The conformations of locked nucleic acids (LNA)

The conformations of locked nucleic acids (LNA)

Michael Petersen,1 Christina B. Nielsen,1 Katrine E. Nielsen,1 Gitte A. Jensen,1

Kent Bondensgaard,1 Sanjay K. Singh,2 Vivek K. Rajwanshi,2 Alexei A. Koshkin,2

Britta M. Dahl, 2 Jesper Wengel2 and Jens Peter Jacobsen1*1Department of Chemistry, University of Southern Denmark, Odense University, DK-5230 Odense M, Denmark2Center for Synthetic Bioorganic Chemistry, Department of Chemistry, University of Copenhagen, Universitetsparken 5, DK-2100Copenhagen, Denmark

We have used 2D NMR spectroscopy to study the sugar conformations of oligonucleotides containing aconformationally restricted nucleotide (LNA) with a 2'-O, 4'-C-methylene bridge. We have investigated amodified 9-mer single stranded oligonucleotide as well as three 9- and 10-mer modified oligonucleotideshybridized to unmodified DNA. The single-stranded LNA contained three modifications whereas theduplexes contained one, three and four modifications, respectively. The LNA:DNA duplexes have normalWatson–Crick base-pairing with all the nucleotides inanti-conformation. By use of selective DQF-COSYspectra we determined the ratio between the N-type (C3'-endo) and S-type (C2'-endo) sugar conformationsof the nucleotides. In contrast to the corresponding single-stranded DNA (ssDNA), we found that the sugarconformations of the single-stranded LNA oligonucleotide (ssLNA) cannot be described by a major S-typeconformer of all the nucleotides. The nucleotides flanking an LNA nucleotide have sugar conformationswith a significant population of the N-type conformer. Similarly, the sugar conformations of the nucleotidesin the LNA:DNA duplexes flanking a modification were also shown to have significant contributions fromthe N-type conformation. In all cases, the sugar conformations of the nucleotides in the complementaryDNA strand in the duplex remain in the S-type conformation. We found that the locked conformation of theLNA nucleotides both in ssLNA and in the duplexes organize the phosphate backbone in such a way as tointroduce higher population of the N-type conformation. These conformational changes are associated withan improved stacking of the nucleobases. Based on the results reported herein, we propose that theexceptional stability of the LNA modified duplexes is caused by a quenching of concerted local backbonemotions (preorganization) by the LNA nucleotides in ssLNA so as to decrease the entropy loss on duplexformation combined with a more efficient stacking of the nucleobases. Copyright# 2000 John Wiley &Sons, Ltd.

Keywords:NMR spectroscopy; DNA; locked nucleic acid; LNA; modified oligonucleotide

Received 13 July 1999; revised 15 September 1999; accepted 30 September 1999

INTRODUCTION

Chemically modified DNA and RNA oligonucleotides haveattracted much attention due to their potential use inantisense strategies. To be considered as a potential geneinhibitor certain requirements must be fulfilled, including anenhanced stability towards cellular nucleases, the ability topenetrate the cell membrane, an efficient hybridization tothe target DNA and activity towards RNaseH (He´lene andTouloue1990; Uhlmann and Peyman 1990). Oligonucleo-tides have been modified in the phosphate linker, the

nucleobase and the sugar moiety in an attempt to achievethese properties. At present, the most successful modifica-tion is the phosphorthioate where one of the non-bridgingoxygens in the phosphate linkage is replaced by sulfur.Antisense compounds of this type have been approved forclinical use (Flanagan, 1998). A second generation ofantisense drugs consisting of a mixture of phosphorthioateand 2'-alkyl modified nucleotides has entered clinical trials(Flanagan, 1998). However, in spite of an extensive search,antisense drugs for many important diseases have yetremained elusive. The field still remains very attractive andcombined efforts in organic synthesis, molecular biologyand structural biology are likely to produce significantresults.

Recently, we have introduced LNA (locked nucleic acid)containing a 2'-O,4'-C-methylene linked bicyclic ribofur-anosyl nucleosides (Scheme 1) locked in an C3'-endo(N-type) conformation (Singhet al., 1998a; Wengel, 1998;Koshkinet al., 1998a–c; Singh and Wengel 1998; Singhetal., 1998b). Oligomerization of 3'-O-phosphoamidite LNAmonomers proceeds efficiently on an automated DNAsynthesizer using standard procedures (Singhet al.,

* Correspondence to: J. P. Jacobsen, Department of Chemistry, University ofSouthern Denmark, Odense University, DK-5230 Odense M, Denmark.E-mail: [email protected]/grant sponsor:The Danish Natural Science Research Council; TheDanish Technical Research Council.Abbreviation used: DQF-COSY, double quantum filtered correlationspectroscopy; dsDNA, double-stranded DNA; ssDNA, single-stranded DNA;DSS, 2,2-dimethyl-2-silapentane-5-sulphonate; HSQC, heteronuclear singlequantum coherence; LNA, locked nucleic acid; ssLNA, single-stranded LNA;NOE, nuclear Overhauser effect; NOESY, nuclear Overhauser effecctspectroscopy; TOCSY, total correlation spectroscopy; TPPI, time proportionalphase incrementation; 2D, two-dimensional.

Copyright# 2000 John Wiley & Sons, Ltd. J. Mol. Recognit.2000;13:44–53

JOURNAL OF MOLECULAR RECOGNITIONJ. Mol. Recognit.2000;13:44–53

1998a; Koshkinet al., 1998b; Singh and Wengel 1998).Thermal denaturation studies of duplexes formed betweenLNAs and complementary DNA and RNA show that LNArecognizes both with remarkable affinities. The duplexesinvolving LNAs (hybridized towards either DNA or RNA)display an unprecedented increase of the melting tempera-tures of between�4.0 and �9.3°C per modificationcompared to the corresponding unmodified referenceduplexes. The LNA:LNA base-pairing constitutes the moststable modified nucleic acid type duplex system hithertodiscovered, and melting temperatures greater than 93°Chave been observed for a duplex between two all-modifiedcomplementary LNAs (Koshkinet al., 1998a). Thus,incorporation of a given number of LNA monomers intoan oligonucleotide appears to be a very convenient andpredictable way of improving the stability of duplexestowards complementary DNA or RNA. Preliminary experi-ments (Wengelet al., 1999) suggest the LNA molecule to bethe most promising candidate for efficient recognition(strand displacement) of a given mixed sequence in anucleic acid duplex so far. These results, together with thedemonstration that LNA is stable towards 3'-exonucleolyticdegradation (Singhet al., 1998a) and that fluorescein-labeled LNA can be transferred into living MCF-7 breastcancer cells (Wengelet al., 1999), have stimulated theevaluation of LNAs as antisense and/or antigen molecules.

