Pure & Appi. C7zem., Vol. 56, No. 8, pp. 989—1004, 1984. 00334545/84 $3.OO+O.OOPrinted in Great Britain. Pergamon Press Ltd.

©1984 IUPAC

DNA FROM A-Z - A SURVEY OF OLIGONUCLEOTIDE STRUCTURES

Olga KennardUniversity Chemical Laboratory, Lensfield Road, Cambridge, CB2 1 EW, UK.

Abstract - The paper discusses the aDplication of single crystal X-ray analysis tosynthetic DNA fragments and surveys oligonucleotide structures Dublishedbetween 1 978 and 1 984. It focuses on the effects of base sequence on localconformation, on the influence of the base stacking and the role of water instabilising global conformations.

INTRODUCTION

Periodically, in different areas of science, a single brilliant idea emerges, which reduces thecomDlexity of experimental observations to one basic concept and provides an insight into themechanism of a wide range of disparate phenomena. A state of euphoria usually follows such adiscovery while the new concept is applied to an even wider range of phenomena only to reveal,ultimately, a higher level of complexity, needing perhaps a new simplification.

The double helical structure of DNA roosed by Watson and Crick in 1953 [1] provided such aunifying concept to molecular biology. It lead to the discovery of the genetic code, to the verysimple mechanism of the transmission of genetic information by complementary hydrogen bonding ofbase-pairs and to an explanation of mutations as a result of errors in the reading of the geneticcode. During the next thirty years, as these ideas were explored in depth, more and more becameknown about the complexity of the biological mechanisms controlling these events. Many of thecontrol functions were found to be mediated by proteins and enzymes which appear to recognise andinteract with specific regions of DNA, characterised by conserved or semi-conserved base sequence.

Indeed, even without a full understanding of the molecular basis of these interactions it has beenpossible to use enzymes, particularly those involved in DNA cleavage and repair to determine theprimary structure of DNA's of a variety of organisms and ultimately to create artificial mutations.A whole new field of genetic engineering was opened u with widespread scientific, medical andsocial implications, of a complexity hardly anticipated in the early days following Watson's andCrick's discovery. One of the keys to understanding this new complexity lies in the detailedsecondary structure of DNA and the influence of specific base sequence on local conformation.

The original model for the double helix was based on X-ray fibre diffraction patterns which, becauseof the inherent disorders of the molecules, and lack of resolution, only indicate gross structuralfeatures, and certainly not conformational details at an atomic level. An added limitation of themethod is that it can not be used to derive a structure directly, but only to test hypotheses andmodels. Even then there may be difficulties in rejecting a wrong model. Although improvedtechniques for preparing fibres and computer fitting of models gave improved structural parameters[2] there was little change in our view of DNA as a monotonous right-handed helix, withconformational versatility limited to a few distinct families, principally the A and B-types.

These same limitations do not apply to X-ray diffraction from single crystals of homogeneousmolecules. With single crystals it is possible to derive the structure of molecules at atomic or nearatomic resolution. The technique was used with outstanding success for the analysis of the

989

uq WTUOL (w) gLOOAG2•

ET 1' 2LflCflLG 0 PDVV LOW LG cooqiuEiG2 (g) 2iqG ATGM (p) ATGM 2[JO/i\TUg W90L (vi)

(g) (p)

GdflJ qGbfli (T J)

flUL[OLWJX 5 GUqO' fiG LG2flJU JGTTX LGJTAGJX 2dng' itp wo uq WTUOL LOOAG2 o uGLJ?bGL UflCJGOJG O -'\J uq g bpO2bp9G tDpO2bpG 2GbgLg1oU O ov jG 2ngL COUOLWJOU 12uq jocGq ou fiG IJGITX gxi' JJJGLG LG J 2G bgii IOL g flJJ pGJTCJ IOLW PATfi gil xTj L!2GpnwqiX oiw o D1V uq DVV Tu ATAoe 12 CJLCGLT2Gq p)' p2G bgu MpTcp LG bGLbGqTcciJLjG OLTgiugJ wOqGJ OL fiG COUTU11O112 p XbG pGJTx fiG COUIOLWTOU 22OCTGq M14p fiG P!P

2JJfflCJS11SV WODEf? - P VI4D V DMV

o bnpjTciougicp o ijj uq pGTL coJJGnG2 /?po 2jiLG #Tp (12 fJGTL LG211J2 OAGL fiG XGL2 OIJGU !uLonb2 o JpoLoLTG2 uq i w LgGnJ bLOG22OL jçr D!CGL2OU g iicr'v uq bLOG22OL y211qTG2 O fiG 2LflC11LG O DIAV LggWGU2 fiG MOLt< O G qG2CLTpGq /t2 cgLLTGq O11 pX fJLGG

CLX2JJOLbpGL21 iu coJJpoLTou Mifi bpA2JcJ uq oLuTc cpGuT22 o Gwb pp LG2oJnToubLoqncTou o qGTuGq 2GdnGucG2 oi rib o bbLOXTWGJX 5 2G bg1L2 fiG itX M2 obGuGq OL2XufiG2G2 bLTC11JLJX fiG TuLoqncTou o 2oJTq 2nbbol4 GcpuTdnG2 [] JJOMG fiG LGJTAGJX LbTq

dnguTTG2 uq bnLWG2 LGdIITLGq oL CLX2JJT29TOU GXbGLTWGU2 EIILfiGL !WbLOAGWGI42 JU C[JGWTC9J

WGfiO O oJToUnCGoTqG 2XUfiG2T2 [] upjq fiG bLGbgLgTOU O 2iOI4 JGUfJ2 O DVV !ULWGU2 OL 2XUfiGTC boJXUflcJGoTqG2' Il fiG wTq-J,2 POMGAGLa fiG qGAGJobwGu O fiG LTG2GL

flifoLnuGJX i UO GGU bo22TpJ& 20 IL 0 OpTU 2TUJG CLX2J2 L0w GTfiGL unij DL'IV

LGJX O 011L 11UqGL2UqTU O fiG CpGW!2LX 0 pToJoiCJ bL0CG22G2'

g bLoor1uq CpUG Il 01W ATGPA 0 fiG 2L11C11LG guq 11UCT0U 0 WCL0W0JGC11JG2 uq c0ULTpnGq

priuqiq bLoGTu2 uq g 1GM ATLI12G2 uq LU2GL g\J2' G2flJ2 01 fiG2G 2nqTG2 iAG pionp gporLflCflLG 01 wuX fi0112U2 01 oigguc uq piojoicjjX Twbor4gu W0JGC(1JG2 TucJnqJu 2GAGLJ

