64246438 NucleicAcidsResearch, 2007, Vol. 35, No. 19 Publishedonline18September2007doi:10.1093/nar/gkm664Crystal structure, stabilityandinvitroRNAi activityof oligoribonucleotidescontainingtheribo-difluorotoluyl nucleotide: insightsintosubstraterequirementsbythehumanRISCAgo2enzymeFengLi1, PradeepS. Pallan1, MartinA. Maier2, Kallanthottathil G. Rajeev2,StevenL. Mathieu3, ChristophKreutz4, YupengFan2, JayoditaSanghvi2, RonaldMicura4, EriksRozners3, MuthiahManoharan2andMartinEgli1,*1Departmentof Biochemistry,Schoolof Medicine, VanderbiltUniversity, Nashville,TN 37232,2DepartmentofDrugDiscovery,AlnylamPharmaceuticals,Inc.,300ThirdStreet,Cambridge,MA02142,3DepartmentofChemistryandChemical Biology, NortheasternUniversity, Boston, MA02115, USAand4Instituteof OrganicChemistry, Centerfor Molecular Biosciences(CMBI), Leopold-FranzensUniversity, 6020Innsbruck, AustriaReceivedJuly20, 2007; RevisedandAcceptedAugust 10, 2007ABSTRACTShort interfering RNA (siRNA) duplexes are currentlybeing evaluated as antisense agents for genesilencing. Chemical modification of siRNAs iswidely expected to be required for therapeuticapplications in order to improve delivery, biostabilityandpharmacokineticproperties. Beyondpotentialimprovements intheefficacy of oligoribonucleotides,chemical modification may also provide insightinto the mechanism of mRNA downregulationmediatedby the RNAproteineffector complexes(RNA-inducedsilencingcomplexorRISC).WehavestudiedtheinvitroactivityinHeLacellsof siRNAduplexes against firefly luciferase with substitutionsin the guide strand of U for the apolar ribo-2,4-difluorotoluylnucleotide(rF)[Xia,J.etal.(2006)ACSChem.Biol.,1,176183]aswellasofCforrF.Whereasaninternal rF:ApairadjacenttotheAgo2(slicer enzyme) cleavage site did not affect silencingrelative to the native siRNA duplex, the rF:G pair andother mismatches such as A:Gor A:Awere nottolerated.The crystal structureat atomicresolutiondeterminedfor anRNAdodecamer duplexwithrFopposite G manifests only minor deviations betweenthegeometriesof rF:GandthenativeU:Gwobblepair. This is in contrast to the previously found,significant deviations between thegeometries of rF:Aand U:Apairs. Comparison between the structures ofthe RNA duplex containing rF:G and a new structureof an RNAwithA:Gmismatches withthe structures ofstandard WatsonCrick pairs in canonical duplexRNAleadstotheconclusionthatlocal wideningofthe duplex formed by the siRNA guide strand and thetargeted region of mRNA is the most likely reason fortheintoleranceof humanAgo2(hAgo2), theRISCendonuclease, toward internal mismatch pairs invol-ving native or chemically modified RNA. Contrary tothe influence of shape, the thermodynamic stabilitiesofsiRNAduplexeswithsinglerF:A,A:A,G:AorC:A(instead of U:A) or rF:Gpairs (instead of C:G) shownoobvious correlationwiththeir activities. However,incorporation of three rF:Apairs intoan siRNAduplexleadstolossofactivity.Ourstructuralandstabilitydata alsoshedlight ontherole of organic fluorine as ahydrogenbondacceptor. Accordingly, UVmelting(TM) data, osmotic stress measurements, X-raycrystallographyat atomic resolutionand the resultsof semi-empirical calculations are all consistent withthe existence of weak hydrogen bonds betweenfluorineandtheH-N1(G) aminogroupinrF:Gpairsof the investigated RNA dodecamers.INTRODUCTIONMessenger RNAs (mRNAs) represent a necessary stepingeneexpression, beingtheintermediarybetweengeneand protein. Short interfering RNAs (siRNAs) areduplexes of RNA, 2123 nt long, and consist of thesense and the guide strand. The latter becomesincorporatedintoa complex of proteins andserves to*Towhomcorrespondenceshouldbeaddressed. Tel: +16153438070; Fax: +16153227122; Email: martin.egli@vanderbilt.edu2007TheAuthor(s)This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/)whichpermitsunrestrictednon-commercialuse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited. at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from identify acomplementary sequence inanmRNA. Thiseither causes the mRNAto be cleaved by the proteincomplexorpreventsitfrombeingtranslatedintoprotein(18), thus eectively silencing the encoding gene.Therefore, RNA interference (RNAi) is a powerfulbiologicalprocessforspecicsilencingofgeneexpressionin diversied eukaryotic cells and has tremendouspotential for functional genomics, drug discovery throughinvivotargetvalidation, anddevelopmentofnovelgene-specicmedicine.Itishopedthatthisapproachmightbeusedtoshutdowndisease-relatedgenesinhumans.TheRNAi processwithincellscomprisesatleastfoursequential steps(915): (i) ATP-dependent processingofdouble-strandedRNAintosiRNAsbyDicer,anRNaseIII-family enzyme; (ii) incorporation of these naturallyderivedsiRNAs,orsyntheticsiRNAsintroducedintothecell, into an inactive ribonucleoprotein complex; (iii) ATP-dependent unwinding of the siRNA duplex to generate anactive RNA-inducedsilencingcomplex(RISC) wherebyATPis usedtomaintain50-phosphates onsiRNAs and(iv) cleavageof theRNAtarget inanATP-independentreaction. RISCisasequence-specicendonucleasecom-plex that contains Argonaute (Ago) proteins and thesingle-stranded guide siRNA strand. RISC mediatescleavage of target RNAs with multiple turnovers inthe absence of additional ATP. The highspecicity ofthe process is based on WatsonCrick pairing between thesiRNAantisense (guide) oligonucleotide andthe targetmRNA. WithintheRISCcomplex, theAgo2enzymeisexclusively responsible for cleavage of the mRNA (16,17).Structural dataindicatethat thePIWI domainof Ago2that adoptsanRNaseH-likefoldisresponsiblefor theSliceractivity(18).The cleavage position on the mRNA in a 21-mer siRNAduplex with mRNAis located between nucleotides 10and11upstreamfromthetargetnucleotidepairedtothe50-terminus of the antisense strand (12). The sites of activeduplexes are not predictable, but numerous siRNAstargetedagainst aparticular messagehavebeenstudiedin attempts to correlate duplex features with