Proc. Natl. Acad. Sci. USAVol. 91, pp. 10546-10550, October 1994Biochemistry

d(G3T4G3) forms an asymmetric diagonally looped dimericquadruplex with guanosine 5'-syn-syn-anti and 5'-syn-anti-antiN-glycosidic conformationsFLINT W. SMITH, FRANCIS W. LAU, AND JULI FEIGON*Department of Chemistry and Biochemistry and Molecular Biology Institute, University of California, Los Angeles, CA 90024

Communicated by M. A. El-Sayed, July 18, 1994 (received for review April 20, 1994)

ABSTRACT The structure formed by the DNA oligonu-cleotide d(G3T4G3) has been studied by one- and two-dimensional 1H NMR spectroscopy. In NaCl solution,d(G3T4G3), like d(G4T4G4) (Oxy-1.5), forms a dimeric quad-ruplex with the thymines in loops across the diagonal of the endquartets. Unlike Oxy-l.5, the dimer is not symmetric, and bothmonomer strands are observed in NMR spectra. Three quar-tets are formed from the GGG tracts. Glycosidic conformationsof the guanines are 5'-syn-syn-anli-(loop)-syn-an-anti in onestrand and 5'-syn-an-an-0(loop)-syn-syn-anti in the otherstrand. Thus, the stkg of the quartets (tail-to-tail, head-to-tail) is unlike all previously described fold-back (tail-to-tail,head-to-head) and paralel-stranded (head-to-tail, head-to-tail)quadruplexes.

The propensity of guanine nucleotides and polymers to formquadruplex structures containing guanines hydrogen bondedin quartets has been known for more than 30 years (1), butonly recently have possible biological roles for DNA andRNA G-quadruplex structures been proposed. Biologicallyrelevant sequences which have been shown to have thepotential to form quadruplexes include the G-rich strand oftelomeres, in particular the two-repeat single-strand over-hang (2, 3), immunoglobulin switch regions (4), and thedimerization domain of the human immunodeficiency virustype 1 genome (5, 6). The (3 subunit ofthe Oxytricha telomerebinding protein has been shown to facilitate the formation ofG-quadruplexes (7). Several other proteins have been foundto interact specifically with quadruplexes, including MyoD(8), QUAD (9), a quadruplex-specific nuclease (10), and aeukaryotic topoisomerase II (11).A number of quadruplex structures formed by short oli-

gonucleotides containing primarily guanines and thymines oruridines have now been solved or modeled on the basis ofNMR (12-26) and x-ray crystallographic (27, 28) data. Thesestructures all contain guanine nucleotides hydrogen bondedin G-quartets as proposed from the early fiber diffractionstudies; however, they vary in molecularity (one, two, or fourstrands), strand polarity (parallel, antiparallel, or both), gua-nosine N-glycosidic conformations (syn and anti), and incases where there are loops, loop orientation (spanning anedge or the diagonal of a G-quartet). These differences leadto differences in such structural features as groove widthsand the positions and accessibility of hydrogen bond donorsand acceptors within the grooves. The factors that determinewhat structure will form for a particular oligonucleotidesequence are unclear. Elucidation of the details of G-quad-ruplex structures and the factors determining them may helpin understanding the formation of G-DNA in biological sys-tems.

We recently presented the quadruplex structure formed byan oligonucleotide based on the repeat sequence d(T4G4)found in Oxytricha telomeres (12, 14, 15). Oligonucleotidescontaining two or four repeats of this sequence readily formdimeric and monomeric quadruplexes, respectively (29). Ithas been proposed that dimerization of Oxytricha telomeres(30, 31) may occur by means of quadruplex formation atsingle-strand overhangs containing these repeats (2). TheNMR structure of the oligonucleotide d(G4T4G4) (Oxy-1.5) istopologically different from both the model originally pro-posed by Williamson et al. (2) and from the x-ray crystalstructure of the same molecule (27). In both the NMRsolution structure and the x-ray crystal structure, a dimericquadruplex with four G-quartets and thymine loops at oppo-site ends of the quadruple helix is formed. However, in thesolution structure, the loops cross the diagonal of the endquartets, while in the crystal structure each loop spans a widegroove. Shafer and co-workers (32) have reported CD, UVmelting, one-dimensional NMR, and gel electrophoresis re-sults on a similar sequence, d(G3T4G3), and suggested that itforms yet another, topologically distinct, dimer.Here we present the results of our NMR investigation into

