Optimization of the methods for small peptide

《Optimization of the methods for small peptide》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

ISSN00268933,MolecularBiology,2010,Vol.44,No.6,pp.958–967.©PleiadesPublishing,Inc.,2010.OriginalRussianText©A.N.Istrate,A.B.Mantsyzov,S.A.Kozin,V.I.Polshakov,2010,publishedinMolekulyarnayaBiologiya,2010,Vol.44,No.6,pp.1075–1085.STRUCTURAL–FUNCTIONALANALYSISOFBIOPOLYMERSANDTHEIRCOMPLEXESUDC577.322.53;531.395OptimizationoftheMethodsforSmallPeptideSolutionStructureDeterminationbyNMRSpectroscopyA.N.Istratea,A.B.Mantsyzova,S.A.Kozinb,c,andV.I.PolshakovdaChemicalDepartment,MoscowStateUniversity,Moscow,119991RussiabResearchInstituteforBiomedicalChemistry,RussianAcademyofMedicalSciences,Moscow,119121RussiacEngelhardtInstituteofMolecularBiology,RussianAcademyofSciences,Moscow,119991RussiadCenterforMagneticTomographyandSpectroscopy,FacultyofFundamentalMedicine,MoscowStateUniversity,Moscow,119991Russia;email:vpolsha@mail.ruReceivedJune29,2010PresentedbyA.A.MakarovAbstract—NMRspectroscopywasrecognizedasamethodofproteinstructuredeterminationinsolution.However,determinationoftheconformationofsmallpeptides,whichundergofastmolecularmotions,remainsachallenge.ThisismainlycausedbytheimpossibilitytocollecttherequiredquantityofthedistanceanddihedralanglerestraintsfromNMRspectra.Atthesametime,shortchargedpeptidesplayanimportantroleinanumberofbiologicalprocesses,inparticularinpathogenesisofneurodegenerativediseasesincludingAlzheimer’sdisease.Therefore,developmentofamethodforstructuresimulationofsmallpeptidesinaqueousenvironmentusingthemostrealisticforcefieldsseemstobeofcurrentimportance.SuchalgorithmhasbeendevelopedusingtheAmber03forcefieldandGromacsprogramaftermodificationofitscode.Calculationalgorithmhasbeenverifiedonamodelpeptidewithaknownsolutionstructureandametalbindingfragmentofratβamyloid,whosestructurehasbeendeterminedbyalternativemethods.Thedevelopedalgorithmsubstantiallyincreasesstructurequality,inparticularRamachandranplotstatistics,anddecreasesRMSDofatomiccoordinatesinsidethecalculatedfamily.Thedescribedprotocolcanbeusedfordeterminationofconformationofshortpeptides,andalsoforoptimizationofstructureoflargerproteinscontainingpoorlystructuredfragments.DOI:10.1134/S0026893310060130Keywords:NMRspectroscopy,moleculardynamics,peptidestructure,βamyloid,AlzheimerdiseaseINTRODUCTIONteases[5,6].Aβisanormalcomponentofbiologicalfluids(blood,cerebrospinalfluid)andpresentedthereinManyneurodegenerativediseasessuchasAlzherelativelylowconcentrations(approximately5–20nM),imer’s,Parkinson’s,andpriondiseasesareassociatedsimilarforbothsickandhealthypeople[7,8].Accordtotheaggregationofrelativelysmallpolypeptides[1,ingtothewidelyacceptedamyloidhypothesis,thekey2].DeterminationofstructuraldetailsofprocessesmoleculareventleadingtotheemergenceoftheADisthatdeterminethetransitionfromnonpathogenicsoltheconformationaltransitionofsolublemonomerAβ,ublepeptidestotheaggregatesresistanttoproteolysisfirst,toneurotoxicdimersandoligomers,then,toisofgreatinterestbothforunderstandingofmolecularinsolublefibrillarpolymericaggregates,whichformmechanismsofthesediseasesanddevelopmentofamyloidplaques[9].methodsfortheirpreventionandtreatment.ThekeySpatialstructureofmodelAβfibrilswasrecentlyroleintheinitiationanddevelopmentofAlzheimer’sdeterminedusingsolidstateNMR[10–12].Itwasdisease(AD)belongstothebetaamyloidAβ[3,4],shownthattwofragmentsoftheAβmoleculeplaysigconsistingof39–43a.a.,whichisformedfromtransnificantlydifferentrolesinfibrilstructure.ThelargermembraneprecursorproteincleavedbyspecificproCterminalfragment(from18to40–42a.a.)formsβsheet,consistingoftwoantiparallelstrands,whichAbbreviations:AD—Alzheimer'sdisease;Aβ—Alzheimer’sdisorganizeintoahydrophobicβsheetstructureintheeasebetaamyloidprotein;RatAβ(1–16)—Nterminalmetalinteractionofβlayersofneighboringmolecules.Nterbindingdomainofratβamyloidprotein;DQFCOSY—DoubleQuantumFilteredCOrrelatedSpectroscopY;TOCSY—minalβamyloidfragment(1–17a.a.)isonthesurTOtalCorrelationSpectroscopY;NOESY—NuclearOverfaceofpolymericaggregateandnotinvolvedinthehauserEnhancementSpectroscopy;NOE—NuclearOverformationofafibrillarhydrophobiccore.NeverthehauserEffect;HSQC—HeteronuclearSingleQuantumCoherless,theabsenceofthisfragmentcompletelyblocksence;AMBER—aforcefieldusedforcalculationofstructureofbiomolecules,whichdescribesvalentandnonvalentinteractionsfibrillogenesisinvivo,indicatingthatitplaysakeyrolebetweenatoms;RMSD—rootmeansquaredeviation.intheformationofamyloidplaques.Itwasestablished958 