0' EW4VJ�D

COLJLWG pA JG 2JuJG cLA2J 9u9JA2G2 qG2cLTpGq Tu pG ojjoitiu 2GCTOu2

2C<TtJ WG JGXTpTJT4A O 9C[<OUG uq /\LTpJG 2flL COLJOLWTOU p9/G TuqGGq GGLJ2G oLTGLJcgTou uq OL aGLTc LG2OU2 Ju flJG 211L couoLwToir JJJG TwboL4gucG O 2GJJJG IJGX!pJG PCI<POLJG GU gqotD2 WG JOMG2 GuGLA COLJOLW!OLJ TucjnqTu CpUG2 lU JG 2flLT2 WOUG JT[<GJA O pG LGJG O WG LGJTAG OLTGUgTOU uq 2gc1<TUJg O p2G2 JU JG MO TOLW2'

1MW 2JJOM GUGLA pLL!GL2 G2lWGq O p 2 JOM 2 OUJG CJ [c]' jG nuqwuj qTIJGLGLJCGOMGAGL G Gwbp2T2Gq 4J JJG 2flL COUOLW!OU lU qGOXAOJTOLJ11CJGOJqG2 12 LGJ!AGJA TGXTPJG

GMGGU 'j uq p OLUJJ2 lu GLW2 o pG C Guqo C uqo 2n9L coUqoLwioUr i 2ponjqflUTJ JJG AGUJ o 2TUJG cLAacJ gUgJA2T2 GLG M92 Cou2iqGLgpJG Gwbp2i2 ou IpG qJIJGLGuCG2

HbU6 b!WQ1u6H hnLu6H Fb('LU6 XLtuiiqKJ6_2 3 2 3% 2, lu v uq p AbG qonpj pGJix2 3 2 3., 2 3.

bflLU6 — —flLiU6 ALiuJqu6- HbnLIu6 )Lwque hi 2G 2gC1<iug QAGUJb2

bnLiuG2 OCCI1L2 9 bALiwiqiuG-bnLiuG 2Gb2

(hJv) JJGLG 12 joa uo bALiwiqTuG-rALiwiqiuG 2gC1<iUJg pft 21UTTjCLU iIuGL2Luq OAGUJ1D Ouq COUL22 MTJJ WG booi 2gc<iug 2GGU g bALiwiqiuG-bALiwiqiuG uq bALTwiqTuG-bnLTuG I'J V-DJAV

TuL92L9uq 2gc1<iug 12 0p2GLAGq: 2G OAGUJtD lU bflUiuG-bALTwiqTuG uq btlUiuG-bflUiuG 2Gb2 J2 ooq

JJGJTCG2 MT4JO1I LGL O JJG gcnj bflUiuG-tDAUTwiqiuG bUTwgLA 2GdnGucG2 ju P-DJAV (ip) oujAlU JJG 2gc[<iu bgGLu2 lu v uq P-DJAV GAGUJ MJJGU G2G bgGLu2 LG LGaRTcGq o W0UJ00U0fl2

2G agcj<iug 12 OIJG o JJG uJgiu 0LCG2 2piJT2iu DIAV C0LJT0LWi0L JJJGLG LG W9UJ<G qTGLGuCG

5 LflG1rIL6 O V—DMV tLOW rpLe cooLqruea () 2rqG ATG (p) ob(g)

(p)

uq UJLLOM wJoL LooAG uq g LGJTAGJA MTqG uq 2JJJJ0M WJUJOL gLooAG•

22iuGq 2 c Giiqo BiG LG2fljLJ GJTX TJJn2LGq iu Li jrn2 u obGu CGULJ COLG g qGGb

COU2i22 O j J 2G bgiL2 MLflJ g XTJ UT2G bGL LG2IqflG O 3y IjJG 2fl9L COUJ0LWJ0u2 LG gjjqi2bJcGq LOW WG JJGJJX xi uq iucJiuGq o WG gxi2 pA pon 5O jjJG CoJJbJGG jJGJJCJ flLUJ2oJniou pA qqTiou o qGpAqLiu Gu2 ju pG ppi woqGJ O V-DJAV fiG 2G bJL2 9LGP-D1V 12 L9u2oLwGq ° V-D1V M}JGU fiG pnwlqLA o fiG LG GUAILOUWGI4 12 LGqI1CGq OL Ju

D2W LOm y-

992 0. KENNARD

SOME CONSTRAINTS OF SINGLE CRYSTAL ANALYSIS

Experimental limitationsAs discussed in the introduction, single crystal X-ray analysis can be carried to the limit of atomicresolution and it is possible to deduce the structure directly from X-ray data without chemicalassumptions other than the existence of discrete atoms. For such an analysis, however, it isessential to measure the diffracted intensities to a resolution around 1 A or better. This requires

. Single crystals of sufficient size to give measurable intensities to the limit ofresolution - usually at least O.2x0. 1 xO. 1 mm3.

. Moderate thermal vibration of atoms since high thermal motions limit resolution.

. The effect of thermal vibrations can be minimised by taking measurements attemperatures well below the melting point of the crystal.

• Highly ordered arrangement of atoms in the crystal structure, including, if thecrystal is solvated, a well ordered solvent structure.