In general, the conformations of the flexible deoxyriboserings determine the overall structure of a (deoxy)ribonucleic acid duplex. Duplexes in A-type conformationcontain nucleotides with an N-type (C3'-endo, Scheme 2)sugar conformation while B-type duplexes contain nucleo-tides with an S-type (C2'-endo, Scheme 2) sugar conforma-tion (Dickersonet al., 1989; Saenger, 1984). N-type sugarsare described by a pseudorotation angle,P, betweenÿ90°and 90°, while for S-type sugarsP resides between 90° and270° (Altona and Sundaralingam, 1972, 1973; Due to thelow energy barrier between these two conformations[�2 kcal/mol (Brameld and goddard, 1999; Olson andSussman 1982)] nucleotides are not trapped completely ineither conformation, but rather exist in a fast equilibriumbetween them [the N, S two-state model (Altona andSundaralingam 1972, 1973)]. Analysis of the fine structureof cross-peaks in DQF-COSY 2D NMR spectra offers areliable estimation of sugar conformations (Wu¨thrich,1986).

In this work we have used two-dimensional (2D) NMRspectroscopy to study the sugar conformations of a singlestranded LNA (ssLNA) oligomer and the correspondingunmodified single stranded DNA (ssDNA) as well as the

sugar conformation of three LNA:DNA duplexes withdifferent compositions and a different number of modifica-tions. The choices of the duplexes were made in order toinvestigate the locked nucleotide in different environmentswith respect to pyrimidine–pyrimidine, purine–purine andpurine–pyrimidine base pair stacking. The aim of this workwas to examine the conformational changes of DNAnucleotides caused by neighboring or opposing LNAnucleotides. The results indicate that the ssLNA ispreorganized and suggest that the increased stability of theduplexes containing LNA nucleotides may be explained byconformational changes from C2'-endoto C3'-endoof theLNA-nucleotide and the neighboring unmodified nucleo-tides combined with an enhanced stacking of the nucleo-bases.

EXPERIMENTAL PROCEDURE

Sample preparation

The modified oligonucleotides were synthesized asdescribed elsewhere (Singhet al., 1998a; Koshkinet al.,

Scheme 2. N- and S-conformations of nucleotides.

Scheme 1. The LNA nucleotide.

Scheme 3. The numbering scheme of the modi®ed oligonucleo-tides used.

THE CONFORMATIONS OF LNA 45


Ta

ble

1.

Co

up

lin

gco

nsta

nts

(Hz)

for

the

su

ga

rp

roto

ns

inssD

NA

an

dssL

NA

an

dth

ed

eri

ve

dsu

ga

rco

nfo

rma

tio

nsa

ssD

NA

ssLN

A

J 1'2'

J 1,2@

J 2'3'

J 2@3'

%Sd

PN

eP

seJ 1

'2'

J 1'2@

J 2'3'

J 2@3'

%Sd

PN

eP

se

C1b

6.5

6.1

4.0

0.5

84(1

5)18

8(28

)5.

26.

35.

83.

854

(12)ÿ1

2(49

)17

2(42

)T

2/T

L2

8.6

5.7

3.2

0.0

95(9

)18

3(14

)17

G3

——

——

——

—6.

66.

77.

03.

057

(23)

29(7

5)14

0(57

)A

4c

——

——

——

—5.

84.

23.

50.

778

(15)

183(

20)

T5/

TL5

8.1

5.6

0.4

0.0

97(6

)18

6(7)

17A

6c—

——

——

——

4.1

9.4

4.8

6.9

30(1

7)29

(14)

T7/

TL7

8.3

5.8

2.0

0.0

95(9

)18

8(10

)17

G8c

7.9

5.9

5.3

0.6

86(1

3)16

7(32

)—

——

——

——

C9b

6.3

6.5

6.7

ÿ0.2

81(1

9)15

6(51

)9.

36.

16.

21.

585

(15)

145(

33)

aT

heJ 2

'2@

coup

ling

cons

tant

sw

ere

fixed

atÿ15

Hz

inal

lC

HE

OP

Ssi

mul

atio

nsan

dth

epu

cker

ing

ampl

itude

sw

ere

fixed

at38

°in

PS

EU

RO

Tca

lcul

atio

ns.

The

stan

dard

devi

atio

nsar

egi

ven

inpa

rent

hesi

s.b

For

the

ssD

NA

C1

and

C9

mig

htbe

inte

rcha

nged

.c

A4,

A6

and

G8

inth

ess

DN

Aan

dG

8in

the

ssLN

Aco

uld

not

bean

alyz

eddu

eto

spec

tral

over

lap.

dP

erce

ntag

eS

-typ

esu

gar

conf

orm

atio

nca

lcul

ated

usin

gR

AN

DP

SE

UR

OT

.T

hepo

pula

tion

ofth

eN

-typ

eco

nfor

mat

ions

isgi

ven

by10

0ÿ%

S.e

Pse

udor

otat

ion

angl

eca

lcul

ated

for

the

two

suga

rco

nfor

mat

ions

usin

gR

AN

DP

SE

UR

OT

.If

the

mol

efr

actio

nof

aco

nfor

mer

was

calc

ulat

edas

belo

w30

%,

the

corr

espo

ndin

gPva

lue

was

not

cons

ider

edto

bede

term

ined

with

suffi

cien

tac

cura

cy.

The

LNA

nucl

eotid

esw

ere

not

incl

uded

inC

HE

OP

Sca

lcul

atio

nsan

dth

eva

lues

ofP

Nw

ere

take

nfr

omO

bika

et

al.

(199

7).