this related sequence, d(G3T4G3). We find that, as previouslyreported by Scaria et al. (32), d(G3T4G3) forms an asymmetricdimeric quadruplex. However, their conclusion that thestructure cannot be diagonally looped is incorrect. Here weshow that d(G3T4G3), like d(G4T4G4), forms a dimeric quad-ruplex containing three G-quartets with TTTT loops acrossthe diagonal of the ends of a stack of guanine quartets.Diagonal looping has also been observed in the unimolecularfour-repeat quadruplexes formed by oligonucleotides con-taining the Oxytricha and human telomere repeats,d[G4(T4G4)31 (Oxy-3.5) (12, 29) and d[AG3(T2AG3)3] (26),respectively. In all fold-back quadruplexes solved to date, theguanine glycosidic conformations alternate syn and anti alongeach strand (12-16, 21, 26, 27, 29). In contrast, in thed(G3T4G3) quadruplex, guanosine glycosidic conformationsare S'-syn-syn-anti in two G3 tracts and 5'-anti-anti-syn in theother two G3 tracts. The resulting G-quartet stacking has not,to our knowledge, previously been observed in quadruplexstructures.

MATERIALS AND METHODSSample Preparation. The DNA decamers d(GGGTTT-

TGGG) [sequence d(G3T4G3)] and d(GGGTTTTGGI) (se-quence I10) were synthesized and purified as described (14).The NMR samples contained 2.5-3.8 mM dimer and 100mMNaCl (pH 7.5) in 400 i4 of90% H20/101oD20((H20 samples;D, deuterium) or 99.996% D20 (D20 samples). Samples were

Abbreviations: D, deuterium; NOE, nuclear Overhauser effect;NOESY, NOE spectroscopy; HOHAHA, homonuclear Hartmann-Hahn; P.COSY, purged correlated spectroscopy; qss, skewedsquared sine bell.*To whom reprint requests should be addressed.

10546

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Natl. Acad. Sci. USA 91 (1994) 10547

exchanged from H20 to D20 by repetitive drying in theNMRtube under a stream of nitrogen gas and resolubilization.NMR Spectroscopy. 1H NMR spectra were obtained on

General Electric model GN500 and Bruker (Billerica, MA)model AMX 500 spectrometers. Spectra in H20 were ac-

quired with a 11 spin echo read pulse (33). A time domaindeconvolution was used to further reduce the H20 signal.Spectra in D20 were acquired with presaturation of theresidual HOD during the recycle delay. Nuclear Overhausereffect (NOE) spectroscopy (NOESY) (34), homonuclearHartmann-Hahn (HOHAHA) (35), purged correlated spec-

troscopy (P.COSY) (36), and 1H-3XP heteroCOSY (37) spec-

tra were acquired as described (14). Spectra were processedwith the FELIX software (Biosym Technologies, San Diego).Two-dimensional spectra shown here were apodized withskewed squared sine bell (qss) and zero-filled to 2048 pointsin both dimensions.

RESULTS

d(G3T4G3) Forms an Asymmetric Dimerc Quadruplex. Fig.1 shows a one-dimensional NMR spectrum of the hydrogen-bonded imino resonances of d(G3T4G3) in 100 mM NaCl.Intensity for 12 different guanine imino resonances is ob-served, as previously reported (32). The presence of bothhydrogen-bonded guanine imino and amino protons is strongevidence for a G-quartet structure (12). Analysis of two-dimensional NOESY spectra (spectra discussed below andothers not shown) indicates that three unique G-quartets are

formed. Assignments of the imino resonances are indicatedon the spectrum.Resonance Assignets. Assignments of the proton reso-

nances ofd(G3T4G3) were initially made by using the methodspreviously described for assignments of quadruplex DNAstructures (14, 16). These assignments were confirmed bycomparison of NOESY spectra to those of the related mol-ecule d(GGGGTTTTGGGG) (Oxy-1.5) (14). NMR spectraused in making assignments of nonexchangeable protonresonances included NOESY (Tm = 200 msec), HOHAHA(Tm = 100 ms), and P.COSY spectra acquired on the samplein D120 at 230C. A NOESY spectrum (Tm = 200 msec) in H20at 50C was used to assign exchangeable proton resonances.Exchangeable proton assignments were checked for internalconsistency against the models. Assignments were furtherconfirmed by analysis of NOESY spectra of the I10 deriva-