If the intensities can only be measured to a resolution of around 2.5 the structure can still besolved, usually either by isomorphous replacement, which requires the introduction of one or more

heavy atoms without serious perturbation of the crystal structure, or by phase measurements usinganomalous scatterers. There are also more or less sophisticated trial and error methods which are

based on trying to fit a hypothetical model structure to the experimental observations to obtain a

starting

Structure refinementOnce they phase problem is solved it is possible to use all measured experimental intensities torefine the positions and thermal vibrations of all atoms in the structure. With high resolutionstructures of moderate size the individual positional parameters can be located with an accuracyof around o.oi-o.ooiA since there are usually about 10 experimental points for each variable beingrefined. For large molecules and especially for large molecules of low (-2A) resolution there areproportionally fewer data points and it may be necessary to refine the structure as if it werecomposed of individual rigid molecular fragments. There are several different approaches to suchconstrained refinements and great care must be taken not to inhibit the refinement of a stucturalfeature - not to impose, for example, planarity on a group of atoms which may in reality not be

planar.

Problems with oligonucleotides

Early experiments with oligonucleotides soon indicated that these compounds are extremelydifficult to crystallise. In the three laboratories mainly involved in this work many thousands of

crystallisation experiments were carried out on a large variety of oligomers between 4-12 base

pairs long. Success has so far been confined to the oligomers listed in Table 1, a few otherswhose structures have been solved but not yet published, and a handful for which experimentalmeasurements are available but whose structures could not, as yet, be solved. All the crystals sofar examined are hydrated, (up to 60% by weight) and contain spermine and various counterions

including Mg2, which aid crystallisation.

With the exception of two structures, intensity data could not be measured below 1 .8-2A probably

because of disordered solvents and very small crystal size. The solution of oligonucleotidestructures also presented its own problems with considerable restrictions on the heavy atommethod, which had been so successful in the analysis of proteins. New trial and error methodswere introduced [7] with spectacular success in favourable cases and sad failure, so far, in thecase of oligomers which crystallise with several helices in the asymmetric unit.

gLJq 2flG24G 4G GXT24GLJCG 0 DIAV PGJ!CG2 IAT4 WOLG 4jJIJ OLJG 2G 0TL TU 4}JG LGbGS4 flUJ4

1j4GLUg41U p woqGj M2 fl2G 40 !U4GLOLG4 20W6 DflJ!Ug 2OGCLO2COtflC uq GUX1JJg4TC Op2GLA4!Ou2

OLOOGI4!G2 LGJGC4G TU fiG 2G 24Ct<!U C<0UG G0WG4LX guq 2u1L C0U0LW4T0U JJJ!2

f1UT4 N'Jfi 2fl91. coLJoLw4ToU [1O J1] jG LG2flJ4U4 }JGJTX [J2 2GdflGUCG qGOGUqGU4

GUGL4G Uoe g2 pq T4GI4O GGU 2nbbo2Gq LOW g 2TuJG p2G-bL flUT4 fl4 ILOW o p2G-b9LLG2bGc4IAGjX jG CLX24J 24LflC4flLG JGq 40 fiG 2flgG24!OU 41Jg4 g COU4TUI1OI12 P-DV GJJXconjq GC0U0LWg4T0u2 g22ocTg4Gq /tLflJ 41JG bXLiwTquG 9uq br1LTUG WOTG4!G2 : guq c i5j5jG W024 !Wb0L4Li4 g2bGc4 O fiG 24LflC4flLG /tg2 p0NGAG1) fiG bLG2GuCG 0 4M0 qTGLGU4 2flL

o q(b\fJ/Jy02GLAG !U fiG CLX24J 24LflC$flLG

•ItT4p J4GLu4Tu coUoLwgTou

jPAO 2G DTL WTUT pGJ!X

40WTC LG2OJfl4!OU iq TUqTcg4Gq1 0L fiG fL24 4TW& fiG OLGCT2G G0WG4L? 0 fi!2 32G t)TLTU

G4IAGGU fiG GUUG uq fiXUJ!UG 2G2 0 uG!Jp0flL!U WOJGCflJG2 jG 24LI1C4I1LG /A2 20JAG 40

cLX24gJ 'J 2pOL4 4M0 2G DgTL ILJJGU4 /tg2 IJOMGAG[) OLWG pX 420U-LiC< 4XbG 2G DgTLJUg

qiGLGU4 L0W 04}JGL2 Ji24Gq JU jpJG J !U pMU ouj? OUG J0JGCflJG !U fiG 92?UJWG4LJC (1UJ4 0 fiG

JJJG fL24 2G D!LGq 0Ji0UnCJG04qG 24LflC4I1LG 20JAG ft2 fiG 4G4LJJGL q(b\fjvj) [ ] j4 12

IHE DEWI2E Oh IHE flL.IIhOKV1 DLIV 211fflC1fl1E - .1vriELn'lvlIwc 1DLIVs

cL?24gJJoLbJGL2

4P!2 TU 2OWG qGLG& !2 qGbGUqGU$ OU GAGU CJO2GL C0JJp3L4T0U pG4NGGU CPGW!242 9U

IwbLoAGq JJGfJ02 0 24LI1C4I1LG 20Jfl4T0U VPOAG gjj fJGLG !2 2COOG 0L DLGO9L!Ug G44GL CLX249J2 uq

!U4GU2!4X 2XUCpO4LOU 2011LCG2 40 0p4TU 24L0UGL qTLC4i0U pOaJ fiG AGIX 2WJJ uqC0flU4GL!0J2 conjq pG irnq JJJGLG 12 2COOG IOL 4GCJUTCJ TUUOAg4TOU2 2flC g2 fiG 112G 0 Pp

H0M\GAGL IMfi fiG GXCGD$!OU 0 fiG p!i LG2OJI14TOU q(c() 24LI1C4I1LG UT) LGJ!pJG g22TUJJGU4 0 fiG

pXqL4T0U 2!UCG g pii bL000L4ToU 0 fiG M94GL WOJGCI1JG2 TU fiG CLX24J 24LflC4flLG MGLG JoCg4Gq

fiG jocj uq jopj pGj!Cj C0U0LW410U2 V ooq qj 0 iU0LW4!0U conjq J20 G qGLTAGq ponfiG 24L11C411LG2 #qiiCp pq GGU 20JAG XTGJqGq1 94GL C9LGfiJ LGfUGWGU4 LGJipJG OgLgUJG4GL2 qGfUU