46 M. PETERSENET AL.


1998b). The purified unmodified oligonucleotides werepurchased from DNA Technology, A˚ rhus, Denmark.Samples of the modified LNA:DNA duplexes, shown inScheme 3, and the corresponding unmodified duplexes wereprepared by dissolving an equimolar amount of the twosingle strands in 0.5 mL of 10 mM sodium phosphate buffer(pH 7.0), 0.05 mM NaEDTA, 0.01 mM NaN3 and 0.1 mM2,2-dimethyl-2-silapentane-5-sulfonate (DSS). The sampledwere heated to 80°C and slowly cooled. For experimentscarried out in D2O, the solid complexes, lyophilized threetimes from D2O, were redissolved in 99.96% D2O (Cam-bridge Isotope Laboratories). A mixture of 90% H2O and10%D2O (0.5 mL) was used for experiments examiningexchangeable protons. The final concentrations of theoligonucleotides were 1–2 mM.

NMR experiments

NMR experiments were performed either on a Varian Unity500 spectrometer or a Varian INOVA 750 spectrometer at25°C unless otherwise stated. NOESY spectra of theduplexes were acquired in D2O using 1024 complex pointsin t2 and a spectral width of 5000 Hz at 5000 MHz. NOESYspectra at 750 MHz were acquired using 2048 complexpoints int2 and a spectral width of 7500 Hz. A total of 512t1experiments were recorded using the States phase cyclingscheme. The residual signal from HOD was removed bypresaturation. The NOESY spectra in H2O were acquired at500 MHz with a spectral width of 10,000 Hz using 2048complex points and a pulse sequence where the last 90°pulse is replaced with a shaped 90° pulse containing a notch

Figure 1. The H2'/H2@/H3' to aromatic region of the 200 ms NOESY spectrum ofthe d(CCGCTLAGCG):d(CGCTAGCGG) duplex obtained at 750 MHz. A solid lineindicates an important part of the the sequential connectivity pathway fromC4H6 ±C4H2'/H2@ ±TL5H6 ±TLH2' ±A6H8 ±C7H2'/H2@.



to avoid excitation of the solvent signal (Steinet al., 1995).This ensured suppression of the water signal together with alinear excitation profile over the whole spectral width.

TOCSY spectra of the oligonucleotides with mixingtimes of 60 ms were obtained in the TPPI mode using 1024complex points int2 and 512t1 experiments. DQF-COSYspectra of the oligonucleotides were recorded using theStates phase cycling scheme with 4096 complex points int2and 1024t1 experiments. In order to enhance the digitalresolution we used a selective DQF-COSY pulse sequencein which the first 90° pulse was replaced by a selectivepulse, hence reducing the spectral width ofF1 to 1000 Hz.

The acquired data were processed using FELIX (versions97.0 and 97.2, MSI/Biosym Technologies, San Diego, CA).The NOESY and TOCSY spectra in D2O were analyzed byconventional methods (Wu¨thrich, 1986; Hareet al., 1983;Scheeket al., 1983; Feigonet al., 1983). The assignment ofthe exchangeable protons was obtained from the NOESYspectrum in H2O (Boelenset al., 1985).

The H1' to H2' and H2@ region of the DQF-COSYspectrum was used as input to CHEOPS to obtainJ-couplinginformation for the H1', H2', H2@ and H3' deoxyribose ringprotons (Macayaet al., 1992).

The deoxyribose conformations were analyzed by use ofPSEUROT assuming a fast, two-state equilibrium betweentwo conformations (N- and S-type). The program uses aleast-squares fit to derive the pseudorotation angles (PN andPS) and the molar fraction of the two states from thecoupling constants obtained from the DQF-COSY simula-

tion. In our randomized routine of PSEUROT (RANDP-SEUROT) we include the effect of uncertainties in thedetermined set of coupling constants. RANDPSEUROTgenerates 1000 sets of coupling constants with a dispersionthat reflects the uncertainty of each determined couplingconstant. Uncertainties of the coupling constants arecarefully assessed by how well the back-calculatedspectrum correlates with the experimental spectrum andby the response of the determined coupling constants tosmall changes in the simulation procedure. Subsequently,PSEUROT calculations are performed on all the couplingconstant sets generated and RANDPSEUROT then calcu-lates mean values and standard deviations of the parametersreturned by PSEUROT. In the statistical analysis N-typeconformations are compared with N-types and S-typeconformations are compared with S-types based on theparameters returned by PSEUROT. In cases where PSEUR-OT returns either two N-type or two S-types conformations,the conformations are sorted according to the majorconformer.

RESULTS

The modified oligonucleotide d(CTLGATLATLGC)

The NOESY spectra of d(CTLGATLATLGC) and d(CTGA-TATGC) in both D2O and H2O show that these oligonu-cleotides occur in single stranded form at 25°C. There was

Table 2. Coupling constants (Hz) for the sugar protons in the d(CCGCTLAGCG):d(CGCTAGCGG) duplex and thederived sugar conformationsa

J1'2' J1'2@ J2'3' J2@3' % Sd PNe Ps

e

C1 7.9 6.2 4.3 ÿ0.5 93(12) 187(22)C2 9.6 5.3 5.2 0.3 92(10) 159(19)G3b — — — — — —C4 8.0 5.6 5.0 1.3 86(14) 168(25)TL5c 17A6 4.3 7.0 3.1 3.5 63(17) ÿ2(20) 197(34)G7 8.0 5.8 5.5 0.9 89(14) 168(31)C8 8.0 6.4 5.9 1.4 80(12) 159(34)G9 7.9 6.2 4.7 1.3 85(11) 183(20)C10 7.4 5.7 2.0 0.0 96(6) 193(7)G11b — — — — — —C12 6.7 6.0 4.1 2.4 85(13) 192(21)T13 7.9 6.3 5.2 1.6 80(14) 159(33)A14b — — — — — —G15 8.6 4.7 4.9 0.3 90(12) 169(23)C16 8.9 6.1 5.9 1.8 79(11) 156(27)G17 9.8 5.4 4.2 0.9 95(7) 170(12)G18 8.2 6.2 4.7 ÿ0.1 91(12) 179(26)

a The coupling constants were derived from the H1'–H2' and H1'–H2@ cross peaks in the selective DQF-COSY spectrum using CHEOPS. TheJ2'2@ coupling constants were fixed atÿ15 Hz. The sugar conformations were calculated using the RANDPSEUROT program with the puckeringamplitudes kept constant at 36°, 38° and 40°, respectively. Only minor differences were observed. The values listed are the ones obtained withan amplitude of 38°. The standard deviations are given in parenthesis.b G3, G11 and A14 could not be analyzed properly due to spectral overlap.c The LNA nucleotides were not included in CHEOPS calculations and the value ofPN were taken from Obikaet al. (1997).d Percentage S-type sugar conformation calculated using RANDPSEUROT. The population of the N-type conformations is given by 100ÿ%S.e Pseudorotation angle calculated for the two sugar conformations using RANDPSEUROT. If the mole fraction of a conformer was calculated asbelow 30%, the correspondingP value was not considered to be determined with sufficient accuracy.