OPTIMIZATIONOFTHEMETHODSFORSMALLPEPTIDESOLUTIONSTRUCTURE959earlierthattheNterminal(1–16a.a.)fragmentisaICMDprogram[30]usesanAMBERforcefield[31]metalbindingdomain[13–16],bindingionsofzincandalsooperatesinthespaceoftorsionanglesinaandsomeothermetals[17,18].Itshouldbenotedthatvacuum,but,however,itincludesthemodelpotentialamyloidplaquesareindeedcharacterizedbyahighofchargeinteractionsfromthefirststepofthecalcucontentofdivalentmetals,particularlyzinc(upto1lation.Furthermore,theproblemoftrajectoryinstamM)[19].Inarecenttheoreticalstudy[20],itwasbilitycausedbytheintroductionofthispotentialisshownthattheconformationoftheNterminalsolvedbyscalingtheweightsoftheexperimentalmetalbindingdomainandpossibleintermolecularNMRrestraintsforMDhightemperaturestage.coordinationofzincionsdrivetheaggregationofRecently,electrostaticpotentialhaveoftenbeenhumanβamyloid.βamyloidisfoundinallmamintroducedintoproteinstructurecalculationatthemals;however,itisinterestingtonotethatratshavenofinalstageaftersimulatedannealing[32].However,weaggregationprocessofAβ[21],whileproteinarenotfamiliarwithalgorithmsthatwouldallowthesequencesofhumanandratpeptidesdifferbyonlyentireprocessofbiomoleculecalculationintheCartethreesubstitutionsatmetalbindingdomain.Itisobvisiancoordinatesinanaqueousenvironmentintheousthatthedeterminationofratmetalbindingβamyrepresentationofwatermoleculesinanexplicitform.loiddomainstructureisofgreatimportanceforexamiSuchanalgorithmcouldprovidethemostrealisticnationofmolecularnatureofAlzheimer’sdisease.results,althoughitrequiressignificantcomputationalBecauseoftheimpossibilitytoobtainβamyloidcosts.Havingopensourcecode,Gromacs[33]hassinglecrystalsandtheircomplexeswithmetalions,goodprospectsfortherealizationofsuchanalgoNMRspectroscopyistheonlymethodtoobtainstrucrithm.Inthegivenpaper,themodificationofthisproturalinformationaboutthestudiedobjects.Determigramwasdeveloped,andanalgorithmwasdesignednationofglobularproteinsstructureusingNMRisaprovidingthecalculationofsmallpeptidestructureincomplexbutrelativelywellsolvedproblem[22].aqueousenvironmentinCartesiancoordinatesusingMethodsfordeterminationofthesystemminimumelectrostaticpotentialontheearlystagesofthecalcuenergyusedinproteinstructuresimulationareusuallylation.Thecalculationprotocolwastestedonamodelbasedonmoleculardynamicsand“simulatedannealpeptideandusedforsimulationofmetalbindingrating”algorithm[23].Forcefieldsthatdonottakeelecβamyloiddomainstructure.Obtainedstructuresweretrostaticinteractionsintoaccountareusedatleastatcomparedtosimilarstructures,calculatedusingstanthestagesofmoleculardynamicsathightemperature.dardprotocols,implementedbyCNSprogram.Thisisbecausetheenergyofelectrostaticinteractionsoftenleadstosysteminstabilityduringsimulation.ThemoredifficulttaskistodeterminethestructureofEXPERIMENTALsmallconformationallymobilepeptides.Inthiscase,NMR.SyntheticpolypeptideDAEFGHDSGFEthenumberofexperimentallydeterminedrestraintsVRHQK,identicaltotheNterminalfragmentofratoninteratomicdistancesanddihedralanglesisusuallyAβ(RatAβ(1–16))waspurchasedfromBiopeptideCo.,small,andelectrostaticinteractionsbetweenchargedLLC(USA),and5mMsolutionofRatAβ(1–16)inagroupsofthepeptideshouldbetakenintoaccountto20mMTrisDpH6.5buffersolutionwasused,con11obtainreliablestructuraldata.taining0.1%NaNtopreventpeptidebiodegradation.3ExperimentersuseanumberofsoftwarepackagesNMRspectraweremeasuredat5°CintheD2OortocalculatethebiomoleculestructureusingNMR.90%H2O/10%D2OonaBrukerAVANCEspectromCNSandXPLOR/NIHprograms[24,25]areeter(Germany)operatingat600MHz1Hfrequency,appliedforcalculationswithouttheuseofelectrostaticequippedwithatripleresonance(1H,13Cand15N)potential.CNSmodificationusinganARIAmodulepulsedfieldzgradientprobe.ForpeptidesignalfortheautomaticassignmentofnuclearOverhauserassignment,DQFCOSY,TOCSY(mixingtimeofeffectsignals(NOE)[26–28]suggeststheuseoftwo70ms),NOESY(mixingtimeof200and500ms)weremethods:structuresimulationinvacuumusingastanmeasured,aswellas13C1HHSQCand15N1HdardCNSprotocolandthesubsequentoptimizationHSQC,measuredatnaturalabundanceof13Cand15Nofstructuresobtainedbycalculationofshorttrajecisotopes.Theassignmentofsignalswasmadeinaccortoryintheaqueousenvironment,involvingtheuseofdancewiththeclassicaltechniqueofK.Wuthrichelectrostaticpotentialatthelatestepsofcalculation.