L(ce)q(jv1Vccc) 1'O b313131 1 V [3]q(cccccccc) 3'3 b3131 I V [51]

q(1ccce) slo b3131 I V [33]

q(cccccccc) 3O b 3 V [30]

q(cclVIVcc) 3 V [11](cccc) ro [3]

q(cccccc) 00 b313131 3 S [33]q(ccccVVncccc) ro b313131 s [13]

q(bVjVj) 10 b31 I P []cowboflUq I�G2OII1fOU / D. 5 24LflC4flLG KGIGLGUCG

IVPFE I orIcoI,lncr'E0IIDE2 X-KVA 2IKflCIflKE2 IL2I

DJ4V 4LOUJ y-

994 0. KENNARD

FULL TURN OF A B-DNA HELIX

There is: only one example, so far, of a single crystal study of a B-type helix which was found inthe dodecamer d(CGCGAATTCGCG). The analysis, by Dickerson and co-workers [12] was a verycareful and extensive one and involved several heavy atom derivatives and measurements at anumber of different temperatures [13]. The helix in the crystal structure is right handed withaverage global helical parameters very similar to those derived from fibre diffraction. The helix isflexible and was observed curved (Fig. 5) or straight, depending on conditions [14].

Fig. 5 Two views of the B-dodecamer d(CGCGAATTCGCG).

The local helical parameters calculated from the atomic positions, show twist angles which rangefrom 22-42 degrees, and sugar conformations which varies from C2' endo through C3' endo to 04'exo. The positions of the bases vary with respect to the helix axis with a marked displacement inthe central GAATTC region towards the minor groove side, characteristic of A-DNA.

An interesting feature of the structure is the pronounced "propeller twist" of individual base pairs,which has the effect of improving the stacking of bases along each strand of the double helix. Thevariations in local helical parameters and propellor twist with base sequence have been successfullycorrelated by Calladine's rules [15, 16] which are based on the principle of elastic beam mechanics.According to these rules a balance is achieved between the propeller twist angles, which maximisebase overlaps and the helical turn and base plane roll angles which relieve the steric clash betweenthe large purine bases. Fig. 6 illustrates the agreement between local helical parameters predictedfrom Calladine's rules and those found experimentally.

42

!3I 1' ____28 — i I 0...C:::) I

I 2 3 4 5 6 7 8 9 10 II 2 3 4 5 6 7 8 9 10 II 12C-G-C-G-A-A-T-T-C -G-C-G C-G- C-G-A-A-T-T-C-G-C -G

Fig. 6 Agreement between observed () and predicted (---) local parameters.

In the extended structure the symmetry related molecules are linked by hydrogen bonding betweenminor-groove atoms and water molecules, which will be discussed in the section on hydration.

DNA from A—Z 99

A-DNA IN SINGLE CRYSTAL STRUCTURES

The A-type conformation was found in several oligomers including three octamers, a tetramer anda decamer RNA-DNA hybrid. The conditions for crystallising these compounds were similar tothose used for the B-dodecamer and involved the addition of alcohol, thought to determine the A-conformation. It is not at all clear at present what the factors are which stabilise a particularconformation when double helices form an ordered crystalline array.

The structure of A-DNA in the octamer d(GGTATACC)

A brief description of the analysis of this structure and the results obtained [17, 18] is a goodillustration of the scope and limitation of present day X-ray methods. The octamer wassynthesised by the solid phase method [5] with yields of 8-10mg per preparation. Crystals weregrown by vapour diffusion with 2-methylpentane-2,4-diol (MPD) from buffered solutions containingspermine hydrochloride, barium, magnesium or strontium chloride and MPD. The conditions ofcrystallisation were systematically varied until crystals large enough for diffraction measurementwere obtained. Intensity data was collected on a diffractometer and later extended to higherresolution using a high intensity synchotron source.

The octamer crystallised with one double helix in the asymmetric unit of a hexagonal cell. The

structure was solved by a specially developed, multidimensional search method which combinespacking criteria and best fit (R-factor) calculations for low resolution X-ray data [7]. The uniformA-DNA helix postulated in the trial searches was refined by a constrained-restrained refinementtechnique [19], making maximal use of the experimental observations to determine the variations inthe backbone conformation and the local helical parameters. The analysis is now being extended tomore variables, using high resolution synchotron data.

Fig. 7 shows stereoscopic views of the refined octamer duplex. The most striking differencebetween the fibre model and the actual structure is the opening u of the major groove by nearly2A as a result of the symmetrical bending, by 150, about the non-crystallographic two fold axis.This bend corresponds to a radius of 80A demonstrating, as did the B dodecamer structure, theintrinsic flexibility of the DNA helix. The width of the minor groove remains nearly constant.

Fig. 7 Stereoscopic view of the octamér duplex

996 0. KENNARD

A detailed analysis of the local helical parameters shows systematic variations with base sequencewhich are most pronounced in the central TATA region. This sequence is of particular interestsince it is found at the binding site of RNA polymerase prior to transcription. Fig. S comparesthe base overlaps in the first four steps of the refined octamer with those calculated from auniform A-DNA fibre model. The overlaps at the first two G-C base pairs are very similar to thefibre model. The T-A step shows the interstrand overlap characteristic of A-DNA between theadjacent adenine bases. The greatest difference is at purine-pyrimidine steps G-T. Here the G.Cbase pair is shifted relative to the T.A base pair by nearly 1 A towards the major groove, resultingin a parallel arrangement of the keto group which is energetically more favourable. The variationsin base stacking have the overall effect of evenly distributing the stacking energy along the helixaxis, and can be correlated with the variations in the local helical parameters.

crFig. 8 Projected views of the first four steps of the octamer compared with the equivalent stepsfrom the A-DNA fibre model. The view is perpendicular to mean 1ane through the upper bases

The extended crystal structure is formed by packing the octamer dulexes in infinite spirals aboutthe crystallographic 6i axis leaving large central cavities filled with solvent molecules and(probably) with counterions. Adjacent spirals related by 2i screw axes interleave, formingcontinuous cylinders. The intimate packing of adjacent molecules is clearly shown in Fig. 9 withthe end G.C base pairs parallel to, and in van der Waals contact with, the sugar phosphatebackbone of symmetry related molecules. The structure is further stabilised by a pair ofintermolecular hydrogen bonds. The only other intermolecular hydrogen bonds involve watermolecules and will be discussed in the section on hydration.