function.Anevaluationoftheactivityof25duplexestargetingtheopenreadingframeand30-untranslatedregionof laminA/Cfoundnobiasbasedonpositionwithinthemessage(19). Variations in siRNA eciency as a function of smallpositional shifts on the target (19,20) couldbe due tocompetition for binding with protein factors, dierentthermodynamicstabilities,orsequencedependenceoftheRISC complex. In addition, the structure and accessibilityof the target message will also play an important role (21)andpredictionalgorithms mayoer some guidance forsiRNA sequence selection (22). The most eective siRNAshaveatendencytopairmoreweaklyat the50-antisensesideoftheduplexrelativetothe30-antisenseside(22,23).Several criteria including base composition, strengthofbasepairingatparticularpositionsintheduplex, overallthermodynamicstabilityandbaseidentitieswereusedtoselect siRNAduplexes against vegenes (24). Duplexesthat were selectedbasedonthese criteria resultedinameansilencingof76%comparedto39%meansilencingfor those that were randomly selected. Introductionofmismatches inthecentral regionof theantisensestrandwas foundtoabolishsilencingactivity (12,25), but thesense strand can tolerate limited mismatches (25,26).Interestingly, whenmismatcheswereplacedinthetermi-nal four-baseregionbetweenthe50-endof theantisensestrand and the 30-end of the sense strand improvements inactivitywereobserved(27).ChemicallymodiedsiRNAslinked by a hairpin loop can still trigger target RNAcleavage through the RNAi pathway (19,2830).However, it appears that recruitment byRISCrequiressiRNA duplexes because most antisense-only, single-strandedsiRNAsshowmuchlessactivitythanthecorre-sponding duplexes (10). Approaches to the successfulidentication of eective and specic siRNAs haverecentlybeenreviewed(25).Chemical modicationof siRNAduplexes onone orboth strands will likely be required to improve theirecacyforinvivoapplicationsdespitethefactthatshortdouble-strandedRNAsshowrelativelyhighstabilityandactivityincell culture(3138). Chemical modicationisdesirable toenhance: (i) the siRNA-bindinganityby,for example, conformationally preorganizing the guidestrand for the mRNA; (ii) the nuclease stability andtherefore the half-life of the siRNA duplexes in circulationinvivoand(iii) thebiodistributionandpharmacokineticpropertiesofsiRNAsandtoallowtargetingofsiRNAtocertain cell types. Incorporation of phosphorothioate (PS)linkages, 20-OMe- or 20-uoro-nucleotides into siRNAswas found to aect their activities dependent onthe position and extent of the modications (3538).Forexample,siRNAduplexeswiththesenseorantisensestrand fully 20-OMe-modied showed a signicantlydiminished activity in HeLa cells despite ecient silencingby the corresponding native RNAs (39). And siRNAs withboth strands fully 20-OMe-modied were inactive.Onthe other handsiRNAs that featuredfullyPS- and20-OMe-modiedsensestrands displayedsimilar ecacyandpotencyastheunmodiedparent duplexesinsomecases (40). The limited activity seen insome cases forduplexeswithcompletelymodiedantisensestrandsmaytherefore relate to altered unwinding of the modiedduplexor loadingofthe unmodiedantisense-strandintoRISCratherthaninaccessibilityofthetarget.Werecentlyevaluatedtheinvitroactivityof siRNAscontainingapolarribo-2,4-diuorotoluyl nucleotides(rF)in place of U (Figure 1Aand B)(41). Modicationof theguide strand at the 50-end withrF was found not to aectsilencing compared with the unmodied control.Moreover, siRNAs with an internal rF modication,despite a slightly reduced anity for their target eectivelysilenced gene expression and they also exhibited improvednucleaseresistanceinserum. ARACEanalysis demon-strated that the siRNA-mediated cleavage occurred atthe predicted site both for the unmodied and therF-containingduplex. The crystal structure of anRNAduplex with rF:Apairs revealed local conformationalvariations relative to canonical RNA (41). rF andApaired more loosely compared with Uand Aandtherewasanabsenceoforderedwatermoleculesaroundthe diuorotoluyl moiety in both grooves. Whereassubstitution of rFfor Uessentially had no eect onNucleic Acids Research, 2007, Vol. 35, No. 19 6425 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from silencing,basemismatchesatthecorrespondingpositionsled to a loss of (G:A, A:A) or reduced (C:A) activity (41).Here, wereport invitroactivitydataof siRNAswithU:G or rF:G pairs (Figure 1C and D) and crystalstructures near 1 A-resolution of RNA dodecamerduplexes with either rF:G(FG12-mer) or A:Gpairs(GA 12-mer) (Figure 1E). To further investigatethe relative stabilityof T:A, rF:A, T:GandrF:Gpairswe have carried out semi-empirical calculations at theMP2/6-31Gand MP2/6-31+Glevels. The combineddata allow insights into substrate specicity by hAgo2 andindicatethatRNAduplexshapeatornearthecleavagesiteis amoreimportant determinant of RISC-mediatedactivityofsiRNAsthanthermodynamicstability.MATERIALSANDMETHODSOligoribonucleotide synthesis and purificationThe rFphosphoramidite building block for incorpora-tionintoRNAs was preparedas describedinref. (41).All siRNAmoleculesandtherF-containingdodecamersforstabilityandcrystallographicstudiesweresynthesizedona394ABIsynthesizerusingpreviouslydescribedpro-tocols (41). The oligonucleotides were cleavedfromtheCPG, with simultaneous deprotection of base andphosphategroups, with1.0 ml of amixtureof ethanolicammonia(ammoniaethanol, 3:1) for 16 hat 558C. Thesolutionwas decanted, lyophilizedandre-suspendedin0.2 ml of triethylamine trihydrouoride (TEA.3HF,Aldrich) and was incubated at 658C for 90 min toremove the t-butyldimethylsilyl (TBDMS) groups. Thecompletely deprotected oligonucleotides were then pre-cipitated from anhydrous methanol (MeOH, 0.4 ml).The oligonucleotides were puried by PAGE(41) andthedesiredbands wereexcisedandshakenovernight in5 ml of 100 mMsodiumacetate. Theextractedoligonu-cleotides were desalted using C-18 Sep-Pak cartridges(Waters). The oligonucleotides were characterized byelectrospray mass spectroscopy (ES/MS) andanalyticalanion-exchange