no indication in the spectrum of any strong sequentialconnectivity pattern that indicated the existence of basestacking in ssLNA. A visual inspection of the H1'–H2'/H2@region of the DQF-COSY spectra of the ssDNA (Plate 1)reveals that most of the resolved cross peaks possess atypical S-type conformation fine structure. For S-typesugars large coupling constants are expected for H1'–H2'and smaller coupling constants for H1'–H2@ and vice versafor N-type sugars. An H1'–H2@ DQF-COSY cross peak ofan S-type sugar has a fine structure which splits the peakinto 16 components while the H1'–H2' cross peak has amuch simpler fine structure. For an N-type sugar the H1'–H2@ cross peak has a simple fine structure while the H1'–H2'cross peak is either of low intensity or totally absent due tocancellation of antiphase peaks. Contrary to the spectrum ofthe ssDNA, the ssLNA spectrum (Plate 1) displays no crosspeaks from a dominating S-type sugar conformation exceptfor the terminal nucleotide C9. Interestingly, the twoadenosine cross peaks (A4 and A6) next to each other inthe spectrum have markedly different appearances.

The quantitative determinations of the coupling constantswere performed by simulation of the H1'–H2'/H2@ regionsfrom the DQF-COSY spectra of either the ssDNA or thessLNA. The coupling constants obtained together with thefraction of S-type sugar and pseudorotation angles calcu-lated using our random application routine of PSEUROT areshown in Table 1.

Our analysis indicates that the sugar conformations of the

ssDNA all are in a nearly pure S-conformation with four ofsix analyzed residues having purities of more than�90%(Table 1). Unfortunately analysis of A4 and A6 provedimpossible due to spectral overlap of the H1' protons as wellas the H2' and H2@ protons.

When analyzing the ssLNA a somewhat differentscenario emerges as can be seen from Table 1. Theadenosine positioned between two LNA nucleotides, i.e.A6, is found in an N-type conformation with a purity of70%. In comparison, none of the analyzed nucleotidespossess more than 20% N-type conformation in the ssDNA.C1 and G3, each flanking just one LNA nucleotide, havesugar conformations with a nearly equal population of N-and S-type conformations. A4, flanking one LNA nucleo-tide, possess only 22% N-type conformation. However, thismay be due to unrealistically low values calculated forJ2'3'and J2@3', since a value ofJ1'2' = 5.8 Hz indicates a largerportion of N-type conformation. C9, the only nucleotide inthe oligomer not flanked by an LNA nucleotide, exhibits asugar equilibrium similar to the ssDNA nucleotides.

The modified oligonucleotided(CCGCTLAGCG):d(CGCTAGCGG)

This duplex contains only one LNA nucleotide in an almostself-complementary strand. This resulted in the formation ofthe LNA:LNA duplex d(CCGCTLAGCG):d(CCGC-

Table 3. Coupling constants (Hz) for the sugar protons in the d(CTLGATLATLGC):d(GCATATCAG) duplexa

J1'2' J1'2@ J2'3' J2@3' %Se PNf PS

f

C1b — — — — — — —TL2c 17G3b — — — — — — —A4 5.3 5.7 4.8 4.6 55 (7) ÿ20 (20) 177 (19)TL5c — — — — — 17A6 ÿ2.1 5.4 8.0 7.7 9 (16) ÿ33 (12)TL7c 17G8d 2.6 5.0 9.4 11.6 — — —C9 5.3 6.5 5.6 2.2 65 (13) ÿ15 (74) 196 (31)G10 5.4 7.6 5.5 2.2 75 (12) 207 (28)C11 7.7 6.0 4.9 1.1 85 (13) 183 (22)A12 5.9 5.7 5.1 1.2 72 (16) 191 (34)T13 7.0 5.9 6.2 1.1 71 (17) 173 (41)A14 6.9 6.1 5.8 1.3 71 (15) 180 (34)T15 7.4 5.3 6.0 0.7 79 (16) 169 (37)C16 7.7 6.1 2.4 0.9 93 (9) 192 (12)A17 8.5 5.5 2.1 0.8 87 (13) 173 (24)G18 8.4 5.8 2.7 1.1 93 (8) 184 (12)

a The coupling constants were derived from the H1'–H2' and H1'–H2@ cross peaks in the selective DQF-COSY spectrum using CHEOPS. TheJ2@2@ coupling constants were varied in the calculation, but values close toÿ15 Hz were obtained. The sugar conformations were calculatedusing the RANDPSEUROT program with the puckering amplitude kept constant at 36°, 38° and 40°, respectively. Only minor differences wereobserved. The values listed are the ones obtained with an amplitude of 38°. The standard deviations are given in parenthesis.b C1 and G3 could not be analyzed properly due to spectral overlap.c The LNA nucleotides were not included in CHEOPS calculations and the value ofPN were taken from Obikaet al. (1997).d G8 did not converge properly in the RANDPSEUROT iterations, but the appearance of the cross peak pattern in the selective DQF-COSYspectrum clearly demonstrated that this nucleus has the major contribution from the N-type conformation.e Percentage S-type sugar conformation calculated using RANDPSEUROT. The population of the N-type conformations is given by 100ÿ%S.f Pseudorotation angle calculated for the two sugar conformations using RANDPSEUROT. If the mole fraction of a conformer was calculated asbelow 30% the correspondingP value was not considered to be determined with sufficient accuracy.



TLAGCG) and DNA:DNA duplex d(CGCTAGCGG):d(CGCTAGCGG) as minor forms, both duplexes contain-ing only eight base pairs and with the end nucleotide in a‘dangling’ position. The amounts of the minor forms wereapproximately 10% of each. We prepared pure samples ofeach of these two minor form duplexes and recorded theone-dimensional(1D) NOESY and TOCSY spectra andperformed a complete assignment of these spectra. Thisenabled us to unequivocally assign the spectrum of theLNA:DNA duplex.