Fig. 9 Stereoscopic view of the packing of the octamer duplexes along the sixfold axis.

I

DNA from A—Z 997

Comparison of the A structure

The A conformation was found in a number of crystal structures (Table 2). One of these, theoctamer d(GGGGCCCC) [20] is isomorphous with d(GGTATACC). The octamer d(GGCCGGCC) [21]and the tetramer (d'CCGG) [22] crystallise isomorphously in a tetragonal rather than a hexagonal

space group but using the same packing motif as the hexagonal crystals with large cavities,accomodating solvent molecules. Interestingly the tetramer d('CCGG) behaves as if it were anoctamer with two tetramers stacked on top of one another but with the connecting phosphate

missing. It seems as if in these A type octamers the ability of the end base-pair to stack againstthe shallow minor groove is the dominant factor in stabilising tle extended crystal structure.

TABLE 2 CRYSTAL DATA FOR A-TYPE STRUCTURES

Compound Cell A Volume A3 Volume/base pairA3

a b cd(GGTATACC) 45.01 45.01 41.55 72899 1519

d(GGPrUAPrUACC) 45.08 45.08 41.72 73425 1530

d(GGGGCCCC) 45.32 45.32 42.25 75151 1566

d('CCGG) 41.10 41.10 26.70 45102 1409

d(GGCCGGCC) [-8°C] 42.06 42.06 25.17 44527 1392

d(GGCCGGCC) [-18°C] 40.51 40.51 24.67 40485 1265

r(GCG)d(TATACGC) 24.20 43.46 49.40 51956 1299

The A-type structures form a particularly interesting set for a detailed examination of sequencedependent structural variations. The sequence d(TATA) is contained in both d(GGTATACC) and inthe RNA-DNA hybrid r(GCG)d(TATACGC) [23] thus allowing comparison of the same sequence intwo different crystal environments. On the other hand the two isomorphous oligomers d(1CCGG)and d(GGCCGGCC) contain the recognition sites CCGG and GGCC of the two restriction endo-nucleases Hpa H and Hae H respectively and can thus be compared in the same environment. Fig.10 illustrates the overall changes in the double helix in these crystals compared with the fibremodel. A very detailed comparison of the A-type structure has recently been carried out [24] andfollowed by a quantitative examination of base stacking using atom-atom potential energycalculation [25]. Such comparative studies are important in evaluating the relative contribution ofintramolecular and crystal forces on the stability of helix conformations and in trying tounderstand the rules governing sequence dependent modulations.

Fig. 10 Comparison of space filling drawings of the three A-DNA structures with the fibre model.

uq 20 conjq G TuAoJAGq iu LGflJTu LGbJTcTou uq GXOLG22TOU 0 gGUG2•

pGJTCJ 2GU2G 0 2Gc4Jou2 01 DVV TJj wL[<GqJX gGc 2flDGLCoTJTu Tu TpGL OLGL 22GWpJ1G2GXLCG MTW o\0 gCGTC cTq [)o] pn L2 piojoicj 1flUC1O 12 O2CflLG [fl] pu1u WGpX x-L' qTLcTou pipjX guTgGuTc uq CS1J G GGCG Hi L020bpiJ CLOW02OWG2 /t4JGu:-DIAV P2 PGGU 2IJOMU J2O O GXJ2 111 201n!0u n21u CD uq JJç WG2flLGWGU2 uq Iii ITPLG2

(g) 2TqG MGM2 (p) lob AJGtA2

O IJJG pGXgLJHCJGOITqG q(ccccCc)GxLboJlGq L0W IJG c0oLqTUlG2

hT 1 2ILHCIHLG 01 TDIAV

L00/\G 21q& Mp!Cp ccowWoqlG2 WG GJTX gXj2 J2 qGGb UJLL0M uq puq /t\Tlp b02blG gLOnb2'

coucgAG 2fl14CG /ATW IPG C?$O2TuG C uq WG 11LJHJG uq Ci 0W2 Gxbo2Gq jjJ WTuOL

CpGW1CJ GLJATLOUWGUI2 jG JOLJg uq LGJlTAGJX JTu -pGJ!X }i2 (JO wglOL L00AG pn oujX

(h1 I )) V2 Tu gJGLugTu PD'1V IIJG bpo2bp1G2 Ju 2GdnGucG q(cbc) uq q(cbc) AG qTGLGIJ1

uq qoiiw uq g g cou2GdnGucG WG JIUG couuGc1u }JG bpo2bpoLon2 g0w2 0JJ0IA2 g COIIL2G

J1GLLJ1TLJ 2XLJ uq 91J11 LLgLJGUJGLJ TwbJTG2 1}J9 IPG uGTpp0nLJu 2flL flUTl2 bOlul J1GLU1GJX no

jG 1/?\0 p2G-b9JL flUT1 12 LGbG1Gq 2!X lTwG2 lo ow nJJ nu o IGII puqq pGJTx JJJG

Ju :-DIAV boJX(qc-qc) boJX(q-qc)1LflC11LG 01 g 2G bgTL LT 15 1C[<TU 01 q1cGu1 2G2 Tu JG1 puqqC2,

02

CbC 21Gb PIll oujX TIGL2IL9Uq 21q<Tu 01 luG CX1021uG2

IPG pouq (hT 2G 2gc<Tu J2 nun2nJ (hT isY MTw uo iuuq 2lgc<Tu 9 lUGC0u0Lw1T0u2 91G qTIjGLGul I!t<G Hi 1JG 21Lu1C1u1LG uq 20 12 1G oLTGugTou pon

oLTGugTou hOL bnLTuG2 flJG oLTGugTou T2 2XLJ pIll bXLTwTqTUG2 11 12 gUI!. JJJG 2flLjG LGbGg nUll 01 lUG GJJX C0U21212 01 l/tO 2G bgTL2 //qJTcp qTGL lu lUG 2ngL-p2G