HPLCand/or capillarygel electrophor-esis(CGE).The GA dodecamer r(CGCGAAUUAGCG) wassynthesizedonaPharmaciaGeneAssemblerPlusinstru-ment using20-O-TOMnucleosidephosphoramidites(42)and previously described protocols (43). Deprotection andcleavage from the solid support was achieved in a mixtureof MeNH2inEtOH(8 M, 0.6 ml) andMeNH2inH2O(40%, 0.6 ml) for5 h. Afterthesolutionwascompletelyevaporated, tetrabutylammonium uoride trihydrate(TBAF.3H2O) in THF(1 M, 0.95 ml) was added. Thereaction mixture was slowly shaken for 14 h at roomtemperaturetoremovethe20-O-silyl ethers. Thereactionwas quenched by the addition of triethylammoniumacetate (TEAA) (1 M, pH7.4, 0.95 ml). The volume ofthe solutionwas reducedto1 ml andthe solutionwasloadedonaAmershamHiPrep26/10Desaltingcolumn(2.6 10 cm;SephadexG25).ThecrudeRNAwaselutedwithH2Oanddried. Analysis of crude RNAproductsafter deprotection was performed by anion-exchangechromatography on a Dionex DNAPac100 column(4 250 mm) at 808C. Flowrate: 1 ml/min; eluant A:25 mMTrisHCl (pH8.0), 6 Murea; eluant B: 25 mMTrisHCl(pH8.0),0.5 M NaClO4, 6 M urea;gradient:060%BinAwithin45 min; UV-detectionat265 nm. ThecrudeRNA(trityl-o)waspuriedonasemi-preparativeDionex DNAPac100 column (9 250 mm). Flowrate:2 ml/min; gradient: 510% B in A within 20 min.Fractions containing RNA were loaded on a C18SepPak cartridge (Waters/Millipore), washed with0.10.2 M (Et3NH)HCO3, H2O, and eluted withH2O/CH3CN(6/4).RNAfractionswerelyophilized.In vitro analysis of luciferase expressionHeLa cells that stably express both rey and Renillaluciferase(HeLaDual-luccells)weregrownaspreviouslydescribed (41). In vitro activity of siRNAs was determinedusingahigh-throughput 96-well plate format assayforluciferase activity. HeLaDual-luc cells were transfectedwithreyluciferasetargetingsiRNAsatconcentrationsFigure1. Chemical structuresof basepairs. (A) T:A; (B) F:A; (C) U:G; (D) F:G; (E) G:A. Hydrogenbondsareindicatedbydashedlines.6426 Nucleic Acids Research, 2007, Vol. 35, No. 19 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from ranging from0.04 to 30 nMusing Lipofectamine 2000(Invitrogen) according to the manufacturers protocol.After24 h,cellswereanalyzedforboth reyand Renillaluciferase expression using a plate luminometer(VICTOR2, PerkinElmer) and the Dual-Glo LuciferaseAssaykit(Promega). Firey/Renillaluciferaseexpressionratios were used to determine the percentage of genesilencingrelativetountreatedcontrols.Thermal denaturation (TM) and osmotic stress studiesMelting of each oligonucleotide (4 mM) was done in10 mMsodiumcacodylate, 0.1 mMEDTAand300 mMNaClinpresenceof0,5,10,15and20%weight/volumeof each of the three organic co-solutes in Table 1.Oligonucleotide concentrations were calculatedfromthenearest-neighborapproximation(44)using" =1000 M1cm1( =260 nm) astheextinctioncoecientfor rFnucleoside (45). Absorbance versus temperatureproles were measured at 280 nm on a Varian Bio100 spectrometer equipped with a six-position Peltiertemperaturecontroller.Thetemperaturewasincreasedat0.58Cper minute. Five samples were measuredconcur-rentlyinthe double-beammode. The meltingtempera-tures and thermodynamic parameters were obtainedusingVarianCarysoftware(Version02.00). Theexperi-mental absorbanceversus temperaturecurves werecon-vertedintoafractionofstrandsremaininghybridized(a)versustemperaturecurvesbyttingthemeltingproletoatwo-statetransitionmodel, withlinearlyslopinglowerand upper base lines. The melting temperatures (TM) wereobtained directly from the temperature at a=0.5.The nal TMwas anapproximationof usually ve toeightmeasurements.Thethermodynamicparametersweredeterminedfromthewidthat thehalf-height of thedierentiatedmeltingcurve. The fractionof strands remaininghybridized()versus temperature curves were convertedintodieren-tiatedmeltingcurves[da/d(T1M )versusTM] usingVarianCarysoftware (Version02.00). The widthof the dier-entiated melting curve at the half-height is inverselyproportional tothevant Hotransitionenthalpy; forabimolecular transition H=10.14/(T11T12) where T1isthelowertemperatureandT2istheuppertemperature(both in K) at one-half of [da/d(T1M )] (46).The changes in the number of water moleculesassociatedwiththemeltingprocessnwweredeterminedas described by Spink and Chaires (47)nw=(H/R)[d(T1M )/d(lnaw)], where H is theenthalpydeterminedfromthewidthatthehalf-heightofdierentiatedmeltingcurveinpurebuer andRis theuniversal gas constant (1.986 calmol1K1). Theexperi-mentally determined values of water activity (lnaw) atgiven co-solute concentrations were provided byProfessors SpinkandChaires. The slope of the plot ofreciprocal temperature (in K) of melting versus thelogarithmofwateractivity(lnaw) atdierent concentra-tions(0, 5, 10, 15and20%)ofsmall co-solutesgavethevalueofd(T1M )/d(lnaw). Thenal nwwereobtainedbylinearttingusingKaleidaGraphsoftware(Version3.51)with a condence level usually better than 98%. Theexperimental uncertainties were obtained as previouslyreported(48).Crystallization of RNAdodecamers containing rF:Gand A:Gpairs and X-ray diffraction data collectionCrystals of the FGdodecamer r(CGCFAAUUGGCG)weregrownat 188Cbythesitting-dropvapor diusionmethod, using the Grid Screen Ammonium Sulfate(Hampton Research Inc., Aliso Viejo, CA, USA).Droplets containing 0.5 mM oligonucleotide, 0.8 Mammonium sulfate and 0.05 MMES, pH 6.0, wereequilibrated against a reservoir of 1.6 Mammoniumsulfate, 0.1 MMES, pH6.0. Crystals appeared after3 days. Crystals of the AG dodecamer r(CGCGAAUUAGCG) weregrownat 188Cbythehanging-dropvapor diusion method, using the Nucleic AcidMiniscreen(Hampton) (49). Bothconditions 14and16resulted in crystals. Crystals for diraction data collectionwere grown from droplets containing 0.5 mM oligonucleo-tide, 5% 2-methyl-2,4-pentanediol (MPD), 20 mM sodiumcacodylate, pH 7.0, 6 mM spermine-4HCl, 40 mM sodiumchloride and 10 mM magnesium chloride that wereequilibratedagainst areservoir of 35%MPD. Crystalsappeared after 2 days. In