11.5ppm

FIG. 1. The 500-MHz one-dimensional 1H NMR spectra of theimino resonances ofd(G3T4G3) at 20"C. Twelve imino resonances are

observed, indicating two distinct strand conformations. Resonanceassignments are indicated (nucleotide number, counting the twostrands sequentially). Sample is 3.8 mM dimer, 100 mM NaCl (pH7.5). The spectrum was acquired with a 11 spin echo pulse sequencewith 8000-Hz spectral width, 128 acquisitions, and 1.8-sec recycledelay. The 4096-point free induction decay was apodized with a qssapodization function of 1800 points shifted 700, skewed 1.1, and zero

filled to 8192 points.

tive as previously described (13, 14). A 'H-31P heteronuclearCOSY spectrum confirmed some 1H sequential assignmentsbut did not show all of the expected crosspeaks. Assignmentswere obtained for all proton resonances with the exception ofthree pairs of H5',H5", the thymine imino, and three pairs ofguanine amino protons. It is worth noting that, unlike in thecase of Oxy-1.5, intra- and intermolecular NOEs can beunambiguously distinguished.d(G3T4G3) Forms a Diagonafly Looped Dimeric Quadruplex

Simiar to Oxy-1.5. For dimeric quadruplexes, there are alimited number of possible strand orientations. Fig. 2 illus-trates eight possible dimeric edge or diagonally looped[d(G3T4G3)h2 quadruplexes containing three G-quartets.Three ofthe model structures have both thymine loops at oneend of the stack of guanine quartets and five have one loopat each end. Of the latter, three are edge looped and two arediagonally looped. Models a and b can be either symmetric orasymmetric and all the other models must be asymmetric.Each of the eight strand-orientation models has a subset offour or eight different possible orientations of the threeguanine quartets (see Discussion).

Fig. 3 shows a comparison of the NOESY spectra ofd(G3T4G3) and Oxy-1.5, which has been previously shown toform a diagonally looped quadruplex similar to model g (12,14, 15). The region shown contains crosspeaks between thearomatic resonances and the deoxyribose H2',H2' and thy-mine methyl resonances. The presence of crosspeaks be-tween two different guanine quartets and thymines from twodifferent loops indicates that d(G3T4G3) forms a dimericquadruplex with loops at opposite ends ofthe G-quartet stack(12). The chemical shifts and NOE crosspeak patterns ofd(G3T4G3) are strikingly similar to those of Oxy-1.5. Onethymine loop and the guanine quartet near it are nearlyidentical to the symmetric loops and end G-quartets ofOxy-1.5. The chemical shifts of the resonances from thesecond loop and its adjacent G-quartet are slightly different,but the NOE crosspeak patterns are essentially the same. Theidentity of the NOE crosspeak patterns with those seen inOxy-1.5 indicates that [d(G3T4G3)h2 forms a similar diagonallylooped quadruplex. The two loops of [d(G3T4G3)12 are inslightly different environments due to the asymmetry of thestructure. The two diagonally looped models (g and h) can bedistinguished on the basis of imino-imino NOEs and theGH8-IH2 NOEs of the inosine derivative as described forOxy-1.5 (14).The Guanosine Glycosidic Conformations in [d(G3T4G3)h

Are Not Alternating synani. Fig. 4 shows a portion of theNOESY spectrum of d(G3T4G3) containing the aromatic-Hi'

ar b c 9

dWe f hW

FiG. 2. Illustrations of the strand polarity in eight possibledimeric quadruplex structure models. Models a-c have two edgeloops at one end; models d-f have edge loops at opposite ends; andmodels g and h have diagonal loops. Models a, b, d, and e have alladjacent strands in antiprallel orientation. Models c and f-h havesome strands in parallel orientation. Only models a and b could besymmetric. Not illustrated are models that have loops from one endquartet to the other (crown model) (31) or the model in which twoloops cross at one end.