[34],supplementedbyanalysisofHSQCspectra.Inthiscase,themodelofsocalledthinlayerofsolventAlmostcompleteassignmentof1Hand13Csignalsofisused.Trajectorystabilityupontheintroductionoftheaminoacidsidechainsand15Nsignalsofamidechargesintothemoleculeisprovidedbylowtemperafragmentofthepolypeptidechainwasperformed.turemoleculardynamicsandtheuseofbiomoleculeSubsequentanalysisofNOESYspectraallowedustostructure,whichwasalreadyformedatthefirstdetermine111restraintsoninternucleardistancesapproximationofsimulatedannealinginavacuum.usedtocalculatepeptidestructureinsolution.DetailsCYANAprogram[29]allowssimulationofproteinofNMRspectraassignmentandcomparativeanalysisstructureinthespaceofinternalcoordinates(torsionofthestructuralinformationisasubjectofanotherangles)withoutregardtoelectrostaticpotential.Thepublication.MOLECULARBIOLOGYVol.44No.62010 960ISTRATEetal.(a)Inordertomaintainaconstanttemperatureandpressure,Berendsenthermostatandbarostatwithafunctionofrescalevelocitywereused.Thelengthsofcova30lentbondsduringMDwerecontrolledbyLINCSalgorithm[37].EnergyofelectrostaticinteractionswascalculatedusingtheEwaldparticlesgridalgo20rithm.ThemaximumdistancesoftheCoulombandvanderWaalsforceswere1.0and1.4nm,respectively.TheintegrationstepforMDcalculationwas2×1010–15seconds.Duringoptimizationoftheprotocol,thefollowingparametersweremodified:temperature,weight0restraintsontheinternucleardistanceandchange51015profileduringthecalculation,incubationlengthdur(b)ingthetemperaturelowering,andtherateoftemperaturedecrease.ToprovidetheabilitytochangetheNOEnumberweightoftheexperimentalrestraintsduringcalcula30tion,thecodeofGromacswaschanged.Detailedchangescanberequiredfromtheauthors.AcompletelistoftestedprotocolsisgiveninTable1.Thefamilyof2010structureswascalculatedforeachprotocol.Convergenceofthecalculationresultswasdeterminedusingrootmeansquaredeviation(RMSD)of10thecoordinatesofatomsC,Cα,andNofproteinchainstructurescalculatedforallthefamily(Fig.2b).Theeffectivenessofsolvationwasdeterminedbythenum0berofhydrogenbondsformedbythepeptide(Fig.2d).51015Thedegreeofsystemcondensationafterheatingwasa.a.numbermonitoredbyvolume(Fig.2b),pressure,anddensity.ThetrajectorystabilitywasestimatedbythenumberofFig.1.Histogramofthenumberandtypeofrestraintsonsuccessfullycompletedcalculations.Thequalityofthetheinternucleardistance(NOE)foraminoacidresiduesofcalculatedstructureswasdeterminedbythepercentofa(a)modelpeptideand(b)RatAβ(1–16).DifferenttypesofhitsofmaindihedralanglesφandψintothemostNOEaremarkedasfollowing:atomsofthesameaminofavorableandprohibitedareasofRamachandranmapacidresidueareconnectedbylightgray,atomsofneighusingProcheckNMR[38].boringresiduesbydarkgray,andatomsseparatedfromeachotherbytwotofourresiduesalongthechainareconIncaseoftheoptimalprotocol(no.18,Table1),anectedbyablackline.calculationwasperformedusingalltherestraintsusedinthecalculationofdepositedinthePDBfamilystructures(145NOE,2hydrogenbondsand15diheOptimizationoftheprotocolofsimulatedannealingdralangles).wasperformedusingGromacsonamodelpeptideconCalculationoftheRatAβ(1–16)structureinGromacs.sistingof17aa(1D9LPDBid),whosestructureconAsystemwaspreparedforcalculationsimilartothattainsasinglealphahelix[35].Duringoptimization,describedaboveforthemodelpeptide.Thenegative119restraintsontheinternucleardistancewereused,chargeofthesystemwasneutralizedbythereplacedepositedbytheauthorstogetherwiththecoordinatesmentofwatermoleculesonthecorrespondingnumoftheatoms(Fig.1a).ExtendedchainconformationberofNa+ions.wasusedastheinitialstructure.ApeptidemoleculeCalculationofRatAβ(1–16)structureswasperformedwasplacedintoacubiccellof0.5×0.5×0.5nm,thenusingtheoptimalprotocol(no.18,Table.1).Atthe13710watermoleculesofSPC/Emodelwereaddedfirststep,thepeptidewasheatedto1500Kandsolvent[36].Thetotalchargeofthesystemiszero,sothetemperaturewas600K.Next,thesystemwasthermointroductionofadditionalionsforneutralizingthestaticallycontrolledwithin1ps,thenpeptidetemperamolecularsystemisnotrequired.Potentialenergywasturewasdecreasedinincrementsof100Kandatminimizedintwostages:first,thealgorithmofsteep100K/pstoequalthetemperatureofthepeptideandestdescentwasused,thentheconjugatedgradientsolvent(600K).Aftereachstep,thesystemwastheralgorithm.ThenMDtrajectoryof20pswithrestrictedmostaticallyequilibratedwithin1ps.Next,thecoolpositionsofallpolypeptideatomswascalculatedingratewasdecreasedto50K/ps,andtheincubation(positionrestrained).AforcefieldAMBER03wastimewasincreasedto3ps,andthesystemwascooledused[31].Theresultingsystemcanbeusedasinitialto0Kwitha100Kincrement.Atatemperatureofforsimulationofstructuresusingdistancerestraints.300K,lengthofsystemtemperatureequilibrationstepMOLECULARBIOLOGYVol.44No.62010 