G gccGblGq MTI}J uX couTqGucG 2JUCG uo qG191JGq 2ILflCIflLJ woqj M2 /TJpJG

GxbGLTwGUlJ Lonuq2 bLTcnJLJX ILOW 2oJnITou 2lnqTG2 01 2AUIpGI!c bojX (qC-qC) [5] pn conjq UOI

COJJOLlOL2 [5] ç'GI puqq pGJTCG2 pq bLGATOn2JX GGU bO2IflJIG ou POl1J GOLG1CJ uqlllO 10LW2 01 lUG GILWGL q(CCCC) pX D!Cl<G120U uq COJJG11G2 [5] uq P jçci uq T2

JJ1JJ LG2OJI1IJOU 2InqX 01 IPG GXWGL q(CCCCCc) [5] jJJG 2WG [JGJTX /A2 onuq 2pOLIJX IGL 'U

LqTCJjX qJGLGU1 IGII jiuqq GJJX jG GJTX M2 qT2COAGLGq pX fçTCJJ uq CO-MOLU<GL2 110W g

jG W021 qLwgI!c LG21111 10 p/\G GWGLG 110W 2TUgJG CLX2IJ UJX2T2 01 oJTouncJGolJqG2 12 IPG

DI,1V - V MEJ1 fEEL HVL'IDED HEfIX

0 KEVI�D

(g)

1'

Sjf

DNA from A—Z 999

HYDRATION IN SINGLE CRYSTALS OF OLIGONUCLEOTIDES

All the crystals that have been analysed contain approximately 40-60% DNA per cell volume, theremaining space being taken up by water and counterions. An examination of difference maps atthe end of the refinement of the double helical molecule alone allowed the location of a substantialnumber of water molecules. Typical difference maps showing the location of water peaks in theB DNA [34] dodecamer are reproduced in Fig. 14.

(a) (b): Hr/fN'(11- -:

1'

(c))

(

IE tii1ti''

UI

jl1!!I !.j:/-----I---- '

,;;,

I!-g1)j3—b-----

I

Fig. 1 4 Combined Fourier and Fourier difference maps for the dodecamer d(CGCGAATTCGCG).(a) first stage of refinement (b) last stage of refinement and addition of water molecules.

Water in crystal structures generally forms either monodentate, bidentate or three centred bonds.The hydrogen bondm being an electrostatic interactions, has variable geometry with 0-H... Xdistances of 2.5-3A and 0-H...X angles of 150°-180°. When oxygen acts as a hydrogen acceptor theX-H...0 bond tends to lie along the lone pair directions. Water has a strong tendency to form themaximum number of hydrogen bonds. Thus four-coordinated water is surrounded ny an approximatetetrahedral arrangement of donor and acceptor groups. These guidelines have been used to interpretthe water peaks in the difference maps. In view of the relatively low resolution of the analysis,however, a tolerance of about 0.251k has been allowed in bonf length thus limiting the distancebetween H-bonded atoms to about 3.5A. Atoms at a distance greater than say 3.7A can not, withany certainty, be regarded as part of an H-bonded network. It should also be remembered that thelimited resolution precludes the identification of all molecules in the crystal structure especially thedisordered water in, fr example, the large cavities of the A-type crystal. In these spaces anumber of alternative water networks may exist and the X-ray analysis gives a time averaged viewof several superimposed structures.

Infra-red spectroscopic studies [32, 33] indicate that the primary hydration shell around doublehelical DNA contains at least 11-12 water molecules per nucleotide, rising to about 20 at a relativehumidity of 80%. The phosphate oxygen atoms have the highest affinity for water, followed by thephosphodiester and sugar oxygen atoms. Above 65% relative humidity the functional groups of thebases are also hydrated, the entire primary hydration shell forming a relatively ordered betworkwith about half the water oxygen atoms hydrogen bonded to the DNA molecule and the remainingforming cross links. The crystal structures offer an excellent opportunity of comparing thehydration of the various atoms in different types of helices and correlating these with solutionstudies. An examination of the extended water structure in the unit cell may also give anindication on the partial order imposed on an aqueous environment by the hydrophobic andhydrohilic atomic grouping in the double helix.

1000 0. KENNARD

Hydration in s-DNA - the dodecamer

The number of water molecules which could be located with certainty in the B-dodecamer wasdependent on experimental conditions. At room temperature 72 peaks could be assigned (of whichsome were thought to be spermine) [34] and a further 48 appeared in difference maps when thestudy was carried out at 1 6°K or when the MPD concentration in the crystallisation mixture wasincreased to about 60% [35]. Under these conditions between one and three water molecules arefound around nost phosphate oxygens, of which about 40% coordinate to both phosphates oxygens.Very few of the phosphate ester atoms are hydrated.

The most striking feature of the B-DNA hydration is the ordered water structure found in the minorgroove and particularly in the central AT rich region. As illustrated in Fig. 15 water moleculeslink the functional groups of adjacent steps on the two strands. These water molecules are in turnlinked by further water molecules and a chain of nine linked water molecules extends along thecentre of the minor groove. The network is further stabilised by coordinating with the sugar 04'atoms. This primary hydration shell becomes less well ordered towards either end of the helix. Ahigher hydration shell links with the Primary shell and involves some of the phosphate oxygenatoms.

Fig. 15 Stereo diagram of the minor groove hydration in the GAATTC segment of the B-dodecamer

The water structure in the major groove is discontinuous and much less regular than in the minor

groove (Fig. 16). Only relatively few water molecules were located at distances of 3.5A or lessfrom the functional groups and only two of these bridge functional groups of the bases betweensteps directly. A second and equally irregular hydration shell links some of these water moleculesand also the phosphate oxygen atoms Pointing into the major groove.

Fig. 16 Stereo diagram of the major groove hydration in the AT rich segment of the B-dodecamer.

DNA from A—Z 1001

Hydration in A-DNA

A large percentage of solvent molecules were located in the A-DNA structures. Some details of thewater structure were published for the tetramer d(1CCGG) [22] and the analysis of the hydrationnetwork in d(GGBrUABrUACC) has recently been completed [36].