The NOESY spectra of the LNA:DNA duplex exhibit thecharacteristic features of dsDNA sequential connectivities.The aromatic protons (H6, H5, H8 and H2), the sugarprotons (H1', H2', H2@, H3', H4', and H5', H5@) as well asthe labile protons were assigned. The chemical shift valuesand the NOE connectivity pattern of the imino protons wereobserved to be in accordance with normal Watson–Crickbase pairing. The NOESY spectrum with short mixing time(50 ms) allows unambiguous assignments of the H2' andH2@ resonances. Figure 1 shows the aromatic H6/H8 to H2'/H2@ region in the NOESY spectrum. This spectrum exhibitsa few characteristic features originating from the absence oftheTL5 H2@ proton and the shift of theTL5 H2' proton to theH3' region.

Only weak cross peaks from the H6/H8 protons to H1'-protons are present in the NOESY spectra. This shows that

the nucleobases are in ananti-conformation. TheTL5nucleotide yields a strong intra-molecular cross peakbetween the H3' proton and the H6 base proton as a resultof the C3'-endoconformation of this locked sugar ring. Theremaining nucleotides, except for the nucleotide A6, do notexhibit a similar cross peak between the H3' protons and thebase protons. This indicates that these nucleotides arepredominantly in the S-type conformation. The nucleotideA6 gives rise to a fairly strong cross peak between A6 H3'and A6 H8, which indicates that this nucleotide may havesome degree of N-type conformation.

The chemical shift values obtained for the LNA:DNAduplex are very close to the ones obtained for theunmodified duplex, except in the case of the protons nearthe modification site. As expected, the chemical shift valuesof the protons on the LNA nucleotide differ from those inthe unmodified nucleotide. Some of the protons on the C4,A6, T13 and A14 nucleotides show differences in chemicalshift values between the modified and the unmodifiedduplex primarily in the order of 0.2–0.4 ppm. The changesof the sugar protons of these nucleotides may, at least partly,be related to conformational changes of the deoxyribosering. However, pronounced shifts of the resonances,especially of the adenine H2 protons, are most probablyrelated to an enhanced stacking of the nucleobases (Nielsenet al., 1999a,b). It is noteworthy that this effect is observed

Table 4. Coupling constants (Hz) for the sugar protons in the d(CTLGCTLTLCTLGC):d(GCAGAAGCAG) duplex andthe derived sugar conformationsa

J1'2' J1'2@ J2'3' J2@3' %Sd PNe PS

e

C1b — — — — — — —TL2c 17G3 4.4 6.0 4.6 4.6 51 (9) ÿ21 (18) 187 (20)C4b — — — — — — —TL5c 17TL6c 17C7 1.6 6.0 4.5 10.6 6 (8) ÿ7 (21)TL8c 17G9 2.6 6.4 3.2 5.1 45 (9) ÿ12 (11) 203 (22)C10 5.0 7.0 6.4 5.4 43 (10) 8 (37) 164 (48)G11 7.5 5.6 4.5 1.7 81 (11) 174 (22)C12 7.7 6.2 5.6 0.9 81 (13) 164 (31)A13 7.9 5.9 5.8 ÿ0.3 86 (14) 153 (32)G14 8.2 5.6 5.8 1.9 78 (11) 154 (26)A15 7.8 5.7 5.5 0.5 82 (14) 165 (28)A16 7.9 5.0 6.1 2.0 74 (9) 148 (20)G17 8.5 5.6 4.4 1.1 89 (10) 177 (18)C18 7.9 6.0 5.6 2.1 76 (12) 163 (26)A19 9.1 5.9 5.6 0.0 91 (12) 153 (28)G20 8.4 6.5 6.4 ÿ0.2 88 (12) 132 (26)

a The coupling constants were derived from the H1'–H2' and H1'–H2@ cross peaks in the selective DQF-COSY spectrum using CHEOPS. TheJ2@2@ coupling constants were varied in the calculation, but values close toÿ15 Hz were obtained. The sugar conformations were calculatedusing the RANDPSEUROT program with the puckering amplitude kept constant at 36°, 38° and 40°, respectively. Only minor differences wereobserved. The values listed are the ones obtained with an amplitude of 38°. The standard deviations are given in parenthesis.b C1 and C4 could not be analyzed properly due to spectral overlap.c The LNA nucleotides were not included in CHEOPS calculations and the value ofPN were taken from Obikaet al. (1997).d Percentage S-type sugar conformation calculated using RANDPSEUROT. The population of the N-type conformations is given by 100ÿ%S.e Pseudorotation angle calculated for the two sugar conformations using RANDPSEUROT. If the mole fraction of a conformer was calculated asbelow 30% the correspondingP value was not considered to be determined with sufficient accuracy.



Copyright © 2000 John Wiley & Sons, Ltd. J. Mol. Recognit. 2000; 13

THE CONFORMATIONS OF LNA

Plate 1. The H1’ to H2’/H2” selective DQF COSY spectrum of the single stranded d(CTGATATGC)oligonucleotide (left) and the modified d(CTLGATLATLGC) oligonucleotide (right) obtained at 500 MHz.

Copyright © 2000 John Wiley & Sons, Ltd. J. Mol. Recognit. 2000; 13

M. PETERSEN ET AL.

Plate 2. The H1’ to H2”/H2” selective DQF COSY spectrum of the d(CTLGCTLTLCTLGC):d(GCAGAAGCAG) duplex obtained at 500 MHz. The experimental spectrum is shown to the leftwhile the spectrum calculated by CHEOPS is shown to the right.

both on the adenine base paired to the LNA nucleotide aswell as on the adenine neighboring the LNA nucleotide onthe same strand.

Cross peaks from 14 deoxyribose residues observed in theselective DQF-COSY spectrum of the H1' and H2'/H2@region could be used to determine the sugar conformations.The appearance of the observed cross peak patternsresembles the one typical for a C2'-endosugar conformationexcept for the nucleotide A6, which differs substantiallyfrom this appearance. The CHEOPS simulations of thisspectrum returned good correlation values with the experi-mental spectrum. TheJ-coupling constants returned byCHEOPS were used as input to RANDPSEUROT. Theminimization with this program converged for all theresidues analyzed by CHEOPS, returning the values givenin Table 2.