OPTIMIZATIONOFTHEMETHODSFORSMALLPEPTIDESOLUTIONSTRUCTURE961Table1.ProtocolsofsimulatedannealingtestedduringoptimizationofmodelpeptidestructurecalculationResultsProtoTemperatureRestraintTemperatureTrajectoryRMSD,colchangemode,weight,Coolingtime,psequilibrationThePeptideSystemlength,psÅNo.Kunitstime,pssystemisstructureiscondenstablecorrectsationProtocolswithaconstantweightofrestraints12000300100014425––––215003001000104252.6–+–31100300100076253.5–––4700300100056253.6+–+5150001000126252.7–+–61500010001402psto300K,then52.7–+–1psLinearchangeofrestraintweight7150000100075252.2–+–Stepwisechangeofrestraintweight870000100020124.7+–+/–970000100035234.5+–+/–107000010005223,20pstemper4.0+–+aturecontrolatT=300K118000010005022,20pstemper4.7+–+/–aturecontrolatT=300K128000010004723,10pstemper4.4+–+/–aturecontrolatT=300K13150000100045124.9+––14150000100060222.7++–15150000100075233.4++–161500001000902psdownto300K,3psdownto2.7++–then4p300K,then6ps171500001000581psdownto700K1psto700K,2.6++–then2ps3psdownto300K,10psat300K,then3ps18.1Peptide:01000581pstoTpeptide=1pstoTpeptide=1.7+++15000downto600K,then2ps600K,3psSolvent:600K,thendownto300K,6000100010psat300K,4000then3ps18.26000400084481.3+++wasincreasedto10ps.Weightlimitshavechangedthermostaticallyequilibratedfor8ps.Atthesecondduringsimulatedannealingfrom0to4000ateachsteps,aconstantweightofrestraintswasused.Thestageoftemperaturelowering.MDtrajectorylengthatlengthoftheMDtrajectoryatthesecondstepwasthefirststagewas58ps.84ps.ThetotaltimeofcalculationintwostagesusingAtthesecondstepthesystemwasheatedto600K,320processorswasapproximately16minutesforathermostaticallyequilibratedwithin8ps,andthenfamilyof20structures.cooledto0Kwithadecrementof100KandarateofIncalculationofthestructureRatAβ(1–16),11125K/ps.Aftereachstepofcooling,thesystemwasrestraintsonadistancewereused.IncontrasttotheMOLECULARBIOLOGYVol.44No.62010 962ISTRATEetal.1600(a)tialsoftheforcefieldwerechangedinaccordancewith1protocolanneal.inp.AsinthecasewithGromacs,as1200theinitialstructureusinganextendedpolypeptidechainwasusedasinitialstructure,andthesamesetof8002experimentalrestraintswasused.Inadditiontothe400standardsetofpotentialsusedinanneal.inp,potentialTemperature,KofRamachandranmapswasactivated[39],which0increasesthepenaltyfunctionofenergy,ifthedihedral(b)800anglevaluesoftheproteinchainfallintotheprohib2itedareaofRamachandranmap.Thisforcefieldmin6001imizesthenumberofobviouslyerroneousconforma2tionsofproteinchain.ThequalityofcalculatedstrucÅ,V400tureswasdeterminedsimilarlytothatdescribedaboveforthecalculationoftheprogramGromacs.200CalculationoftheRatAβ(1–16)structureinCNS1.0(c)withsubsequentoptimizationofsolventthinlayer.0.8RatAβ(1–16)structures,obtainedaftercalculationina0.6vacuumusingCNSwerefurtheroptimizedwitha1refine_water.inpprotocol,whichuseselectrostatic0.4RMSD,nmpotential[27,32]butnotpotentialofRamachandran0.2map.WatermoleculesofTIP3Pmodelwereaddedto20thecellcontainingthestructurewithafurthersimula80(d)tionofasolventthinlayer.Potentialenergywasmini60mizedusingthePowellalgorithm,afterwhichthemolecularsystemwasheatedupto500K.TheMD40trajectoryof4pslengthwascalculatedatthistemper2ature,andthenastandardprocedureforcoolingto200KandthefinalstageofenergyminimizationwereNumberofHbonds1performed.Thepotentialofdistancerestraintshave0204060notchangedfromthelevelof20kcal/mole.AsaresultLengthofMDtrajectoryofcalculation,weobtainedafamilyof40structures.Fig.2.Changeofthesystemparametersduringthetrajectoryofmoleculardynamicsprotocol18andcalculationofRESULTSANDDISCUSSIONthe1D9Lmodelpeptide.Boldline(1)hightemperatureOptimizationoftheSimulatedAnnealingProtocolstage(protocol18.1);thinline(2)stageoffurtheroptimization(protocol18.2).(a)ChangesintemperatureoftheinGromacspeptide(K).(b)Changethevolumeofthesystem,Å2.(c)OptimalstructureofpolypeptidewiththelowestStandarddeviationofcoordinatesofheavyatomsofthevaluesofRMSDincoordinatesofatomsofthemainchainpeptide,nm.