In the octamer virtually all the phosphate 01 and 02 atoms are hydrated. Neighbouring 01 atomsalong the backbone are frequently bridged by one water molecule but none of the water moleculesare linked to the two oxygen atoms of the same phosphorous atom.

The most remarkable feature of the first hydration shell is the ribbon of water molecules extendingalong the centre of the major groove (Fig. 17). This ribbon is formed by pentagonal arrangementsof five water molecules, each tetrahedrally coordinated either to the functional groups of the basesor to the water molecules which in turn are linked to the phosphate 01 atoms pointing into themajor groove. The network has pronounced two-fold symmetry, particularly in the central TATAregion. The network becomes less regular towards either end of the groove. Pentagonalarrangements involving water were observed in small molecule structures, notably in a complex ofd(CpG) and proflavine [37] and in hydrated cyclodextin [38, 39].

Fig. 17 Stereo diagram of the hydration in the major groove of the ATA region of the A-octamer.

The minor groove in d(GGBrUABrUACC) is less accessible to solvent molecules than the majorgroove because the end pairs of one helix are stacked in the minor groove of a symmetry relatedhelix. Nevertheless, most of the functional groups in the minor groove are hydrated (Fig. 18). Thenetwork of water molecules is much less regular than in the major groove and the ordered chain ofwater molecules, observed in the AT rich region of the minor groove of the B-dodecamer, is notseen. However, there is a well defined chain of water molecules in the central ATA region, wherehydrogen bonds are formed to the bases of opposite strands and also to the sugar 04' atom pointinginto the minor groove. Less regular networks cross link the functional group of the bases at eitherend of the octamer.

Fig. 18 Stereo diagram of the hydration in the minor groove of the A octamer.

1002 0. KENNARD

In the other A-type structure d(1CCGG) [22, 40] where the water structure was analysed in somedetail, the hydrogen bonded network in the major groove is much more irregular than ind(GGTATACC) possibly because of the intrusion of the terminal OH groups after only four basepairs along the helix. The primary hydration shell consists mostly of local clusters of watermolecules coordinating the functional groups and providing intrastrand links. There is only onehydrogen bonded network stretching from the phosphate oxygen of one strand via various functionalgroups to a phosphate oxygen atom of a symmetry related molecule. In the minor groove very fewof the functional groups are hydrated, possibly again because of end effects.

In the octamer d(GGCCGGCC) eighty water molecules were located mostly in the large solventchannels near the major groove. No detailed analysis of the water is as yet available. The

position is similar for the hydration in the Z-structures.

It has been suggested that the regular water molecules in the minor groove of B-DNA stabilise theconformation. On dehydration this regular structure is disrupted and only the highly polar phosphateoxygen atoms remain hydrated. This induces a conformational transition and the A-DNA form isstabilised by the hydration in the major groove. The very regular ribbon of water molecules foundin the A-octamer structure lends support to the view that the A-conformation is indeed maintained

by the hydration of the major groove.

SUMMING UP

The behaviour of DNA in vivo is very complex and becomes apparently more complex as it isinvestigated in detail. The same seems to hold true for the structure of DNA. The handful ofsingle crystal analyses of deoxyoligonucleotides published in the last few years have made us realisethe structural flexibility of DNA, the existence of left handed as well as right handed helices,helices generated from a variable number of repeat units and variable local conformations. We areready to accept that the classical B-structure is but one of the conformations adopted in vivo andhave ourselves become more flexible when interpreting biochemical, as well as physiochemical,experiments on DNA in structural terms.

The search for unifying principles and predictive rules is one of the driving forces behind mostscientific investigations. There are already tantalising glimpses of such principles, which mightaccount for some of the observed structural variability. Local conformational changes in particularseem to be governed by base stacking and the tendency for equipartitioning the stacking energyalong the entire helix. The different hydration patterns observed in the B and A type structureshave suggested some theories about the role of water in mediating conformational transitions.However, these generalisations are based on structural data from just a few short, self-complementary DNA fragments. We need to analyse many more deoxyoligonucleotides of varyinglength and base composition, selected to answer specific questions. New information will also come

from the rapidly expanding field of single crystal analysis of DNA-protein and DNA-drug complexes,and the wide variety of solution studies particularly high resolution nuclear magnetic resonancemeasurements. It seems a safe prediction that this wealth of data will lead to a new simplificationin our understanding of the mode of action of DNA in biological systems.

ACKNOWLEDGEMENTS

I thank my colleagues Dr. T. Brown, Dr. G.G. Kneale, Dr. M.J. McCall, Dr. W.B.T. Cruse, Mr.1. Nachman, Professor D. Rabinovich, Dr. S. Salisbury, Dr. Z. Shakked and Professor M.A. Viswamitrafor their collaboration during the period covered by this survey; Dr. M.3. McCall and Dr. S. Salisburyfor their helpful comments on the manuscript and Mrs. S. Berry for all her work in preparingthe camera ready copy. Financial support was provided by the Medical Research Council.

(O6L) OZOlr9lOli 'LL PS PPV N D0Jd '1PTI V pirn JJJIpsJ UA J WOO UA H'C "I!O E0 'U1 EHV '>I1T0N 'Ei 'PJ0JM 11

(os6L) Lc-L9c'9z (pUO) IflN 'UOSJ>pTQ pu Jfl>J ')[ ')jU 'S '0>IJ. 'I "-'U 'I-I

'(6L61) 99O9 ' (puofl '1Pi1 V

pirn jJJJ\J ip UA "3 'W00 UPA 'pOJiJ3 '>c1IoN 'C'l "TTO 'C'3 '1i 'E'I-I'V 9Z

'(61) uoIDrqnc1 J0J pwqfl5 'Pa>PPs 'Z pu z 'urn lEt'! 'c '(16t) ssid Uj 'AN 'suo pu AJTJ UqOC

'>UJfl0E 'J pu U0SJiJdD 'V 'spa 'Aojoi 'PUUN '0 pu P>PI11S z '(Z6l) ?O9-LO9 ' (pU0)