Biochemistry: Smith et A

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Natl. Acad. Sci. USA 91 (1994)

8.0 7.5 7.0pprr

H B F A DE CI I

8.0 75ppmr

FIG. 3. Comparison ofNOESY spec-tra of d(G3T4G3) (Left) and Oxy-1.5(Right). This region contains crosspeaksbetween the aromatic resonances and theH2',H2' and methyl resonances. Reso-nances from the thymine loop and theG-quartet near it that are nearly identicalto the symmetric loops and end quartetsofOxy-1.5 are indicated by capital lettersand connected by bold lines. The chem-ical shifts ofthe resonances from the loopand quartet at the opposite end (lower-case letters) are slightly different, but theNOE crosspeak patterns (broken lines)are essentially unaffected. The Oxy-1.5sample was 5.2 mM strand in 50 mMNaCl (pH 6.0). The spectrum was takenat 250C and is the 200-msec NOESYdescribed by Schultze et al. (15).

crosspeaks. Six syn and six anti guanines can be identified onthe basis of the intensity of the GH8-H1' crosspeaks. TheP.COSY crosspeak patterns (not shown) are indicative ofS-type sugar puckers found in B-DNA-type helices (14), aswith other quadrupled structures determined to date. The 31pchemical shifts are also consistent with a regular B-DNA-likephosphodiester backbone.Comparison ofthe NOESY spectrum ofd(G3T4G3) with that

of Oxy-1.5 (Fig. 3) provides confirmation of assignments ofmost of the guanine nucleotides in the end quartets. These

(DCd

8.0 7.6pprn

7.2

FiG. 4. Region of the two-dimensional NOESY spectrum showingthe crosspeaks between aromatic and Hi' resonances of d(G3T4G3) at230(. Sequential n-to-n and n-to-n + 1 connectivities along each strandare indicated by solid (G1-G1O) and broken (G11-G20) lines. Intranu-cleotide crosspeaks are numbered. The n-to-n + 1 connectivities are ofthree types: "normal" sequential, nHl' to n+lH6,8 (+); "reverse"sequential, nH6,8 to n+1H1' (-), or both (+). The n-to-n + 1 connec-tivities observed in this region are G'-G2tG3+T4+T5/T*+T7/G8±G9+G10 and Gll±Gu2+Gl3+Tl4+Tl5/Tl6+Tl7/Gls-Gl9-GC,where / indicates no crosspeak observed. Two cross-strand cross-peaks, G1 to T16 and G1' to T, are indicated by *. The NOESYspectrum was acquired with rm = 200 msec and 48 acquisitions foreachof 255 tl increments. The 700-shifted 1.1-qss apodization functions of400 and 255 points were applied in t2 and ti, respectively.

include the guanines at the 5' and 3' ends ofeach strand and theguanine 5' to the T4 loop (Fig. 3, peaks A, H, and C, respec-tively). G-quartets have two different orientations (i.e., faces).We define the "heads" and "tails" sides of the G-quartet byapplying a right-hand rule to the arrows pointing firm hydrogenbond donors to acceptors with the thumb pointing towards the"heads" face (see Fig. 6). The phosphodiester backbone con-formation and glycosidic conformation together determinequartet orientation. Since we know that the structure has aregular phosphodiester backbone and we know the glycosidicconformations ofthe guanine nucleotides, the orientation ofthequartets is also known. In d(G3T4G3), like Oxy-1.5, the "head"sides of the end G-quartets face the thymnine loops. Thus, bycomparison of the NOESY spectra of d(G3T4G3) and Oxy-1.5,both the backbone orientation (model g) and the quartet ornen-tation are determined. Note thatthe two models constructed byplacing the central quartet in either orientation are identical-i.e., they simply reverse the identity of the two strands. Theglycosidic torsion angles for the two strands of the quadruplexformed by d(G3T4G3) are 5'-GsynGs2nGa~ntT4aniTa6"TianidrG8 G9 G10 Iand 5'11 Ge12 G13t Ta14t T15t Ta6t Ta17t G16,,natiu anti an syn atuat1aint aiannt 1isyn

