleic acids: conventions and principles. J. Biomol. Struct. Dynam. 6, 655 ± 667. Marshall, W. S., Beaton, G., Stein, C. A., Matsukura, M. & Caruthers, M. H. (1992). Inhibition of human immunode®ciency virus activity by phosphorodithioate oligodeoxycytidine. Proc. Natl Acad. Sci. USA, 89, 6265± 6269. Mazumder, A., Neamati, N., Ojwang, J. O., Sunder, S., Rando, R. F. & Pommier, Y. (1996). Inhibition of the human immunode®ciency virus type I integrase by

651 guanosine quartet structures. Biochemistry, 35, 13762± 13771. Miura, T., Benevides, J. M. & Thomas, G. J., Jr (1995). A phase diagram for sodium and potassium ion control of polymorphism in telomeric DNA. J. Mol. Biol. 248, 233± 238. Nicholls, A., Sharp, K. & Honig, B. H. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins: Struct. Funct. Genet. 11, 281± 296. Nilges, M. (1995). Calculation of protein structures with ambiguous distance restraints. Automated assignment of ambiguous NOE cross-peaks and disul®de connectivities. J. Mol. Biol. 245, 645±660. Padmanabhan, K., Padmanabhan, K. P., Ferrara, J. D., Sadler, J. E. & Tulinski, A. (1993). The structure of a-thrombin inhibited by a 15-mer single-stranded DNA aptamer. J. Biol. Chem. 268, 17651± 17654. Patel, D. J., Kozlowski, S. A., Nordheim, A. & Rich, A. (1982). Right-handed and left-handed DNA studies of B-DNA and Z-DNA by using proton nuclear Overhauser effect and phosphorus NMR. Proc. Natl Acad. Sci. USA, 79, 1413± 1417. Patel, D. J., Bouaziz, S., Kettani, A. & Wang, Y. (1999). Structures of guanine-rich and cytosine-rich quadruplexes formed in vitro by telomeric, centromeric and triplet repeat disease DNA sequences. In Oxford Handbook of Nucleic Acid Structure (Neidle, S., ed.), Oxford University Press, Oxford, in the press. Phillips, K., Dauter, Z., Murchie, A. I. H., Lilley, D. M. J. & Luisi, B. (1997). The crystal structure of a paralÊ resolution. lel-stranded guanine tetraplex at 0.95 A J. Mol. Biol. 273, 171± 182. Pilch, D. S., Plum, G. E. & Breslauer, K. J. (1995). The thermodynamics of DNA structures which contain lesions or guanine tetrads. Curr. Opin. Struct. Biol. 5, 334± 342. Pinnavaia, T. J., Marshall, C. L., Mettler, C. M., Fisk, C. L., Miles, H. T. & Becker, E. D. (1978). Alkali metal ion speci®city in the solution ordering of a nucleotide, 50 -guanosine monophosphate. J. Am. Chem. Soc. 100, 3625± 3627. Rando, R. F., Ojwang, J., Elbaggari, A., Reyes, G. R., Tinder, R., McGarth, M. S. & Hogan, M. E. (1995). Suppression of human immunode®ciency virus type 1 activity in vitro by oligonucleotides which form intramolecular tetrads. J. Biol. Chem. 270, 1754± 1760. Rhodes, D. & Giraldo, R. (1995). Telomere structures and function. Curr. Opin. Struct. Biol. 5, 311 ±322. Saenger, W. (1984). Principles of Nucleic Acid Structure, Springer-Verlag Inc., New York. Schultze, P., Smith, F. W. & Feigon, J. (1994). Re®ned solution structure of the dimeric quadruplex formed from the Oxytricha telomeric oligonucleotide d(GGGGTTTTGGGG). Structure, 2, 221± 233. Sen, D. & Gilbert, W. (1988). Formation of parallel fourstranded complexes by guanine-rich motifs in DNA and its implications in meiosis. Nature, 334, 364 ± 366. Smith, F. W. & Feigon, J. (1992). Quadruplex structure of Oxytricha telomeric DNA oligonucleotides. Nature, 356, 164± 168. Sundquist, W. I. & Klug, A. (1989). Telomeric DNA dimerizes by formation of guanine tetrads between hairpin loops. Nature, 342, 825± 829. Williamson, J. R. (1993). Guanine quartets. Curr. Opin. Struct. Biol. 3, 357± 362.

652

K Cation-coordinated d(G-G-G-C-T4-G-G-G-C) Quadruplex

Williamson, J. R., Raghuraman, M. K. & Cech, T. R. (1989). Monovalent cation-induced structure of telomeric DNA: the G-quartet model. Cell, 59, 871 ± 880. Wyatt, J. R., Vickers, T. A., Roberson, J. L., Buckheit, R. W., Jr, Klimkait, T., DeBaets, E., Davis, P. W., Rayner, B., Imbach, J. L. & Ecker, D. J. (1994). Combinatorially selected guanosine-quartet is a potent

inhibitor of human immunode®ciency virus envelope-mediated cell fusion. Proc. Natl Acad. Sci. USA, 91, 1356± 1360. Xu, Q., Deng, H. & Braunlin, W. H. (1993). Selective localization and rotational immobilization of univalent cations on quadruplex DNA. Biochemistry, 32, 13130± 13137.

Edited by I. Tinoco (Received 19 February 1998; received in revised form 19 June 1998; accepted 21 June 1998)