The modified oligonucleotided(CTLGATLATLGC):d(GCATATCAG)

The 1D1H NMR spectrum of the sample of this modifiedduplex exhibits sharp lines from the expected LNA:DNAduplex. The assignment of the 2D NMR spectra wasperformed similarly to the case of the d(CCGCTL

AGCG):d(CGCTAGCGG) duplex. The chemical shiftvalues obtained are very close to the ones obtained for theunmodified duplex, except in case of the protons near themodification site. As expected, the protons on the LNAnucleotides differ from those in the unmodified nucleotide.Some of the protons at the G3, A4, A6, A12, A14 and A17nucleotides show differences in chemical shift valuesbetween the modified and the unmodified duplex primarilyin the order of 0.2–0.4 ppm. As previously stated for thed(CCGCTLAGCG):d(CGCTAGCGG) duplex, this can beexplained as due to a change of the sugar conformation andan enhanced stacking near the modification sites. However,since three modifications are present in this duplex thetrends in the changes are somewhat more obscure.

The H3' to H6/H8 region of the NOESY spectrum showsthat the locked modified nucleotides are in the C3'-endoconformation with strong intra-molecular cross peaksbetween H3' and H6. Fairly strong intra-molecular crosspeaks between H3' and H6/H8 are also observed for A6, C8and G9. Consequently, these three nucleotides may have ahigh degree of N-type conformation. The remaining nucleo-tides seem to be predominantly in S-type conformation.

Cross peaks from 13 deoxyribose residues observed in theselective DQF-COSY spectrum of the H1' and H2'/H2@region could be used to determine the sugar conformations.The nucleotides C1 and G3 yielded cross peaks in regions ofoverlap. The appearance of the cross peak pattern resemblesthe one typical for a C2'-endosugar conformation except forthe nucleotides A4, A6 and G8, which differ substantiallyfrom this appearance. The CHEOPS simulations of thespectrum returned good correlation values with the experi-mental spectrum. TheJ-coupling constants returned byCHEOPS were used as input to PSEUROT. The RANDP-SEUROT minimization converged for 12 of the 13 residuesanalyzed by CHEOPS, returning the values given in Table3. The RANDPSEUROT iterations of the nucleotide G8 didnot converge properly, but the appearance of the cross peak

pattern in the selective DQF-COSY spectrum clearlydemonstrated that this nucleotide has the major contributionfrom the N-type conformation.

The modified oligonucleotided(CTLGCTLTLCTLGC):d(GCAGAAGCAG)

The 1D1H NMR spectrum of the sample of this modifiedduplex shows only lines from the expected LNA:DNAduplex. No signs of alternative hybridization wereobserved. The spectra of the LNA:DNA as well as thecorresponding DNA:DNA duplex show the characteristicfeatures of dsDNA sequential connectivities and thenormal Watson–Crick NOE connectivity pattern. Thechemical shift values observed demonstrate close agree-ment between the modified and unmodified duplex exceptfor the protons near the modification site. As expected fromthe sugar conformation, the protons in the LNA nucleotidesdiffer from those in the unmodified nucleotide. The protonson the other nucleotides show differences in chemical shiftvalues between the modified and the unmodified duplexprimarily at the A13, A19, C7 and G14 nucleotides. Thelargest effect is observed for the adenine H2 protons,especially of A13 and A19 with downfield shifts of 0.51and 0.38 ppm, respectively. A large effect is also observedat some of the other aromatic protons, e.g. C7H6 andG14H8 with an upfield shift of 0.40 ppm and a downfieldshift of 0.36, respectively. The chemical shifts of the otheraromatic proton are changed more moderately.

The H3' to H6/H8 region of the NOESY spectrum alsocontains the cross peak between the H2' and the baseprotons of the modified nucleotide. This part of the NOESYspectrum shows that the locked modified nucleotides are inthe C3'-endo conformation with strong intra-molecularcross peaks between H3' and H6. Fairly strong intra-molecular cross peaks between H3' and H6/H8 are alsoobserved in case of G3, C7 and G9. These three nucleotidesmay therefore have some degree of N-type conformationwhile the remaining sugar rings seem to be in the S-typeconformation.

The experimental and calculated H1' and H2'/H2@ regionsof the selective DQF-COSY spectrum of thed(CTLGCTLTLCTLGC):d(GCAGAAGCAG) duplex areshown in Plate 2. The cross peaks from 15 nucleotidesobserved in the spectrum could be used to determine thesugar conformations. The CHEOPS simulations of thisspectrum returned good correlation values with the experi-mental spectrum. TheJ-coupling constants returned byCHEOPS were used as input to RANDPSEUROT. TheRANDPSEUROT minimization converged for all theresidues analyzed by CHEOPS, returning the values givenin Table 4.

DISCUSSION

Our results on the ssLNA and ssDNA indicate that an LNAnucleotide is able to alter the sugar conformation of itsneighbors since nucleotides flanking one LNA nucleotidehave sugar conformations with nearly equal populations ofN- and S-type conformation. The nucleotide positioned



between two LNA nucleotides is subject to an ‘additive’effect altering the sugar conformation to a predominantly N-type conformation. This change in the dynamic behavior ofthe ssLNA compared to the ssDNA may be a consequenceof a quenching of the concerted local backbone motions inthe modified single stranded oligonucleotide due to thelocked conformation of the LNA nucleotide.

Qualitatively, the NOESY spectra of the three LNA:DNAduplexes indicate that they all adopt a right-handed helixconformation. All bases are in theanti-conformation andthey all form normal Watson–Crick base pairs. Theunmodified nucleotides are found predominantly in S-typeconformations except for those nucleotides flanking an LNAnucleotide. These nucleotides adopt a higher degree of N-type conformation than in normal DNA. The largest effect isobserved on the nucleotide following the modification in the3'-direction.

The d(CCGCTLAGCG):d(CGCTAGCGG) duplex con-taining only one LNA modification exhibits the smallesteffect on the sugar conformations. Except for the LNAnucleotide, only the nucleotide following the modificationin the 3'-direction (A6) has an appreciable amount of N-typeconformation, but even for this nucleotide, the S-typeconformation remains the most populated one.