(d)ThenumberofhydrogenbondswithinthepeptideandbetweenthepeptideandobtainedfamilywasformedonlyonthelongMDtrawatermolecules.jectorieswithinitialtemperatureof1500K(Table1).However,inthetestingofsimulatedannealingofpolypeptideinthesolventcell,itwasfoundthatatthemodelprotein,RatAβ(1–16)ispoorlystructured,whichtemperatureabove1000KandstandardweightsofisevidentfromtheNOEdistribution(Fig.1b).Theexperimentalrestraintsontheinternucleardistancesqualityofobtainedstructureswasdeterminedbytheanddihedralanglesthesystemisunstable,resultinginabsenceofrestraintviolations,aswellasthecharacteranemergencystopofcalculation.ToovercomethisisticparametersofRamachandranmap.problem,wemodifiedtheprogramcodeofGromacsSimulationoftheRatAβ(1–16)structureinCNSinatomakeitpossibletochangetheweightsoftheexpervacuum.CalculationusingtheCNSprogram[25]imentalrestraintsduringcalculation.Introductionof(version2.1)wasperformedusingastandardprotocolstepwiseincreaseofrestraintsweightprovidestosigofsimulatedannealinganneal.inpinCartesiancoordinificantlyincreasethetrajectorystability(Table1).nates.Thisprogramusesitsownforcefieldparallhdg,AnotherproblemcausedbytheuseofhightemperaoptimizedforthecalculationofshortmolecularturesinMDprotocols,isassociatedwiththelackofdynamicstrajectoriesusingexperimentalrestraints.solventcondensationaftercellenlargementduetoProtocolcalculationincludedtwosteps:(a)slowcoolheating.Asaresultofbadsolventcondensation,sysingfrom1000to0Kand(b)finalenergyminimizationtemvolumeincreasesbyalmostanorderofmagniofthestructureusingthePowellalgorithm.Thetude.However,theprotocolswithinitialtemperatureweightsoftheexperimentalrestraintswereconstantnotexceeding700Kshowedagoodcondensationofduringeachofthestages.Theweightsofotherpotenthesystem,butthistemperaturewasinsufficientforMOLECULARBIOLOGYVol.44No.62010 OPTIMIZATIONOFTHEMETHODSFORSMALLPEPTIDESOLUTIONSTRUCTURE963(a)(b)180180TRP2~b~bbbββ135b~b135b~b~l~l9090ll45L45Laaαα00~a~aPsi,degrees–45TRP2Psi,degrees–45TRP2–90–90~b~p~b~p–135–135bpbp~b~b–180–135–90–4504590135180–180–135–90–4504590135180Phi,degreesPhi,degreesFig.3.Ramachandranmap(proteinchaindihedralanglesφandψ)of(a)20conformersof1D9Lfamilyand(b)20conformerscalculatedinGromacsusingprotocol18.ThemostfavorableareaofRamachandranmapisshownindarkgray.ImageswerecreatedwithProcheckNMR[38].correctproteinfolding.Thisproblemissolvedbymined(Fig.4d).Atthesametime,aromaticgroupsofusingdifferenttemperaturesforpeptidemoleculestheseresiduesaredeterminedlessaccuratelyintheandaqueousenvironment(Table1,protocol18.1),family1D9L,resultinginhighRMSDofatomiccoorandadditionaloptimizationofthesystemaftercomdinatesofthesegroupsandsignificantdifferencesofpletionofthesimulatedannealing.Accordingtoprotheirorientation,ascomparedwiththefamilycalcutocol18.1,proteintemperatureattheinitialmomentlatedusingtheGromacs(Fig.4c).was1500K(Fig.2a),andsolventtemperaturewas600K.Aftercoolingofproteinmoleculedownto600K,afurthertemperaturedecreaseoccurredsimulComparisonofRatAβ(1–16)StructuresCalculatedtaneouslyfortheproteinandaqueousenvironment.UsingDifferentProtocolsProtocol18showedthebestresultsamongallTheRatAβ(1–16)polypeptidehastwodomainsconexaminedprotocols.Itconsistsoftwosteps:highnectedbymobilelinker:Nterminaldomain(a.a.1–temperaturesimulatedannealing18.1andfurther6)andCdomain(a.a.7–16).Theirlocationwithinoptimizationofthesystem18.2(Fig.2).Fourfoldthepolypeptidesequenceisconsistentwiththedistriincreaseoftherelativeweightofexperimentallimitsbytheendofeachstage(Table1)allowsforobtainingbetterconvergenceinthefamilyofcalculatedstrucTable2.Comparisonoffamiliesofmodelpeptidecalculatedtures.usingprotocol18and1D9LfamilydepositedintoPDB[35]Comparisonofmodelpeptidestructuresderived1D9LGromacs,usingprotocol18withthecharacteristicsofthestrucStatisticsofcalculatedstructuresfamilyprotocol18turesdepositedinPDBandcalculatedwithouttheelectrostaticpotentialandaqueousenvironment[35]Violationofdistancerestraints,Å0.0660.096showsthatthedevelopedprotocolsignificantlyPercentageofaminoacidresidues80.397.6improvestheparametersofRamachandranmapinthemostfavorableareaofRam(Fig.3,Table2).ItisimportanttonotethattheusedachandranmapcalculationprotocoldoesnotusepotentialofRamachandranmaps.Percentageofaminoacidresidues0.90.0intheprohibitedareaofRamThestructurescalculatedusingtheprotocol18areachandranmapsignificantlybetterstructuredintheterminalfragments(Fig.4b)ascomparedwiththefamily1D9LRMSDofcoordinatesofheavyatoms2.52±0.560.99±0.06(Fig.4a).Theyalsohaveanimprovedfoldingofaminoofproteinchain(C',Ca,OandN)*acidsidechains.Forexample,conformationofphenyl*RMSDvalueofcoordinatesofheavyatomsofproteinchain(C',residuesofPhe5andPhe10inafamilyofstructuresCα,andN)inpairwisesuperpositionofthefamilyofstructurescalculatedusingtheprotocol18isuniquelydetercalculatedinGromacsand1D9Lfamilyis2.46±0.03.MOLECULARBIOLOGYVol.44No.62010 964ISTRATEetal.