Gy anG aBase-Hi' Region ofNOESY Spectrum Shows Unusual Con-

nectivities. Sequential connectivities between the base andHi' resonances of d(G3T4G3) are indicated by the solid andbroken lines in Fig. 4. An NOE crosspeak from nHl' ton+lH6,8 is expected between two sequential anti bases inB-DNA (38). These "normal" sequential NOEs are observedbetween anti guanines in [d(G3T4G3)12 and between some ofthe loop thymines. In NOESY spectra of Oxy-1.5, both nHl'to n+1H6,8 and nH6,8 to n+1Hl' are observed between a5'syn base and a neighboring anti base (12). These doubleconnectivities are also observed in the NOESY spectrum of[d(G3T4G3)12. Where there are two sequential syn bases, onlythe "reverse"' sequential nH8 to n+1H1' crosspeak is ob-served. Using these normal-sequential and reverse-sequential NOE connectivities, we are able to sequentiallyconnect all guanines and parts of both loops.NOESY Spectra of Inosine Derivative Shows Head-to-Tail

G-Quartet Interactions. As in the case of Oxy-1.5, NOESYspectra ofthe inosine derivative I10 provided confirmation ofthe diagonally looped structure (14). Fig. 5 shows the regionof the D20 NOESY spectrum of 110 contg the aromaticresonances and their crosspeaks. The G-quartet structureand assignments are confirmed by NOE connectivities be-tween the inosine H2 protons and the H8 protons of theadjacent guanine. In addition to the expected intraquartetG3H8-I20H2 and G13H8-170H2 crosspeaks, an interquartetNOE crosspeak is observed between G12H8 and I2°H2. This

GAGGGJT, TI.TLLGGGG,

i.

nflln {! ,f)

H1

- -016

I15~ 6 4t~~~2? i 14

;______ 14

1x- ? ---- '

'13 - it 7,y 0Q8

20

10(1;

10548 Biochemistry: Smith et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Biochemistry: Smith et al.

0 CD

8 0ppm

Proc. Natl. Acad. Sci. USA 91 (1994) 10549

5

IlC

anti

syn

7 5 syn

FIG. 5. Region of the two-dimensional NOESY spectrum of theinosine derivative, showing the crosspeaks between aromatic reso-nances at 25TC. Peak A, G3H8-I12H2, is an intraquartet NOE similarto the one observed for the inosine derivative of Oxy-1.5; peak B,G13H8-110H2, is an intraquartet NOE at the head-to-tail end of thedimer; peak C, G12H8-110H2, is an interquartet NOE. The NOESYspectrum was acquired with a spectral width of 10,000 Hz and 48acquisitions for each of 255 tl increments. The 700-shifted 1.1-qssapodization functions of400 and 255 points were applied in t2 and t1,respectively.

interquartet NOE is consistent with the head-to-tail stackingbetween the quartets at this end of the molecule.

DISCUSSIONMultiple Dimeric Quadruplex Models for [d(G3T4G3)J2. The

models presented in Fig. 2 make up a family of three quartetquadruplexes which vary in strand polarity and quartetorientation. These represent a total of 48 different structuralmodels for a dimeric quadruplex ofd(G3T4G3) with noncross-ing loops at the ends of the quadruplex. With only theinformation presented by Scaria et al. (32) that d(G3T4G3)forms an asymmetric dimer, none of these models could beexcluded. Models a and b could be symmetric, since there isa potential axis of symmetry parallel to the helix axis, buteven for these models an asymmetric interaction between theloops would destroy the symmetry of the structure. Modelsa-c and fhave distinct ends, with either both loops at one end(a-c) or one loop spanning a narrow groove and the otherspanning a wide groove (f). These models have eight distinctstacking variations ranging from all quartets "heads up" toall "heads down." The other four models (d, e, g, and h) haveessentially symmetric loops and lose the distinction between"up" and "down." These models have only four differentvariations each. Ofthese 48 distinct asymmetric dimers, onlythe model structure in Fig. 6, which is one ofthe four possiblemodel g variations, is consistent with the NMR data.