both cases, crystals weremountedinnylonloops andfrozeninliquidnitrogen.Diractiondatawerecollectedat120 Kontheinsertiondevice beamlinesof theSoutheast RegionalCollaborativeAccessTeam(SER-CAT,22-ID,FGdodecamer)andtheDuPontNorthwesternDowCollaborativeAccess Team(DND-CAT, 5-ID, GAdodecamer) at the AdvancedPhoton Source (APS), Argonne, IL. The wavelengthsforthedatacollectionsvariedbetweenca.0.9and1.0 A.In both cases high- and low-resolution data sets wereTable1. Thermodynamicstability(TM) andosmoticstressdataforself-complementaryRNA12-merswithU:G, rF:A, rF:GandA:GpairsSequence Name TaM(8C) nwEthyleneglycol nwGlycerol nwAcetamideCGCUAAUUGGCG UG 57.7 0.2 18.5 3.0 20.3 3.4 33.1 3.0CGCFAAUUGGCG FG 54.9 0.2 14.6 2.1 26.0 3.3 29.2 2.4CGCGAAUUFGCG GF 53.9 0.3 13.5 2.4 24.7 3.7 29.6 3.1CGCFAAUUAGCGbFA 53.2 0.3 13.2 3.1 12.7 3.3 23.6 2.9CGCGAAUUAGCG GA 58.2 0.2 21.2 2.6 23.8 3.1 Not measuredCGCAAAUUUGCGcAU 58.7 0.6 22.8d18d43.2da4 mMRNA, 300 mMNaCl, 10 mMNacacodylate, pH6.5.bThecrystal structureof thisduplexisdescribedinref. (41).cTakenfromref. (48).dExperimental uncertaintieswerenot determined.Nucleic Acids Research, 2007, Vol. 35, No. 19 6427 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from collectedseparately.DiractiondatawereprocessedwiththeprogramsHKL2000(50)(FG)andXDS(51)(GA).Structure determination and refinementBoth structures were determined by the molecularreplacementmethodusingtheprogramEPMR(52). FortheFGRNAstructure, acanonical A-RNAdodecamerconstructedwiththeprogramTURBO-FRODO(53)wasused as the search model. A previously published structureat lowresolution (54) served as the search model forstructuredeterminationoftheGARNAduplex.Inbothcases, thestructureswereinitiallyrenedwithCNS(55)andat alater stage withthe programSHELX-97(56).Inthesubsequentrenement cycles, solvent watermole-cules were added, and anisotropic temperature factorsrenement was carried out. The program TURBO-FRODOwas usedtodisplayelectrondensitymaps formanualrebuildingofthemodels.TocalculatetheR-free,5%randomlychosenreections were set aside inbothcases.Computational methodsAb initio calculations were carried out using theGAUSSIAN03 code (57) on a 64-bit Linux machine.Optimizations were initiated at the HF-3-21G level,followedbycalculations at theMP2/6-31Gandnallyat the MP2/6-31+Glevels. No constraints, i.e. tomaintain planarity of the base pairs were used.Interactionenergies were computedas the dierence inenergy between the optimized base pair on one hand, andthe sum of energies of optimized components on the other.Tocalculatethepointenergiesofbasepairsbasedonthecrystallographic conformations, hydrogen atoms wereaddedbytheprogram. All quantitieswerecorrectedforbasis set superposition error by the standard counterpoiseprocedure(58).CoordinatesFinal coordinates andstructurefactors for theFGandGA RNAs have been deposited in the Protein Data Bank(http://www.rcsb.org.) The PDB IDcodes are 2Q1O(FGduplex)and2Q1R(GAduplex).RESULTSGene silencing activity of siRNAs with U or C replaced by rFToexaminetheeectsongenesilencingofrFincorpora-tion into siRNAs, we employed an in vitro rey luciferaseassayinHeLacells. Tworegions of theluciferase genewere targetedseparatelywithRNAduplexeswhoseguidestrands contained single or multiple rF residues (Figure 2)(41).Thesilencingactivitiesofmodiedduplexesappliedinconcentrationsrangingbetween0.04and30 nMweremeasuredintheformof adecreasedluciferaselumines-cence and compared with the signals resulting fromtreatment withnativecontrol siRNAduplexes. Asingleuridinepositionedthreeormorenucleotidesupstreamordownstreamfromthe cleavage site inthe guide strandcould be replaced by rF without a signicant change in the50%inhibitory concentration (IC50) (41). Moreover,replacementoftheUadjacenttothecleavagesitebyrFonlyincreasedtheIC50from0.21to1.2 nM(Figure2A).However, substitutionof either C9or C11for Uor rF,resulting in the formation of U:G and rF:G pairs,respectively, signicantly diminished or even abolishedthe silencing activity altogether (IC50>30 nM;Figure 2A). Similarly, replacement of U10by Aor Gresultedinalossofactivity, whereasaU10!Csubstitu-tion (leading to a C:Amismatch) yielded an IC50of0.92 nM, comparabletotheactivityofansiRNAduplexwiththeU10!rFsubstitution(41). Therefore, itappearsthat asinglepurinepurinemismatchpair or U:G/rF:Gwobble pairs (assumingthe rF:Gpair adopts awobblegeometry) at or adjacent to the cleavage site are nottoleratedbythehAgo2endonuclease.TheC:Amismatchand rF:Aconstitute noteworthy exceptions. ARACEanalysis of the cleavage products had established thatthepositionof thecleavagesiteremainedunchangedasaresult of usingsiRNAduplexeswithsinglerF:ApairscomparedwithnativesiRNAs(41).Whereas a single rF residue in place of U was toleratedintheguidestrand, consecutivereplacementofthreeUsalsoledtoaloss of activity(Figure 2B). The thermo-dynamicstabilities(UVmeltingTMs) of siRNA21-merduplexes withsingle rF:A, A:G, A:Aor C:Apairs arecomparable(between65and678Crelativetothenativeconstruct =738C) (41). Therefore, the observed dierenceinsilencingactivitybetweensiRNAduplexescontainingrF:A or any of the above mismatch pairs is unlikelyto bea consequence of reduced stability. Conversely, threeconsecutiverF:Apairs mayleadtolocal meltingof theRNAduplex and considerably distort or destroy thesecondaryandtertiarystructureofthehAgo2substrate.Thermodynamic stability of and osmotic stress measurementsfor oligoribonucleotides with U!rF or C!rF substitutionsToassesstheeectofrFincorporationoppositeAorGonthe thermodynamic stabilityof anRNAdodecamerduplex we carried out UVmelting experiments. Initialruns at a wavelength of 260 nm showed nonlinear upper orlower base lines, a problem that was eliminated bymeasuringoptical densitiesasafunctionof temperatureat 280 nm. The resulting TMs are listed in Table 1. For thecomparisonherebetweenthe sixsequenceswehavelistedTMdierencespermodiedbasepair(notethattherearetwo substitutions per self-complementary