(a)(b)(a)(b)(c)(d)(c)Fig.4.Superpositionof(a,c)20structuresof1D9LandFig.5.Superpositionof(a)threefamilieson40structures(b,d)20structuresofthemodelpeptide,calculatedinofpeptideRatAβ(1–16)calculatedinGromacsusingproGromacsusingprotocol18.Theconformationofthe(a,b)tocol18;(b)inCNSinvacuum,and(c)CNSwiththemainpolypeptidechainand(c,d)foldingofthesideoptimizationinathinlayerofsolvent.Thesuperpositionischainsofaminoacidresidues.madeofallatomsofC',Cα,Nofpolypeptidechain.butionofthenumberofdistancerestraints(Fig.1b).protocols(Fig.6).However,thepeptidestructurehasFigure5showstheresultofsuperpositionofcalculatedthehigherqualityintheaqueousenvironmentcalcufamiliesusingcoordinatesofallheavyatomsoflationexperimentaboveusingGromacs.Itisdemonpolypeptidechain.ItisevidentthatthemutualorienstratedintheparametersofRamachandranmapsandtationofdomainsvariesfromstructuretostructure,RMSDvaluesofcoordinatesofproteinchainheavybutthefamilysimulatedusingGromacs(Fig.5a)isatoms(Table3).Thisprotocolismosteffectiveincasemoredenselypackedascomparedwiththestructuresofinsufficientnumberofexperimentalrestraints.determinedusingCNS(Figs.5b,5c).Thus,orientationofthearomaticPhe4andPhe10resThealgorithmofsimulatedannealinginaqueousidues,havingasufficientnumberofrestraintsontheenvironmentwithelectrostaticpotentialfromtheveryinternucleardistance,isalmostidenticalforallthreefirststepofsimulationcansignificantlyimprovetheprotocols.However,thepositionofthelongsidechainparametersofproteinchainfolding(Table3)comofArg13,whichhasasmallnumberofexperimentalparedwiththecalculationinavacuum,evenusingrestraints,isdeterminedbetterusingtheprotocolartificialpotentialofRamachandranmap.Calculabasedonanaqueousenvironment(Fig.6).tionprotocolinCNSsoftware,whichdoesnotusetheAswasnotedabove,determinationofthestructurepotentialofRamachandranmap,leadstoamoreofsmallpeptidesisassociatedwithdifficultiescausednoticeabledecreaseinthenumberofaminoacidresibytheirhighmobility.EvenwhenthereisadominantduesinthemostfavorableareaofRamachandranmapconformationofthepolypeptidechain,itisusually(Table3).impossibletoobtainalargenumberofconstraintsonGeneralfoldingoftheproteinchainoftwotheinternucleardistancesanddihedralanglesfromdomainsofthepeptide(NandCdomains)isalmostNMRspectra.However,theuseofrealisticforcefieldsidenticalforfamiliescalculatedusingthreedifferentforcalculationoflongmoleculardynamictrajectoriesMOLECULARBIOLOGYVol.44No.62010 OPTIMIZATIONOFTHEMETHODSFORSMALLPEPTIDESOLUTIONSTRUCTURE965Table3.Parametersof40structuresofRatAβ(1–16)calculatedusingdifferentprogramsandprotocolsUsedprogramandcalculationprotocolStatisticsofcalculatedstructuresGromacsCNSCNScalculationcalculationinvacuumusingoptimizationinthethininaqueousRamachandranmaplayerofsolventenvironmentParametersofRamachandranmap%ofa.a.inthemostfavorablearea79.877.160.0ofthepotentialofRamachandranmap%ofa.a.intheprohibitedarea0.00.20.6ofRamachandranmapAveragenumberofviolationsofdistancerestraintsperonestructure0.1–0.2Å7.452.122.70.2–0.5Å2.650.080.50>0.5Å2.220.000.00RMSDofcoordinatesofatomsC',Ca,N,andOinsuperpositionoffamilystructures,ÅRMSD1–162.31±0.553.33±0.933.23±0.73RMSD1–61.98±0.362.05±0.692.25±0.41RMSD7–161.66±0.361.87±0.372.04±0.40inaqueousenvironmentallowsustosimulatefoldingmenthavebeenincreasedinmorethanfourtimesofsmallpolypeptidesabinitiousingtheexistingcomcomparedwiththecalculationinavacuum.putationalcapabilities[40].ItisbelievedthattheuseCalculationofthestructureinanaqueousenvironofevenarelativelysmallnumberofexperimentalmentleadstoanincreaseinthenumberofviolationsrestraintscansignificantlyspeedupsuchsimulationofexperimentalrestraintsoninternucleardistancesandmakethemmorerealistic.Thegivenstudycon(Table3).However,thismaybeduetothefactthatfirmsthishypothesis.optimizationofthelistofexperimentalrestraintswasTheresultsobtainedindicatethatamorerealisticcarriedoutinthecalculationofthepeptidestructureforcefield,includingelectrostaticpotential,andinavacuum.Then,thelistwasusedwithoutanymodmoleculardynamicsinaqueousenvironment,canificationforstructureoptimizationinathinlayerofsubstantiallyreducetheRMSDofflexiblefragmentsolventandcalculationintheaqueousenvironment.coordinatesofpeptide,whilewecannotdetermineaOnecaneffectivelyperformadjustmentofthespatialsufficientnumberofNMRrestraintsforthem.restraintsthatarederivedfromthespectraldataandcontainerrorsduetosucheffectsasspindiffusionandHowever,theintroductionofthechargepotentialoverlapping,usingtheresultsofcalculationinanandaqueousenvironmentintroducesadditionalcomaqueousenvironment.Additionalstructuringcapabilplexityintosearchforoptimalstructureusingagivenityprovidedbyarealisticforcefieldcansimplifythissetofrestraints.Thisisprobablyduetosolvationoftheprocess.initialextendedproteinchainandanadditionalstabilizationoftheintermediatestates.Thus,tofindanThedevelopedprotocolcanbeusedforthecalcueffectiveminimumenergyconformation,weshouldlationofshortpeptidestructurewithalimitednumberuseahighinitialdynamictemperatureandperformofexperimentalrestraints.Itsmajoradvantageisaslowcoolingofthesystem.Therefore,thedurationofhighqualityofobtainedstructuresdefinedbyparamethetrajectoryinMDcalculationinaqueousenvirontersofRamachandranmaps.ThisapproachcanalsoMOLECULARBIOLOGYVol.44No.62010 966ISTRATEetal.Phe42.GrigorenkoA.P.,RogaevE.I.2007.Molecularbasisof(a)Phe10Arg13(b)Alzheimer’sdisease.Mol.Biol.(Moscow)41,294–307.3.GlennerG.G.,WongC.W.1984.Alzheimer’sdiseaseGlu11andDown’ssyndrome:Sharingofauniquecerebrovascularamyloidfibrilprotein.Biochem.Biophys.Res.Commun.122,1131–1135.His6His144.MastersC.L.,SimmsG.,WeinmanN.A.,MulthaupG.,Ser7McDonaldB.L.,BeyreutherK.1985.AmyloidplaquecoreproteininAlzheimerdiseaseandDownsyndrome.Phe4(c)Arg13(d)Proc.Natl.Acad.Sci.USA.82,4245–4249.Phe10Glu35.HaassC.,SchlossmacherM.G.,HungA.Y.,VigoPelGlu11freyC.,MellonA.,OstaszewskiB.L.,LieberburgI.,KooE.H.,SchenkD.,TeplowD.B.,etal.1992.AmyloidbetapeptideisproducedbyculturedcellsduringHis6normalmetabolism.Nature.359,322–325.His146.ToropyginI.Y.,KugaevskayaE.V.,MirgorodskayaO.A.,ElisseevaY.E.,KozminY.P.,PopovI.A.,NikolaevE.N.,Phe10MakarovA.A.,KozinS.A.2008.TheNdomainofangio(e)Phe4Arg13(f)tensinconvertingenzymespecificallyhydrolyzestheArg5His14His6bondofAlzheimer’sAbeta(1–16)peptideanditsisoAsp7analoguewithdifferentefficiencyasevidencedbyGlu3Glu11quantitativematrixassistedlaserdesorption/ionizationtimeofflightmassspectrometry.RapidCommun.MassSpectrom.22,231–239.His67.MayeuxR.,TangM.X.,JacobsD.M.,ManlyJ.,BellK.,MerchantC.,SmallS.A.,SternY.,WisniewskiH.M.,MehtaP.D.1999.Plasmaamyloidbetapeptide1–42Fig.6.SuperpositionofNterminaldomain(1–6a.a.;a,andincipientAlzheimer’sdisease.RapidCommun.Massc,e)andCterminaldomain(7–16a.a.;b,d,f)familiesofSpectrom.22,412–416.(a,b)40structuresofRatAβ(1–16)peptidecalculatedinGromacs;(c,d)inCNSinavacuum;in(e,f)CNSwith8.SeubertP.,VigoPelfreyC.,EschF.,LeeM.,DoveyH.,optimizationinathinlayerofsolvent.ThesuperpositionisDavisD.,SinhaS.,etal.1992.IsolationandquantifimadeofC',Cα,andNatomsand(a,c,e)fragments1–6cationofsolubleAlzheimer’sbetapeptidefrombiologofthepolypeptidechainand(b,d,f)fragments7–16.icalfluids.Nature.359,325–327.Atomsofthemainproteinchainareshowninblack,and9.HardyJ.,SelkoeD.J.2002.Theamyloidhypothesisofthesideatomsareshowningray.SeveralaminoacidresiduesofRatAβAlzheimer’sdisease:Progressandproblemsontheroad(1–16)arelabeled.totherapeutics.Science.297,353–356.10.LuhrsT.,RitterC.,AdrianM.,RiekLoherD.,BohrbeusefulforoptimizationoflargerproteinstructuremannB.,DobeliH.,SchubertD.,RiekR.2005.3Dcontainingunstructuredfragments,toobtainthemoststructureofAlzheimer’samyloidbeta(1–42)fibrils.reliabledataonthedynamicensembleofconformaProc.Natl.Acad.Sci.USA.102,17342–17347.tionsinsolution.11.OylerN.A.,TyckoR.2004.AbsolutestructuralconstraintsonamyloidfibrilsfromsolidstateNMRspecCalculationsinGromacssoftwarepackageweretroscopyofpartiallyorientedsamples.J.Am.Chem.performedontheChebyshevSKIFMSUsupercomSoc.126,4478–4479.puter.12.PetkovaA.T.,YauW.M.,TyckoR.2006.ExperimentalconstraintsonquaternarystructureinAlzheimer’sbetaamyloidfibrils.Biochemistry.45,498–512.ACKNOWLEDGMENTS13.GuilloreauL.,DamianL.,CoppelY.,MazarguilH.,ThestudywasfinanciallysupportedbystateconWinterhalterM.,FallerP.2006.Structuralandthermotracts(02.512.11.2334and02.740.11.0776),andynamicalpropertiesofCuIIamyloidbeta16/28comICGEBgrant(CRP/RUS0802),program“MolecuplexesassociatedwithAlzheimer’sdisease.J.Biol.Inorg.Chem.11,1024–1038.larandCellBiology”ofRussianAcademyofSciences,andagrantoftheRussianFoundationforBasic14.KozinS.A.,ZirahS.,RebuffatS.,HoaG.H.,DebeyP.Research(080400582).2001.ZincbindingtoAlzheimer’sAbeta(116)peptideresultsinstablesolublecomplex.Biochem.Biophys.Res.Commun.285,959–964.REFERENCES15.MekmoucheY.,CoppelY.,HochgrafeK.,GuilloreauL.,TalmardC.,MazarguilH.,FallerP.2005.Character1.AguzziA.,O’ConnorT.2010.ProteinaggregationdisizationoftheZnIIbindingtothepeptideamyloideases:Pathogenicityandtherapeuticperspectives.beta1–16linkedtoAlzheimer’sdisease.Chembiochem.NatureRev.DrugDiscov.9,237–248.6,1663–1671.MOLECULARBIOLOGYVol.44No.62010 