"-P!1 V PU jD0 UA 'V'V'S 'I1A1 p UA 'V"J 'W00 UA H'C 'T!1nJ 'S 'UJ 'C'H'V '(z61) 66Z fl6Z 'c6z (pU01)

'U0SJ>jDiQ J>J pu >lI 'N ')iU,L 5 '0U)jJ .1 'UU03 'N'c3

'(Z61) ZL696 '6L 'V'S'rl 'PS 'PDV 'Wd 'ipi 'V W00 UA 'C '!!nJ 'S 'U 'C'H'V

'(fl61) pL1sqnd q oj 'pJ?UUN '0 PU IID v'ctt '(LL6L) flOFi00 'JC V S J0 JSJJ3 DV

'WIN 'V'S 't4DJflq3 '\'rj 'UJJj 'jJ '>joqJo 'WS 'UWSSfl5 'TI'S

'(c61) 1OZ-1 '991 'T!I T01, ' 'TwMsiA 'V't, pU iCrnqsrJ5 'V'S 'PJUUN '0 ' 'sni- 'j'' 'ipiAoUiq 'ç 'p>pjq5 ' '(161) L1r6Lfl 'L 'D05 '1coj 'DOJd

'JTwMsTA 'V'1' PU Anqsiji 'V'S 'SflJT3 'I'i ' PJUUN '0 'qDIA0UTq '0 'P>PI1IS 'Z 'L I

'(c61) çflfl-fl '991 'I°'1 'I°IAJ 'C 'U0SJ>1JT0 'IEI')l '91

'(z61) cçl—JflcJ '191 'joicj 'joj 'C 'UTPT1) "?I'D 'ci '(61) L0L1i1991i1 'Lc 'Wt43 'f0rcJ 'C 'U0S>PtU '1E1'J pU MJQ 'J' '>jd0N ''J 'rUiJd 'A'V 'fit

'(Z6I) flfl0r0fl0fl '6L VSF1 'PS 'PV 'N 'DOJd 'U0SJTU ''J JU UOSWP5 '5 'AJQ 'J'J-4 '[ '(061) cL-ccL 'LS (pUoTI)

'U0SJJiQ 'lE1'1 DU fl>II 'N 'P)Uj '5 '>j0J '3 '0UPj '1 'M1Q ' 'UTj '1 '(6L61) 09-669 'ftl 'JoIcj 'JOy\i 'C

'ZT5 'V'J. PU P>PII1S 'Z 'PJUUN '0 'UJWMSTA 'V' '>PC 'V '9IN 'V '(L61) Z6Z-6Z '6 'PS

'0JAjJ ' pU AJnqsTJ5 'V'S '>pJppq5 '"3 'SU0c 'D'd 'P>PP-IS 'Z 'PJUUN '0 'flTWMSTA 'V'l' '01

'(z86 1) zc- i ç ' SJWXf0d0rcJ

'PJUUN '0 DU Xnqsrj 'V'S '>P!PI1S '"3 'SU0f 'O'd '>f>j15 'Z 'flTWASTA 'V' '(L6 1) 9L69 'LZ (pUoTI)

'0JJAJJ 'TI pU Anqsij 'V'S '>puppq5 '1t'D 'SU0C "3'd 'PUUN '0 'JTW\SIA 'V'IAI

(fl61) SSU UI 'V 'SAfl 'P>PI1IS 'Z qDIA0UTqJ 'ci '(L61) l19Z-L09Z '001 'Dos 'Wq3 'WV 'C 'TqSJ 'V TATI 't,

'(61) flZ9fl9 '01 'S)J PPV 'IDflN 'SWTj '3's pU oc15 'S'c3 'q9Ui5 ''J 'U'í'H '!D 'C'

'(9L61) 199-cc9 ' 'D05 'W)q) 'JWV 'C 'pOJSUflTI 'J'j )U JaU!STI 'T1 (q

'(161) 19961 'Z 'flfl '1 'Sfl) 'H'AJ pu D1W 'TI'S ( '(LL6I) '1L—lc 'i 's PPV 'PAN

'fl 'J UP N 'd'3 'U2JN 'V'S 'TWnzoJ4 '1 'ic>JSU!AJ5 'C

'(3L61) 60c1-flOcI 'Lfl SU0!DTUflWW03 DJSJ jisAqcior pUP JPDTWipOT 'SUT)jflH 'TI'i'U DUP OUJV 'S

'(cc6I) 01iLLL ' (pUOTI) '>P!D '3'H'd pUP U0SPJ 'I'C SI?JlEL1lElF8

:00I z—v woij vci

0. KENNARD 1004

29. F.M. Pohi and T.M. 3ovin, J. Mo!. Biol. 67, 375-396 (1972).

30. R.M. Sandella, D. Grunberger, I.B. Weinstein and A. Rich,Proc. Nat. Acad. Sci. U.S.A. 78, 1451-1455 (1981).

31. R.3. Hill and B.D. Stollar, Nature (Lond) 305, 338-340 (1983).32. M.3. Falk, K.H. Hartman, 3r and R.C. Lord, 3. Amer. Chem. Soc. 85, 387-391 (1963).33. M. Falk, A.G. Poole and C.G. Goymour, Can. 3. Chem. 48, 1536-1542 (1970).34. H.R. Drew and R.E Dickerson, 3. Mo!. Biol. 151, 535-556 (1981).

35. M.L. Kopka, A.V. Fratini, H.R. Drew and R.E. Dickerson, 3. Mo!. Bio!. 163, 129-146 (1983).36. 0. Kennard et. a!. In preparation.37. 5. Neidle, H.M. Berman and H.S. Shieh, Nature (Lond) 288, 129-133 (1980).38. W. Saenger, Nature 279, 343-344 (1979).39. W. Saenger and K. Lindner, Agnew. Chem. mt. Ed. EngI. 19, 398-399 (1980).40. R.E. Dickerson. personnal communication.

Figs. 1, 2, 3, 5, 11, 12, 13 were reproduced from W. Saenger Principles of Nucleic Acid Struture,Springer Verlag, New York, 1984 and Figs. 6, 14, 15, 16 from 3. Mo!. Biol. by kind permissionof the Editors.