dodecamerduplex). Remarkably, substitutingUfor rFoppositeGresultedinaTMreductionofjust1.48Cpermodication(UG versus FG duplex). In the case of the dodecamer withGandrFswitched(GF)thereductioninTMrelativetothe UG duplex amounts to 1.98C per modication.Switching a U:G(UG) to an rF:Apair (FA) lowerstheTMby2.38C.Therefore,therF:GismorestablethantherF:A pair(TM=0.98C;FG versusFAduplex).ThedierenceinTMbetweentheFAandA:Upairs(AU)is2.88C. For completeness, we also measured the stability oftheGAduplexwithtwoG:Amismatchpairs. TheGAduplex was slightly less stable than the AU duplex(TM=0.38C; GA versus AUduplex) but slightly6428 Nucleic Acids Research, 2007, Vol. 35, No. 19 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from more stable than the UGduplex (TM=0.38C; GAversus UG duplex). Overall the observed order of stabilityofthepairsatpH6.5is: A:U>G:A>U:G>rF:G>G:rF>rF:A.In order to gain insight into changes in RNA hydrationasaresultofrFincorporation,weemployedtheosmoticstress method (47,59). The reductions in the RNA meltingtemperatures uponadditionof ethylene glycol, glycerolor acetamide were measured at dierent concentrations ofthesesmallco-solutes(48,60).Thechangesinthenumberof water molecules nwassociated with the meltingprocess were determined as described by Spink andChaires (47): nw=(H/R)[d(T1M)/d(lnaw)], whereHis the measured enthalpy, Ris the universal gasconstant (1.986 calmol1K1) and lnaw are experimentallydetermined values of water activity at given co-soluteconcentrations.Thisapproachtoestablishthenumberofwater molecules bound to a nucleic acid duplex is based onthe assumption that the co-solutes (ethylene glycol,glycerol, acetamide) onlychange the water activityanddonotdirectlyinteractwiththeRNA.Rather thaninterpretingtheabsolutevalues obtainedfornw webelieveit maybemore instructiveto considerthe relative dierences in amounts of water released whencomparing the duplexes (Table 1). Interestingly, intheethylene glycol and acetamide series the general trendregarding hydration followed the above order ofthermodynamic stabilities (TM): A:U>G:A>U:G>rF:G>G:rF>rF:A.Thus,introductionoftheG:AandU:G mismatches gradually decreased the hydration.Introductionof rFas a base-pairing partner decreasednwevenmore, althoughthe eect maybe considereda relatively minor one, especially if one considers thehighlyhydrophobicnatureoftherFmodication. Inallthree series, hydration of the duplexes with either rF:G orG:rFpairswasmorefavorablethanthehydrationoftheduplex with rF:A pairs, which is in agreement withthe observations made incrystal structures (vide infra).Overall, the trends were less clear in the glycerol series. Aspreviouslyobserved(48,60), thenwvalues weresome-what higher in the acetamide than in the ethylene glycol orglycerol series. Apossible explanation for such discre-pancies might be the fact that the organic co-solutesdisplaysomeinteractionwiththeRNA(61,62). Recordand coworkers (62) showed that most small organicco-solutes (including glycerol) were not completelyexcluded fromthe surface of bovine serumalbumin.Whenusedinosmoticstressexperiments,suchco-solutesare expected to underestimate the magnitude in change ofhydration(62). Whereas the inertness of the co-solutesand reliability of the absolute numbers of nw is thereforestillunderdebate(61,62),we(48,60)andothers(47)haveused the relative trends to gain insight into the changes inhydrationinnativeandmodiednucleicacids.X-ray crystallography of RNA dodecamers with rF:Gand A:G mismatchesInorder togainabetter understandingof the invitrosiRNAactivity and thermodynamic stability/hydrationdataofRNAscontainingrFresiduesandtocomparethe0204060801001200 5 10 15 20 25 30 35% Luc. Activity1:9 1:10 1:11 1:2 1:12 1:19 AL-1 dTdTAGCUUCAUGAGUCGCAUUC 5 (sense strand)AL-25 UCGAAGUACU C AGCGUAAGdTdT(guidestrand) AL-95 UCGAAGUA FUCAGCGUAAGdTdTAL-105 UCGAAGUACF|CAGCGUAAGdTdTAL-115 UCGAAGUACU F AGCGUAAGdTdTAL-125 UCGAAGUACU U AGCGUAAGdTdT AL-195 UCGAAGUAUU C AGCGUAAGdTdT 12345 6789 10 11 1213 14 1516 17 1819 2021 A AL-18dTdT AACCACUCCAAACUAGGCG 5(sense strand)AL-175 UUGGUGAGGUUUGAUCCGCdTdT(guide strand)AL-135 UUGGUGAGGF|UUGAUCCGCdTdT AL-145 UUGGUGAGGUFUGAUCCGCdTdT AL-155 UUGGUGAGGUUFGAUCCGCdTdT AL-165 UUGGUGAGGFFFGAUCCGCdTdT 123 45678910 11 1213 14 1516 17 1819 2021B siRNA [nM]Figure 2. Concentration-dependent in vitro silencing activity of siRNAduplexes directed against rey luciferase relative to Renilla luciferase.(A) Incorporationof asinglerFresidueoppositeAinthesense[datanot shown, seeref. (41)] andguidestrands (1:10) is toleratedalbeit witha decreasein activityfor the latter(reference duplex1:2). On theother hand, incorporation of rFor U oppositeG is notor signicantlyless tolerated(1:9, 1:11and1:12, 1:19, respectively). (B) Contrarytothe presence of asingle rFresidue inthe guide strandopposite A(13:18, 14:18, 15:18;referenceduplex17:18),multiplerF:ApairsinansiRNAduplexpreventsilencing(16:18).IntheAL-10,AL-11,AL-13andAL-14oligonucleotides,therFsubstitutionisadjacent tothecleavagesite(markedbyavertical line).Nucleic Acids Research, 2007, Vol. 35, No. 19 6429 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from geometry of the rF:Gpair withthat of the previouslyvisualized rF:A pair (41), we determined the crystalstructure at 1.10 Aresolution of the RNAdodecamerduplex [r(CGCFAAUUGGCG)]2with rF:G pairs(FG duplex, Table 1). The structure of the RNAdodecamer duplex [r(CGCGAAUUAGCG)]2with A:Gmismatch pairs (GAduplex, Table 1) at 1.12 Awasdeterminedtopotentiallyprovide astructural rationalefor the inactivity of siRNA duplexes with an incorporatedmismatch pair at the cleavage site in in vitro assays.A summary of crystal data and renement parameters forboth structures is provided in Table 2. Sum(2FoFc)electron density maps based on the nal models of the FGandGAduplexesaroundselectedregionsaredepictedinFigure3.The geometries of the FG and GA duplexes (all ribosesadopt a C30-endo pucker) deviate froma more or lesscanonical