The situation is dramatically changed in the d(CTLGA-TLATLGC):d(GCATATCAG) duplex with three LNAmodifications. The modified strand in this duplex isidentical to the ssLNA strand. The flanking nucleotidesG3, A6 and G8 have the N-type conformations as the mostpopulated ones. The nucleotide A6, which is flanked by amodification in both directions, is exclusively in the N-typeconformation. A similar situation is observed in thed(CTLGCTLTLCTLGC):d(GCAGAAGCAG) duplex inwhich the nucleotide C7 is exclusively in the N-typeconformation whereas G3 and G9 are in mixed conforma-tions with a substantial N-type contribution.

The nucleotides base-paired to the LNA nucleotideremain predominantly in the S-type conformation. Thus,the conformational effect of introducing an LNA nucleotidein a DNA duplex seems to be very local. Yet, the effect onthe thermal stability is substantial. Similar strong effectshave been reported in other modified DNA duplexes(Gelfandet al., 1998).

The chemical shift values of the LNA:DNA duplexes arevery close to the values observed in the correspondingDNA:DNA duplexes except for the H2 and H6/H8 of thebases close to the modifications. The H2 protons on theadenine base paired to the LNA nucleotide show the largestchemical shift changes. This indicates that an increased ringcurrent effect influences these protons and consequently thata more efficient stacking of the involved nucleobases occurs(Nielsenet al., 1999a,b).

The results obtained in this work reveal two importantchanges when introducing a modified LNA nucleotide in aDNA duplex. The locked conformation of the LNAnucleotide organize the phosphate backbone in such a wayas to introduce preference for the sugar rings to adopt the N-type conformation for the nucleotides following in the 3'-direction. Furthermore, an improved stacking of thenucleobases is associated with this ordering of thephosphate backbone.

Searle and Williams (1993) have discussed the stability of

nucleic acid structures in solution in terms of enthalpy–entropy compensation. Following their outline, the effect ofintroducing a modification can be explained in terms of achange of the free energy of the ‘internal rotors’ of thephosphate backbone (DGr) and the free energy related to thebase stacking (DGs). The change of the free energy of thebase stacking is dominated by an enthalpy change (DGs �DHs), whereas the change of the free energy of the internalrotors of the backbone is dominated by entropy (DGr� DSr)(Searle and Williams 1993).

The conformations of the nucleotides in the ssLNAand the corresponding LNA:DNA duplex d(CTLGA-TLATLGC):d(GCATATCAG) are found to show thesame trend. Consequently, the ssLNA strand is to somedegree preorganized to duplex formation. This impliesthat a smaller change of backbone entropy (DSr) has tobe paid in the duplex formation. The organization of thebackbone and the sugar rings in a C3'-endo fashionfavors a more efficient stacking. This is equivalent to thestructural changes from B-DNA (C2'-endo conformation)to A-DNA (C3'-endo conformation) (Saenger, 1984).Thus, in LNA modified duplexes there is a local changeof the phosphate backbone geometry that favors a higherdegree of stacking and hence an increased value ofÿDHs.

A straightforward consequence of the explanation givenabove is that the increase in the stability of LNAmodified LNA:DNA duplexes should saturate as thenumber of modifications increases. As observed pre-viously, in the d(GTLGATLATLGC):d(GCATATCAC)duplex the melting temperature has increased 5.3°C permodification compared to the unmodified DNA duplex,whereas in the fully modified d(GLTLGLAL-

TLALTLGLCL):d(GCATATCAC) duplex the increase inthe melting temperature was 4.5°C (Singh et al., 1998a;Koshkin et al., 1998b). Thus, the largest effect permodification is indeed observed in duplexes that are notfully modified.

It seems difficult to specify the amount of stabilizationthat arises from the ordering of the phosphate backboneand the part caused by increased stacking. Breslaueret al.(1986) have determined the nearest neighbor thermo-dynamic parameters related to the base pairing in DNAduplex formation and used these values to predict duplexstability. They found very convincing agreement betweencalculated and experimentally determined values. Thevalues ofDGo, DHo andDSo for duplex formation of twonearest neighboring base pairs are very sequence-specificwith increasing value ofÿDGo between 4 and 15 kJ/mol(0.9 –3.6 kcal/mol),ÿDHo between 23 and 50 kJ/mol (5.6–11.9 kcal/mol) andÿDSo between 54 and 116 J/mol(12.9 –27.8 cal/mol). Within these limits the stability ofLNA:DNA duplexes is easily explained by a slightlyincreased value ofÿDHo and a slightly decreased valueof ÿDSo.

CONCLUSIONS

The stability of partly modified LNA:DNA duplexes can beexplained as follows. The locked C3'-endoconformation ofthe LNA nucleotide locally organizes the phosphate back-



bone (decreased loss of entropy upon duplex formation) inthe direction of a conformation that in the duplex favors amore efficient stacking of the nucleobases (increased loss inenthalpy upon formation). Thus, the formation of anLNA:DNA duplex is favored by both enthalpy and entropycompared to the corresponding DNA:DNA duplex.

Acknowledgements

The Danish Natural Science Research Council and The Technical ResearchCouncil are thanked for generous financial support. We thank theInstrument Center for NMR-Spectroscopy of Biological Macromoleculesat The Carlsberg Laboratory, Copenhagen for providing spectrometer timeat the 750 MHz spectrometer.

REFERENCES

Altona C, Sundaralingam M. 1972. Conformational analysis of thesugar ring in nucleosides and nucleotides. A new descriptionusing the concept of pseudorotation. J. Am. Chem. Soc. 94:8205±8212.

Altona C, Sundaralingam M. 1973. Conformational analysis of thesugar ring in nucleosides and nucleotides. Improvedmethods for the interpretation of proton magnetic resonancecoupling constants. J. Am. Chem. Soc. 95: 2333±2344.

Boelens R, Scheek RM, Dijkstra K, Kaptein R. 1985. Sequentialassignment of imino- and amino-proton resonances in 1HNMR spectra of oligonucleotides by two-dimensional NMRspectroscopy. Application to lac operator fragment. J. Magn.Reson. 62: 378±386.