Strand and Loop Orientations in the [d(G3T4G3)J2 Quadru-plex. The quadruplex structure formed by [d(G3T4G3)h2 isdiagonally looped, in the same way as Oxy-1.5. The conse-quences of this loop arrangement, in terms of strand orien-tation and groove widths, have been presented (12, 15). Themodel structure presented here for [d(G3T4G3)h2 is essentiallythat of Oxy-1.5 from which one of the central quartets hasbeen deleted. The two end quartets and the thymine loops arerelatively unaffected. The grooves consist ofone wide groovebetween the two 5' guanine tracts, one narrow groovebetween the two 3' guanine tracts, and two medium groovesbetween the parallel guanine tracts.

Quartet Orientations in the [d(G3T4G3)J2 Quadruplex. Thethree quartets are arranged such that two are oriented head-

FIG. 6. Schematic of the quadruplex structure formed byd(G3T4G3) and I10. The numbering scheme is indicated. Stranddirection along each edge of the quadruplex is indicated by largearrows; note that there are both parallel and antiparallel adjacentstrands. The four grooves are labeled narrow, medium, or wide.Guanine bases are shown as rectangular solids; syn bases are drawnwith thick lines. Shading indicates the "heads" face ofa quartet (topquartet). The figures at the right illustrate the definition of "heads"and "tails" by applying a right-hand rule to the hydrogen bonddirections (from donor to acceptor) indicated by arrows on the edgesof the bases.

to-tail and two are oriented tail-to-tail. Within a quartet, allguanines must be in the same orientation to hydrogen bondcorrectly-i.e., all heads on one face and all tails on the otherface of the planar quartet (Fig. 6). Since the quadruplex hasstrands with antiparallel polarity, some bases must beflipped. The fact that the phosphodiester backbone isB-DNA-like indicates that the bases are flipped to the synconformation. These syn nucleotides are unlike those inZ-DNA, which are formed by rotating the sugars in thephosphodiester backbone. The guanosine glycosidic confor-mations, 5'-syn-syn-anti-3' and 5'-syn-anti-anti-3', are thusdirectly related to the stacking orientation of the quartets.The syn-syn and anti-anti steps are between the two quartetsstacked head-to-tail. This type of stacking has previouslybeen seen only in parallel-stranded quadruplexes having onlyanti nucleotides (18-20, 22, 23). The syn-anti steps arebetween the quartets stacked tail-to-tail, an orientation seenonly in fold-back structures (12-16, 21, 26, 27, 29).Compariso to Other Quadruplex Structures. As discussed

above, the structure formed by d(G3T4G3) is very similar tothe solution structure of Oxy-1.5 (12, 14, 15). The details ofgroove widths and loop conformation are presumably onlyslightly affected. The main difference can be described as thedeletion of one of the central quartets. This destroys thetwofold symmetry axis by placing the center of the moleculein the plane of the center quartet. This quartet has twodistinct faces and interacts differently with the end quartetson either side of it. Thus a separate set ofresonances for eachof the two strands in the dimer is observed. Oxy-3.5 also hasa very similar diagonal loop and tail-to-tail, head-to-head,tail-to-tail stacking (12, 29). Thus, all solution NMR struc-tures reported to date for oligonucleotides having more thanone repeat of sequences of the type GJT4 contain at least onediagonal loop and tail-to-tail stacking. The quartet stacking inthe crystal structure of Oxy-1.5 is the same as the solution

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Nati. Acad. Sci. USA 91 (1994)

structure, although the loops span wide grooves (27). Thethrombin binding aptamer, d(GGTTGGTGTGGTTGG),forms a unimolecular quadruplex with two G-quartets (13,21). This structure has only groove-spanning loops, and thetwo G-quartets are oriented tail-to-tail.The solution structure of the oligonucleotide d[AG3

(T2AG3)3], which contains 3.5 repeats of the human telomererepeat sequence TTAGGG, also has a diagonal loop at oneend and three quartets (26). However, the structure is verydifferent from the [d(G3T4G3)J2 quadruplex. The stacking ofthe quartets alternates like that in Oxy-1.5 and Oxy-3.5,head-to-head and tail-to-tail, and the first 5' guanine is antirather than syn. This may be due to the fact that the actual 5'nucleotide is an A rather than a G. The sequence andstructure of the diagonal loop, TTA, are substantially differ-ent from those of the TTTT loop.