A-form only in the vicinity of the rF:G and A:Gpairs(Figure4). rF:Gpairsexhibit thesheartypical forthe U:Gwobble pair wherebythe guanine anddiuor-otoluenemoieties arepushedintotheminor andmajorgrooves, respectively (Figures 1Cand 3C). In the GAduplex, A and G are both in the standard anticonformation and pair in a pseudo WatsonCrick fashion(Figures 1E and 3D). As a result, the minor groove of theGAduplex is widened by 1 Awith a concomitantnarrowing of the major groove. The FG structure featurestwo independent duplexes per crystallographic asymmetricunit, leading to four observations of the rF:Gpair.Both duplexes have very similar geometries as revealed byasuperimposition(Figure4B). Thestructureof theGAduplexwaspreviouslyreportedat alowerresolutionbyothers(1.9 A)(54).However,inthatstructurewithspacegroup P21 the asymmetric unit consisted of a single duplexin a general position (the presence of a strong non-crystallographic 2-fold symmetry was noted by theauthorsatthetime). Conversely, thespacegroupof theGARNAcrystalsanalyzedhereisC2, but theunit cellconstantsarevirtuallyidentical tothosereportedearlier.The conversion frompseudo-2-fold to crystallographic2-foldsymmetryismostlikelyaconsequenceofdierentcrystallization conditions (see Materials and Methodssection). IntheC2structure, theduplexislocatedonacrystallographicdyad andthe asymmetricunitconsistsofasingle dodecamer strand(Table 2). Summaries of thehelicalparameterscalculatedwiththeprogramCURVES(63) for the twoFGduplexes andthe GAduplex areprovidedin theSupplementary Data.Residuesin the GAduplexarenumberedC101toG112andresiduesinthefour independent strands of the FG duplexes arenumbered C101 to G112 (strand 1), C201 to G212(strand2), C301toG312(strand3) andC401toG412(strand4)(Figures3and5).The four rF:G pairs exhibit similar geometries(Figure 5); the (rF)C3-H . . . O6(G) hydrogen bonddistancevariesbetween3.18and3.31 A. Inthreeof thepairsthe(rF)F2. . .N1(G)distanceisbelow3.1 Aandinthe rF304:G409 pair this distance is somewhat longer(3.45 A, Figure5D). Therefore, theuorineamineinter-action here is considerably shorter than the sum of the vander Waals radii. If weassumethe radii of uorineandhydrogentobe1.35 Aand1.2 A,respectively(64),andaN-Hdistanceof1 A,theF. . .NdistancesinthreeofthefourrF:Gpairsareca0.5 Abelowthesum(3.55 A).Thewisdomof assigningalarger radius touorine thantohydrogen has been questioned recently (65). Still, even weassume the same radius for uorine andhydrogen(i.e.1.2 A ), theobserveddistancesbetweenF2andN1intherF:Gpairsareindicativeofanattractiveinteraction.Theshortestdistance(3.03 A)betweenF2(rF)andN1(G)seenhereisactuallymorethan0.8 Ashorterthantheaveragedistance betweenF4(rF) andN6(A) inrF:Apairs (41)!Thisdierenceisnoteworthybecausethedistancesofthe(rF)C3-H. . . O6(G)/N1(A) hydrogenbonds inthe rF:GandrF:Apairsareverysimilar.ThecloseresemblanceoftherF:GandU:Gpairs(Figure5)canitselfbetakenasevidence for an attractive interaction between uorine andN1-H(G). If the (rF)C3-H. . . O6(G) hydrogen bondprovided the only or the main contribution to the stabilityoftherF:Gpair, theshearseenbetweenFandGwouldseem quite unnecessary. Likewise, shearing does notbenet stacking interactions; if the geometry of rF:Gand rF:Apairs were chiey determined by stacking itwouldbediculttoexplaintheirdierentgeometries.ThesimilaritybetweenrF:GandU:Gisnotlimitedtopairinggeometrybutextendstonearestneighborinterac-tions and hydration. Thus, superimposition of (CpU):(GpG) and (CprF):(GpG) base-pair steps reveals virtuallyidentical conformations (Figure 6). The four rF:G pairs intheFGstructuredisplayaconservedhydrationpatternTable 2. Crystal data, data collection parameters and structurerenement statisticsStructure 50-CGCFAAUUGGCG-30(FG)50-CGCGAAUUAGCG-30(GA)Crystal dataSpacegroup P1 C2Numberofstrandsperasym.unit4 1Cell dimensionsa, b, c(A ) 22.90, 32.06, 42.93 41.34, 34.89, 32.12, , g(8) 96.2, 83.9, 93.9 90.0, 129.3, 90.0DatacollectionNumberofreections45 752 13 514Resolution(last shell)1.10(1.141.10) 1.12(1.161.12)Completeness(%, last shell)93.9(90.7) 96.4(89.3)Rmerge(last shell) 0.059(0.239) 0.035(0.277)RenementRwork/Rfree0.112/0.142 0.147/0.187Numberof atomsRNA 1016 255Water 324 76Ions 2Mg2+B-factorsNucleicacid(A2) 9.9 13.4Water(A2) 29.0 29.7r.m.s.d.Bondlengths(A ) 0.014 0.014Bondangles(1 3; A)0.029 0.0316430 Nucleic Acids Research, 2007, Vol. 35, No. 19 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from Figure 3. Quality of the nal Fourier (2FoFc) sumelectron density around the RNAduplexes in the region of the chemically modiedormismatchedbasepairs.(A)PortionoftheFGduplexviewedintotheminorgroove;(B)portionoftheGAduplexviewedintotheminorgroove;(C) rF104:G209intheFGduplexand(D) G104:A209intheGAduplex. Atoms arecoloredyellow, red, cyanandorangefor carbon, oxygen,nitrogenandphosphorus,respectively.Carbonatomsofthediuorotoluylmoietyarehighlightedinmagentaanduorineatomsareshownasgreenspheres.Figure4. Overall conformationof RNAdodecamerduplexeswithrF:Gpairs. (A) Oneof thetwoRNAduplexes(duplex1) percrystallographicasymmetricunit viewedintothemajorgroove. Carbonatomsof theFmoietyarehighlightedinmagentaanduorineatomsareshownasgreenspheres.(B)Superimpositionofthetwoduplexes;duplex1isredandduplex2isblue.