Brameld KA, Goddard III WA. 1999. Ab initio quantum mechanicalstudy of the structure and energies for the pseudorotation of5'-dehydroxy analogues af 2'-deoxyribose and ribose sugars.J. Am. Chem. Soc. 121: 985±993.

Breslauer, FR, BluÈ cker H, Marky L. 1986. Predicting DNA duplexstability from base sequence. Proc. Natl. Acad. Sci. 83: 3746±3750.

Dickerson RE, Bansal M, Calladine CR, Diekmann S, Hunter WN,Kennard O, von Kitzing E, Nelson HCM, Lavery R, Olson WK,Saenger W, Shakked Z, Soumpasis DM, Tung C-S, Sklenar H,Wang AHJ, Zhurkin VB. 1989. De®nitions and nomenclatureof nucleic acid structure parameters. EMBO J 8: 1±4.

Feigon J, Leupin W, Denny WA, Kearns DR. 1983. Twodimensional proton nuclear magnetic resonance investiga-tion of the synthetic deoxyribonucleic acid decamerd(ATATCGATAT)2. Biochemistry 22: 5943±5951.

Flanagan WF. 1998. Antisense comes of age. Cancer MetastasisRev 17: 169±176.

Gelfand CA, Plum GE, Grollman AP, Johnson F, Breslauer KJ.1998. The impact of an exocyclic cytosine adduct on DNAduplex properties: signi®cant thermodynamic consequencesdespite modest lesion-induced structural alterations. Bio-chemistry 37: 12507±12512.

Hare DR, Wemmer DE, Chou S-H, Drobny G, Reid BR. 1983.Assignment of the non-exchangeable proton resonances ofd(C-G-C-G-A-A-T-T-C-G-C-G) using two-dimensional nuclearmagnetic resonance methods. J. Mol. Biol. 171: 319±336.

HeÂ leÁ ne C, ToulmeÂ J-J. 1990. Speci®c regulation of geneexpression by antisense, sense and antigene nucleic acids.Biochem. Biophys. Acta. 1049: 99±125.

Koshkin AA, Nielsen P, Meldgaard M, Rajwanshi VK, Singh SK,Wengel J. 1998a. LNA (locked nucleic acid): an RNA mimicforming exceedingly stable LNA:LNA duplexes. J. Am. Chem.Soc. 120: 13252±13253.

Koshkin AA, Singh SK, Nielsen P, Rajwanshi VK, Kumar R,Meldgaard M, Olsen CE, Wengel J. 1998b. LNA (lockednucleic acids): synthesis of the adenine, cytosine, guanine, 5-methylcytosine, thymine and uracil bicyclonucleoside mono-mers, oligomerisation, and unprecedented nuclei acid re-cognition. Tetrahedron 54: 3607±3630.

Koshkin AA, Rajwanshi VK, Wengel J. 1998c. Novel convenientsyntheses of LNA [2.2.1]bicyclo nucleosides. TetrahedronLett. 39: 4381±4384.

Macaya RF, Schultze P, Feigon J. 1992. Sugar conformation inintramolecular DNA triplexes determined by coupling con-stants obtained by automated simulation of P.COSY crosspeaks. J. Am. Chem. Soc. 114: 781±783.

Nielsen CB, Singh SK, Wengel J, Jacobsen JP. 1999a. Thesolution structure of a locked nucleic acid (LNA) hybridized toDNA. J. Biomol. Struc. Dyn. 17: 175±191.

Nielsen KE, Wengel J, Jacobsen JP. 1999b. Solution structure ofan LNA hybridized to DNA: NMR study of thed(CTLGCTLTLCTLGC):d(GCAGAAGCAG) duplex containingfour locked nucleotides. Bioconjugate Chem. (in press).

Obika S, Nanbu D, Hari Y, Mori K, In Y, Ishida T, Imanishi T. 1997.Synthesis of 2'-O, 4'-C-methyleneuridine and -cytidine. Novelbicyclic nucleosides having a ®xed C3'-endo sugar puckering.Tetrahedron Lett. 38: 8735.

Olson WK, Sussman JL. 1982. How ¯exible is the furanose ring?1. A comparison of experimental and theoretical studies. J.Am. Chem. Soc. 104: 270±278.

Saenger W. 1984. Principles of Nucleic Acid Structure; Springer:New York.

Scheek RM, Russo N, Boelens R, Kaptein R. 1983. Sequentialresonance assignments in DNA 1H NMR spectra by two-dimensional NOE spectroscopy. J. Am. Chem. Soc. 105:2914±2916.

Searle MS, Williams DH. 1993. On the stability of nucleic acidstructures in solution: enthalpy±entropy compensations,internal rotations and reversability. Nucleic Acid Res. 21:2051±2056.

Singh SK, Wengel J. 1998. Universality of LNA-mediated high-af®nity acid recognition. Chem. Commun. 1247±1248.

Singh SK, Nielsen P, Koshkin AA, Wengel J. 1998a. LNA (Lockednucleic acids): synthesis and high-af®nity acid recognition.Chem. Commun. 455±456.

Singh SK, Kumar R, Wengel J. 1998b. Synthesis of 2'-amino-LNA:a novel conformationally restricted high-af®nity oligonucleo-tide analogue with a handle. J. Org. Chem. 63: 10035±10039.

Stein PC, Jacobsen JP, Spielmann HP. 1995. Design and imple-mentation of a simple shaped notch ®lter. J. Magn. Res. Ser.B 109: 93±96.

Uhlmann E, Peyman A. 1990. Antisense oligonucleotides: a newtherapeutic principle. Chem. Rev. 90: 543±584.

Wengel J. 1998. Synthesis of 3'-C- and 4'-C-branched oligonu-cleotides and the development of locked nucleic acid (LNA).Acc. Chem. Res. 32: 301±310.

Wengel J, Koshkin A, Singh SK, Nielsen P, Meldgaard M,Rajwanshi VK, Kumar R, Skouv J, Nielsen CB, Jacobsen JP,Jacobsen N, Olsen CE. 1999. LNA (locked nucleic acids).Nucleosides and Nucleotides 18: 1365±1370.

WuÈ thrich K. 1986. NMR of Proteins and Nucleic Acids. Wiley, NewYork.



The conformations of locked nucleic acids (LNA)

Documents

Transcript of The conformations of locked nucleic acids (LNA)