All of the tetrameric quadruplexes studied to date formwith parallel strand orientation (18-20, 22, 23). In thesecases, there is no geometric requirement for syn nucleotides,and all of the nucleotides are in the anti conformation. TheG-quartets are all stacked head-to-tail.

Tight binding of cations within or between G-quartets isproposed to contribute to the stability of G-quadruplexes(1-3). The different stacking arrangements of G-quartetsresult in different environments for ion binding.Foling Rules for Quadruplexes? The structure of the

[d(G3T4G3)12 quadruplex is important because it illustratesthat guanine tracts in fold-back quadruplexes do not neces-sarily alternate syn-anti. The underlying question of whythese oligonucleotides fold as they do is as yet unresolved,although some trends begin to appear. Parallel strand orien-tation is favored. In tetrameric quadruplexes, in which eachguanine tract is independent, all strands are parallel (18-20,22, 23). In solution structures of fold-back quadruplexes,diagonal looping results in some tracts being parallel (14, 26).In calorimetric studies of quadruplex dimers it was observedthat quadruplexes designed to have all parallel strands (witha nonnatural linkage) formed structures with a higher stabilitythan Oxy-1.5 (39).

All anti conformations and B-DNA-like phosphodiesterbackbones are found in tetrameric quadruplexes. To have aregular B-DNA-like phosphodiester backbone in foldbackstructures, the guanine base must flip over when the stranddirection changes. For fold-back quadruplex structuresformed from oligonucleotides with a 5'-terminal G, this G isalways syn (12-16, 21, 26, 27, 29). This conformation for the5'-G may be stabilized by a hydrogen bond between the GN3and the 5'-OH (40).For both Oxy-1.5 and d(G3T4G3) we observe a single

diagonally looped conformation in solution. This is somewhatsurprising, because there is very little difference between thetwo diagonally looped models (g and h) (Fig. 2). Changing thepolarity ofone strand yields a model having identical guaninetract polarity, groove widths, and quartet stacking (rotatemodel h 900 clockwise for a comparison of the guanine core).The only difference would be in the interactions between theloops and the end quartets, which may contribute to theoverall stability. We propose that the formation of the singlediagonally looped structure observed may be a consequenceof the folding pathway through a specific intermediate (15).

This work was supported by National Institutes of Health GrantsRO1 GM37254 and RO1 GM48123-01 to J.F. and National Institutesof Health Predoctoral Training Grant GM07185 (F.W.S.).

1. Guschlbauer, W., Chantot, J.-F. & Thiele, D. (1990) J. Biomol.Struct. Dyn. 8, 491-511.

2. Williamson, J. R., Raghuraman, M. K. & Cech, T. R. (1989)Cell 59, 871-880.

3. Sundquist, W. I. & Klug, A. (1989) Nature (London) 342,825-829.

4. Sen, D. & Gilbert, W. (1988) Nature (London) 334, 364-366.5. Sundquist, W. I. & Heaphy, S. (1993) Proc. Nati. Acad. Sci.

USA 90, 3393-3397.6. Awang, G. & Sen, D. (1993) Biochemistry 32, 11453-11457.7. Fang, G. & Cech, T. R. (1993) Cell 74, 875-885.8. Walsh, K. & Gualberto, A. (1992) J. Biol. Chem. 267, 13714-

13718.9. Weisman-Shomer, P. & Fry, M. (1993) J. Biol. Chem. 268,

3306-3312.10. Liu, Z., Frantz, J. D., Gilbert, W. & Tye, B.-K. (1993) Proc.

Natl. Acad. Sci. USA 90, 3157-3161.11. Chung, I. K., Mehta, V. B., Spitzner, J. R. & Muller, M. T.