(C)OverallconformationoftheRNAdodecamerduplexwithA:Gpairs; theviewisintothemajorgroovethat containsfourMg2+hexahydratecomplexes.Nucleic Acids Research, 2007, Vol. 35, No. 19 6431 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from (Figure 5BE) that is similar to the water structure aroundthe U:G pair (Figure 5A) (6669). In both cases the waterstructure extends to the periphery of the grooves (data notshown),withribose20-hydroxylgroupsservingasbridge-heads in the minor groove (70). The extensive hydration ofthe diuorotoluyl moiety observedinthe FGstructurefurther distinguishes the rF:G from the rF:A pair and is ingoodagreement withthe osmotic stress data discussedabove. Nowater molecules wereobservedat hydrogen-bonding distance from diuorotoluyls in rF:A pairs,althoughweneedtokeepinmindthattheresolutionsoftheFGandFA(41) structures aredierent (1.1versus1.6 A). Conversely, water molecules are located as close as2.93 Afrom uorine in the FG structure (Figure 5B). If weassume that hydrogenatoms fromwater molecules arepointing more or less in the direction of uorine (theresolution of the structure is not sucient to allowvisualization of H atoms), the observed OW-uorinedistances are inconsistent with mere van der Waalsinteractions.ACEBDFFigure 5. Geometry and hydration of rF:G, U:Gand G:A base pairs. (A) U:Gpair in structure NDB AR0009 (66); (B) rF104:G209;(C)rF204:G109;(D)rF304:G409;(E)rF404:G309;(F)G104:A209#(#designatesasymmetry-relatedresidue).HydrogenbondsintheU:GandG:AbasepairsandthecorrespondinginteractionsintherF:GpairsareshownasdashedlineswithdistancesinA . WatermoleculesandMg2+ionsarepinkandbluespheres, respectively.Figure6. StereoillustrationofthesuperimpositionofCpU-GpG[blue,NDBAR0009(66)]andCprF-GpG(red)dinucleotidesteps.TherFresidueislocatedonthelowerleft.6432 Nucleic Acids Research, 2007, Vol. 35, No. 19 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from TheGAstructurealsorevealedextensivehydrationofA:Gpairs inthemajor andminor grooves (Figure 5F)which again is in good agreement with high nw obtainedusingosmotic stress. Our structure at higher resolutionrevealed the coordination of four Mg2+-hexahydrate ionsinside the narrowand deep major groove. No metalcations wereobservedinthepreviouslyreportedcrystalstructureoftheGAdodecamer(54).Interestingly,Mg2+ions engage in water-mediated contacts to the majorgrooveedgesofbothGandA(Figure5F).Semi-empirical calculations of F:G and F:A pairs and theirnative counterpartsAll calculations were carried out with the programGAUSSIAN (57) and a counterpoise correction (58)was appliedineachcase. Weranequilibriumgeometryoptimizations as well as point-energy calculationsbased on the crystallographically observed geometries.Energies were computed both at the MP2/6-31GandMP2/6-31+Glevelsof theory. ThesearecomparedinTable 3for F:GandF:Apairs alongwiththeir nativecounterparts T:GandT:A. Notethat weusedTas thereference nucleobase due to the presence of a methylgroupin2,4-diuorotoluene. Equilibriumgeometriesforthe various pairs obtainedat bothlevels of theoryaredepictedinFigure7. Therelativerankingof energiesisT:G > T:A F:G > F:A with energy dierences betweentheT:GandT:AandtheF:GandF:Apairsofaround3 kcal mol1. TheT:Gpairis 8 kcal mol1morestablethanthe F:Gpair (MP2/6-31+Glevel). Interestingly,the order of stabilities for isolatedpairs obtainedfromsemi-empirical calculations is similar to the one estab-lished experimentally based on UV melting data forRNAduplexes containing these pairs (Uinsteadof T;in UVmelting UAis slightly more stable than UG).The geometry calculatedfor the F:Gpair (Figure 7A)showscloseresemblancetothegeometriesofrF:Gpairsinthecrystal structureoftheFGduplex(Figure5BE).On the other hand, the calculated and experimentallyobservedgeometriesoftheF:Apairdierdrastically;theF4 N6 distance in the calculated pair is ca. 0.7 Ashorter than the corresponding distance in the crystalstructure (41). Despite this divergence between thecomputedandexperimental pairinggeometries for F:Gand F:A, the dierences between the average point energybased on crystallographic data and the equilibriumgeometryenergyfortheF:G(5.1versus7 kcal mol1,respectively) andF:Apairs (2.9versus 4.1 kcal mol1,respectively)arequitesimilar.DISCUSSIONInsights into Ago2 substrate recognitionInvitrosiRNAexperiments inHeLacells usingadualluciferase assaydemonstratedthat unlike asingle rF:Apair, an rF:G pair at or adjacent to the cleavage site is nottoleratedbythe Ago2slicer enzyme. Similarly, asingleU:G mismatch drastically reduces activity and theonly mismatch pair exhibiting activity (somewhat reducedrelative to a canonical siRNA duplex) is the A:C pair (41).ThisisconsistentwitharequirementforaWatsonCrickgeometry of base pairs anking the cleavage site. It isunlikelythattheobserveddierencesinactivitybetweensiRNAs containing mismatches or the rF residue atvarious locations are related to other factors, i.e.eciencies in siRNAstrand dissociation and biases instrand loading. Importantly, there is no correlationbetweenthe thermodynamic stabilityandsilencingacti-vity of the investigated siRNAs (Figure 2A, Table 1).Moreover,detailedpositionalanalysesdemonstratedthata siRNA with the rF residue in the guide strand, adjacenttotheAgo2cleavagesiteledtoreducedsilencingactivityrelative topositions further removedfromthe cleavagesiteorat the50-terminus(41,71). rFinthesensestrandof the siRNAhad no eect on the activity (71). However,an earlier RACE(rapid amplication of cDNAends)analysis of the siRNA-induced cleavage of luciferasemRNAprovidesthestrongestsupportthattheobservedsilencing activities are directly relatedtoAgo2-inducedcleavage (41). Thus, cleavage at the normal site wasfoundwithsiRNAAL-10(rF:A,Figure2A),whereasnocleavage was observed with siRNA AL-9 (rF:G,Figure2A).AlthoughrFandAaremorelooselypairedrelativetoU:A, geometric constraints of the enzyme active sitemay produce a geometry that is not too dierent from thatof astandardWatsonCrickbase pair. Observations inconnection with in vitro DNA polymerization experimentsinvolvingF-containingtemplates or incomingdFTPareinstructive in this context (7275). Accurate incorpo-rationof dATPopposite template For dFTPoppositetemplate A is obviously not due to complementaryhydrogenbondformationbetweenthetwopartners(76).Rather,theactivesitesofhigh-delityDNApolymerasesact as amoldthat provides tight geometric constraintsforareplicatingbasepair,virtuallyforcingittoadopttoaWatsonCrick-likearrangement (77). Bycontrast, theTable 3. Energies (E in kcal mol1) obtained fromsemi-empiricalcalculations(CC=CounterpoiseCorrection)Basisset level/BasepairMP2 MP26-31G6-31+GMP2 MP26-31G6-31+G(CC) (CC)Point