OPTIMIZATIONOFTHEMETHODSFORSMALLPEPTIDESOLUTIONSTRUCTURE96716.ZirahS.,StefanescuR.,ManeaM.,TianX.,CecalR.,forcefieldforNMRstructurecalculation.J.Biomol.KozinS.A.,DebeyP.,RebuffatS.,PrzybylskiM.2004.NMR.13,51–59.Zincbindingagonisteffectontherecognitionofthebeta28.RiepingW.,HabeckM.,BardiauxB.,BernardA.,Malamyloid(4–10)epitopebyantibetaamyloidantibodies.liavinT.E.,NilgesM.2007.ARIA2:AutomatedNOEBiochem.Biophys.Res.Commun.321,324–328.assignmentanddataintegrationinNMRstructurecal17.TsvetkovF.O.,MakarovA.A.,ArchakovA.I.,culation.Bioinformatics.23,381–382.KozinS.A.2009.Effectofisomerizationofaspartate729.GuntertP.2004.AutomatedNMRstructurecalculaonthebindingofcopper(II)ionbythebetaamyloidtionwithCYANA.MethodsMol.Biol.278,353–378.peptide.Biofizika.54,197–201.30.BardiauxB.,MalliavinT.E.,NilgesM.,MazurA.K.18.TsvetkovP.O.,PopovI.A.,NikolaevE.N.,2006.ComparisonofdifferenttorsionangleapproachesArchakovA.I.,MakarovA.A.,KozinS.A.2008.forNMRstructuredetermination.J.Biomol.NMR.34,IsomerizationoftheAsp7residueresultsinzinc153–166.inducedoligomerizationofAlzheimer’sdiseaseamy31.HornakV.,AbelR.,OkurA.,StrockbineB.,RoitbergA.,loidbeta(1–16)peptide.Chembiochem.9,1564–1567.SimmerlingC.2006.ComparisonofmultipleAmber19.LovellM.A.,RobertsonJ.D.,TeesdaleW.J.,CampbellJ.L.,forcefieldsanddevelopmentofimprovedproteinbackMarkesberyW.R.1998.Copper,ironandzincinAlzheimer’sboneparameters.Proteins.65,712–725.diseasesenileplaques.J.Neurol.Sci.158,47–52.32.LingeJ.P.,WilliamsM.A.,SpronkC.A.,BonvinA.M.,20.MillerY.,MaB.,NussinovR.2010.ZincionspromoteNilgesM.2003.RefinementofproteinstructuresinAlzheimerAbetaaggregationviapopulationshiftofexplicitsolvent.Proteins.50,496–506.polymorphicstates.Proc.Natl.Acad.Sci.USA.107,33.vanderSpoelD.,LindahlE.,HessB.,GroenhofG.,9490–9495.MarkA.E.,BerendsenH.J.2005.GROMACS:Fast,flexible,andfree.J.Comput.Chem.26,1701–1718.21.ShiversB.D.,HilbichC.,MulthaupG.,SalbaumM.,BeyreutherK.,SeeburgP.H.1988.Alzheimer’sdisease34.WuthrichK.1986.NMRofProteinsandNucleicAcids.amyloidogenicglycoprotein:ExpressionpatterninratNY:Wiley.brainsuggestsaroleincellcontact.EMBOJ.7,1365–35.OhD.,ShinS.Y.,KangJ.H.,HahmK.S.,KimK.L.,1370.KimY.1999.NMRstructuralcharacterizationofcecropinA(1–8)magainin2(1–12)andcecropinA22.BilleterM.,WagnerG.,WuthrichK.2008.Solution(1–8)melittin(1–12)hybridpeptides.J.Pept.Res.53,NMRstructuredeterminationofproteinsrevisited.578–589.J.Biomol.NMR.42,155–158.36.HessB.,vanderVegtN.F.2006.Hydrationthermody23.CloreG.M.,BrungerA.T.,KarplusM.,Gronennamicpropertiesofaminoacidanalogues:AsystematicbornA.M.1986.Applicationofmoleculardynamicscomparisonofbiomolecularforcefieldsandwaterwithinterprotondistancerestraintstothreedimenmodels.J.Phys.Chem.B.110,1761617626.sionalproteinstructuredetermination.Amodelstudyofcrambin.J.Mol.Biol.191,523–551.37.ShenT.,WongC.F.,McCammonJ.A.2001.AtomisticBrowniandynamicssimulationofpeptidephosphory24.BrüngerA.T.1992.XPLOR:ASystemforXrayCrystallation.J.Am.Chem.Soc.123,9107–9111.lographyandNMR,17thed.NewHaven,Connecticut:38.LaskowskiR.A.,RullmannnJ.A.,MacArthurM.W.,YaleUniv.Press.KapteinR.,ThorntonJ.M.1996.AQUAand25.BrungerA.T.,AdamsP.D.,CloreG.M.,DeLanoW.L.,PROCHECKNMR:ProgramsforcheckingthequalGrosP.,GrosseKunstleveR.W.,JiangJ.S.,KuszewskiJ.,ityofproteinstructuressolvedbyNMR.J.Biomol.NilgesM.,PannuN.S.,ReadR.J.,RiceL.M.,SimonNMR.8,477–486.sonT.,WarrenG.L.1998.Crystallography&NMRsys39.KuszewskiJ.,GronenbornA.M.,CloreG.M.1997.tem:AnewsoftwaresuiteformacromolecularstructureImprovementsandextensionsintheconformationaldetermination.ActaCrystallogr.DBiol.Crystallogr.54,databasepotentialfortherefinementofNMRand905–921.Xraystructuresofproteinsandnucleicacids.J.Magn.26.LingeJ.P.,HabeckM.,RiepingW.,NilgesM.2003.Reson.125,171–177.ARIA:AutomatedNOEassignmentandNMRstruc40.VoelzV.A.,BowmanG.R.,BeauchampK.,PandeV.S.turecalculation.Bioinformatics.19,315–316.2010.Molecularsimulationofabinitioproteinfolding27.LingeJ.P.,NilgesM.1999.InfluenceofnonbondedforamillisecondfolderNTL9(1–39).J.Am.Chem.Soc.parametersonthequalityofNMRstructures:Anew132,1526–1528.MOLECULARBIOLOGYVol.44No.62010