(1992) Nucleic Acids Res. 20, 1973-1977.12. Smith, F. W. & Feigon, J. (1992) Nature (London) 356, 164-

168.13. Macaya, R. F., Schultze, P., Smith, F. W., Roe, J. A. &

Feigon, J. (1993) Proc. Nati. Acad. Sci. USA 90, 3745-3749.14. Smith, F. W. & Feigon, J. (1993) Biochemistry 32, 8682-8692.15. Schultze, P., Smith, F. W. & Feigon, J. (1994) Structure 2,

221-233.16. Schultze, P., Macaya, R. F. & Feigon, J. (1994) J. Mol. Biol.

235, 1532-1547.17. Wang, Y., de los Santos, S., Gao, X., Greene, K., Live, D. &

Patel, D. J. (1991) J. Mol. Biol. 222, 819-832.18. Wang, Y. & Patel, D. J. (1992) Biochemistry 31, 8112-8119.19. Aboul-ela, F., Murchie, A. I. H. & Lilley, D. M. J. (1992)

Nature (London) 360, 280-282.20. Cheong, C. & Moore, P. B. (1992)Biochemistry 31, 8406-8414.21. Wang, K. Y., McCurdy, S., Shea, R. G., Swaminathan, S. &

Bolton, P. H. (1993) Biochemistry 32, 1899-1904.22. Gupta, G., Garcia, A. E., Guo, Q., Lu, M. & Kallenbach,

N. R. (1993) Biochemistry 32, 7098-7103.23. Wang, Y. & Patel, D. J. (1993) J. Mol. Biol. 234, 1171-1183.24. Wang, Y., Jin, R., Gaffney, B., Jones, R. A. & Breslauer, K. J.

(1991) Nucleic Acids Res. 19, 4619-4622.25. Jin, R., Gaffhey, B. L., Wang, C., Jones, R. A. & Breslauer,

K. J. (1992) Proc. NatI. Acad. Sci. USA 89, 8832-8836.26. Wang, Y. & Patel, D. J. (1993) Structure 1, 263-282.27. Kang, C., Zhang, X., Ratliff, R., Moyzis, R. & Rich, A. (1992)

Nature (London) 356, 126-131.28. Padmanabhan, K., Padmanabhan, K. P., Ferrara, J. D., Sad-

ler, J. E. & Tulinsky, A. (1993) J. Biol. Chem. 268, 17651-17654.

29. Feigon, J., Smith, F. W., Macaya, R. & Schultze, P. (1993) inStructural Biology: The State of the Art, Proceedings of theEighth Conversation, State University of New York, Albany,NY, 1993, eds. Sarma, R. H. & Sarma, M. H. (Adenine,Schenectady, NY), Vol. 2, pp. 127-136.

30. Oka, Y. & Thomas, C. A., Jr. (1987) Nucleic Acids Res. 15,8877-8898.

31. Acevedo, 0. L., Dickinson, L. A., Macke, T. J. & Thomas,C. A., Jr. (1991) Nucleic Acids Res. 19, 3409-3419.

32. Scaria, P. V., Shire, S. J. & Shafer, R. H. (1992) Proc. Natl.Acad. Sci. USA 89, 10336-10340.

33. SklenM, V. & Bax, A. (1987) J. Magn. Reson. 74, 469-479.34. Kumar, A., Ernst, R. R. & Wfithrich, K. (1980) Biochem.

Biophys. Res. Commun. 95, 1-6.35. Davis, D. G. & Bax, A. (1985) J. Am. Chem. Soc. 107,

2820-2821.36. Marion, D. & Bax, A. (1988) J. Magn. Reson. 80, 528-533.37. Sklendf, V., Miyashiro, H., Zon, G., Miles, H. T. & Bax, A.

(1986) FEBS Lett. 208, 94-98.38. Wfithrich, K. (1986) NMR of Proteins and Nucleic Acids

(Wiley, New York).39. Lu, M., Guo, Q. & Kallenbach, N. R. (1993) Biochemistry 32,

598-601.40. Saenger, W. (1984) Principles of Nucleic Acid Structure

(Springer, New York).

10550 Biochemistry: Smith et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0