energycalculationbasedoncrystallographicdataF104:G209 8.3 6.9 4.4 4.8F204:G109 9.2 7.7 5.1 5.6F304:G409 7.2 6.4 4.2 4.7F404:G309 8.4 7.1 4.7 5.1F4:A21 5.4 4.9 2.6 2.8F16: A9 5.6 4.9 2.6 2.9averageF:G 8.3 7 4.6 5.1averageF:A 5.5 4.9 2.6 2.9EnergycalculationfrombasepairgeometryoptimizationT:A 18.2 16.5 12.3 12.7F:A 7.6 7.2 4 4.3T:G 19.6 19.3 14.1 15.5F:G 11.5 8.4 6.4 7Nucleic Acids Research, 2007, Vol. 35, No. 19 6433 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from A:G, U:G and rF:G mismatches cannot be accommodatedin this fashion. A superimposition of these mismatch pairsdemonstrates that they result inlarger deviations fromacanonical backbonegeometrythantherF:AandA:Cpairs (Figure 8). In fact the observed siRNA activities areinversely correlated with the backbone distortions presentin the RNAduplexes as a result of mismatch pairs.Althoughthe A+:Cpair withits twohydrogenbondsisexpectedtobeisostericwiththeU:Gwobblepair(78),theparticular exampleof theformer usedinthesuper-impositionexhibits a smaller backbone distortionthanboth U:Gand rF:G. Moreover, all measurements ofsiRNA activity were carried out at pH 7 and it is unlikelythat either A or C is protonated under these conditions. Aneutral A:Cpair withasingle hydrogenbondbetweenN6(A) and N1(C) or N1(A) and N4(C) may in fact displaya geometry that is even more similar to that of a standardWatsonCrickbasepair.Contrary to the correlation between local distortions inthe backbone geometry and in vitro siRNA activities, thereis noobvious relationshipbetweenactivityandthermo-dynamicstability. Thus, boththerF:ApairandtheA:Cpair (41) are more destabilizingthanthe U:Gor rF:Gpairs (Table 1). But the latter two lead to a loss of siRNAactivity whereas the former result in a somewhatdiminished activity compared to a canonical (andT:G (cc MP2-6-31G*)F:G (cc MP2-6-31G*)F:G (cc MP2-6-31+G**)T:G (cc MP2-6-31+G**)A B F:A (cc MP2-6-31G*)F:A (cc MP2-6-31+G**)T:A (cc MP2-6-31G*)T:A (cc MP2-6-31+G**)C D Figure7. Equilibriumgeometriesofthe(A)F:G, (B)T:G, (C)F:Aand(D)T:Apairsbasedonsemi-empirical calculations[ccMP2/6-31GandccMP2/6-31+Glevels; cc =counterpoisecorrection(58)]. DistancesareinA ngstrom.6434 Nucleic Acids Research, 2007, Vol. 35, No. 19 at National Institute of Science Education and Research, Bhubaneswar on October 13, 2014http://nar.oxfordjournals.org/Downloaded from thermodynamically more stable) U:Apair at the Ago2cleavage site (Figure 2). The stability data for thedodecamerduplexeswithmismatchesprovidedhereandthose based on longer siRNAduplexes published pre-viously (41) are consistent with the thermodynamics at pH7 of RNAhairpinscontainingsingleinternal mismatches:G:C A:U U:G>G:A>A:C(79).UnlikesiRNAswithsinglerFresiduesinplaceofUintheguidestrandthat retain activity [ref. (41) and this article], and appear toenhance sequence selectivitybeyondthat of the naturalbaseinsomecases(71),ansiRNAwiththreeconsecutiverF:Apairsspanningthecleavagesitedidnotdisplayanyactivity(Figure2B).Thisobservationcouldindicatethatmultiple rF:Apairs result inmoreseverehelical distor-tions (80) that are not toleratedby the Ago2 enzyme.Alternatively, the loss of activity may be related toa signicant weakening of the hybridization betweenguide strand and mRNAtarget, or a combination ofchangesinstabilityandconformation.rF:G and rF:A pairs display considerable differencesThe main surprise regarding the geometry of the rF:G pairishowcloselyitresemblesthenativeU:Gpair,includingthe shear that is ahallmarkof the wobble pair. Basedon the signicantly larger separation between rF andArelativetoUandA(>0.8 Aonaveragebasedonthevarious atompairs), one wouldperhaps have expecteda looser association of rF and G as well, weakly stabilizedby a C-H Nhydrogen bond. Instead, the averagedistance between atom pairs in the four independent rF:Gpairs(Figure5BE)presentinthecrystalstructureoftheFGduplexis just 0.3 Alonger thanatypical hydrogenbonding distance of ca 2.9 Ain the U:G pair (Figure 5A).Not only is the pairing tighter than expected, but the rF:Gpairsdisplayaconservedhydrationpatternthatmatchesthe arrangement of water molecules aroundthe nativeU:Gwobble pair. Bycomparison, the hydrationof therF:Apairwasmuchmorelimited(41).Itisunlikelythatthisobservationissimplyduetothelowerresolutionoftheearlierstructurecomparedtothat of theFGduplex(1.6 A). Osmoticstressmeasurementsareconsistentwiththe structural data and support the notion that therF:Gpair is surrounded by more solvent molecules.Furthermore, the thermodynamic stability (TM) of anRNAdodecamer duplex containing two rF:Gpairs ishigherthanthat of thecorrespondingduplexwithrF:Apairs (Table1).ThisisnoteworthybecauseU:G pairsaretypicallylessstablethanU:Apairs(79).Byanalogy,onewouldhavepredictedthatrF:GpairsarelessstablethanrF:Apairs. Indeed, our ndings with modied RNAduplexes dier frompublished stability data for DNAdodecamer duplexes withsingle F:Gor F:Apairs (81).Accordingly, the relative order of stability in DNA is T:A(TM=508C)>T:G(438C)>F:A(368C)>F:G(358C).Although there are no crystal structures of DNA duplexeswithincorporatedF:AorF:Gpairsitisunlikelythatthedierences between DNA and RNA arise from dierentialstacking interactions involving the diuorotoluene moiety.Incorporation of Fin place of Tinto DNAleads toimproved stacking (82). By comparison, the particularpairingmodebetweenF/TandGpushestheformerintothemajorgroove, thuspresumablyreducingtheoverlapwiththesurroundingnucleobases.Overall,X-raycrystal-lography, osmoticstress measurements andUVmeltingdataallclearlyindicatethatFpairsmorefavorablywithGthanwithAinthecaseofRNA.Organic fluorine as a hydrogen-bond acceptorThequestionwhetheruorinecanact asanacceptorofhydrogenbondshasrecentlybeendiscussedindetailandwith authority (65,83). In general, uorine is a pooracceptordespitebeingveryelectronegative, andascreenofsmall moleculecrystal structuresonlyrecoveredafewexampleswithshort X-Hd!F-Ycontacts(2.1> G:rF G:U A:G local backbone distortion: A:U < A:C A:rF