Identification of geographical origins of Panax notoginseng ...

Identification of geographical origins of Panax notoginseng based on HPLC multi-wavelength fusion profiling combined with average linear ... Skiptomaincontent Thankyouforvisitingnature.com.YouareusingabrowserversionwithlimitedsupportforCSS.Toobtain thebestexperience,werecommendyouuseamoreuptodatebrowser(orturnoffcompatibilitymodein InternetExplorer).Inthemeantime,toensurecontinuedsupport,wearedisplayingthesitewithoutstyles andJavaScript. Advertisement nature scientificreports articles article IdentificationofgeographicaloriginsofPanaxnotoginsengbasedonHPLCmulti-wavelengthfusionprofilingcombinedwithaveragelinearquantitativefingerprintmethod DownloadPDF DownloadPDF Subjects ChemistryDrugdiscovery AbstractTheaimofthisstudywastoestablishamethodforgeographicaloriginsidentificationofPanaxnotoginseng(P.notoginseng)basedonabundantchromatographicspectralinformation.CharacteristicfingerprintsofP.notoginsengextractssamplesweregeneratedbyMulti-wavelengthFusionProfiling(MWFP)methodbasedontheHPLCfingerprintsestablishedatthreewavelengthsof203 nm,270 nmand325 nm.Thesamplesgroupingresultscalculatedwiththeaveragelylinearquantifiedfingerprintmethod(ALQFM)andtheunsupervisedstatisticalmethodsbasedonfusionfingerprintsmatcheswiththegeographicalorigins.TheMulti-wavelengthFusionProfiling(MWFP)methodhasbeensuccessfullyappliedtoidentificationofgeographicaloriginsofP.notoginsengandshowstheadvantagescomparedwithsingle—channelfingerprints.Inaddition,eightphysiologicallyactivecomponents,includingfoursaponins,twoflavonesandtwoaminoacids,wereidentifiedfromthemostrelevantingredientsofP.notoginsenggeographicaloriginsbyfusionfingerprint-efficacyrelationshipanalysis.Besidestherecognizedactivesaponins,othercategoriesofactiveingredientssuchasflavonoidsandaminoacidsshouldbepaidattentiontointheproducingareasidentificationorthequalityjudgmentofP.notoginseng. IntroductionPanaxnotoginsengnamed“San-qi”and“Tian-qi”isthedriedrootsandrhizomesofPanaxnotogitiseng(Burk.)F.H.ChenlistedintheChinesepharmacopoeia.ItisavirtualChinesemedicine(TCM)withanannualoutputvalueexceeded70billionRMByuanandmorethan3600relateddrugapprovalnumbersinChina1.Itsmainefficacyispromotingbloodcirculation,removingbloodstasis,detumescenceandanalgesia2,3,4. ThegenuineproducingareaofP.notoginsengwasinYunnanprovince,China5.Duetothegreatdemandcannotbemet,theproductionofP.notoginsengexpandedtotheadjacentregion.SuiningCity,SichuanProvincewasthefirsttointroducesuccessfully,andtheonlyproductionareaoutsideYunnanwithstableoutputforthreeconsecutiveyearsupto2019.Thedifferenceofproducingareainvolvesthedistinctionincomplicatenaturalandartificialfactors,whichwillleadtothediversityoftheintegralcompositionandeventuallytheefficacyofTraditionalChineseMedicine6,7.Intermsofgeographicalconditions,theP.notoginsengorigininYunnanissignificantlydifferentfromthatinSichuan.Theformerisplateaumountainareaandthelatterishillsinthehinterlandofthebasin.Thenaturalenvironmentandthegrowingtechniquesarebothvariousinthetwoprovincesproducingarea.Therefore,itisnecessarytodistinguishP.notoginsengproducedinthetwoprovincesbyorigindiscriminanttechnique.BasingonthecurrentlegalqualityindicatorsofP.notoginsengregulatedinPharmacopoeiaofChina,theindividualindicatorsofGinsenosideRb1,GinsenosideRg1andNotoginsenosideR1content,theidentificationofgeographicaloriginsisdifficulttoachieve.ThereweresomestudiesoftheidentificationofgeographicaloriginsofTCM8,9includingP.notoginseng10basedonstableisotoperatios.Exceptthestrategyrelatedtotheinorganicstableisotopetheoriginsidentifiedstrategybasedoncomplexorganiccompositioninformationhavebeenworkedout.P.notoginsenggeographicaldiscriminationmodelshavebeenestablishedbasedonspectraldataandvariousdataminingalgorithms.InWang’sstudy11,themodelisestablishedbypartialleastsquaresdiscriminantanalysis(PLS-DA)oftheoptimizedUVspectradataofP.notoginsengsamples.TheFouriertransform-infraredspectrum12ornear-infraredspectrumdataofP.notoginsengareusedtodistinguishtheproducingarea13.Inordertoexploitthesynergeticandcomplementaryinformation,researchersestablishedmodelsoforigindiscriminationbasedonhigh-levelfusionofFouriertransformmid-infraredspectroscopyandnearinfraredspectroscopycombinedwiththealgorithmofrandomforest14,15.Differentfromthemethodsbasedonspectraldatamatrixmentionedabove,thediscriminationmodelbasedonnear-infraredspectraimageanddeeplearningstrategyisalsoappliedtoP.notoginseng16.Thesediscriminationmethodsbasedonspectralanalysishaveanadvantageinapplyingthespectralinformationinacertainwavelengthrangeofsamplescomprehensively.However,theinformationofspectralwithoutseparationiscomplexandcouldonlybeusedfordiscriminantanalysisaftercalculationandtransformation.Atthesametime,itisdifficulttodirectlyassociatetheinformationofcomponentswiththespectralandestablisheddiscriminantmodel.Atpresent,researchersdiscussedSpectralandChromatographicOverallAnalysistechnique(SCOA),aimingtoextract,integrateandtransformtheinformationfromcomplexmulti-channelintoanalyzableforms17.Multi-wavelengthFusionProfiling(MWFP)isarepresentativeoftheSCOA.ItisobtainedbytheprojectionalongthewavelengthaxisoftheHPLCchromatogramscollectedfromthesamesampleanalyzedwithmultiplewavelengths(orDAD).ItischromatogrammaximizestheavailableMulti-wavelengthchromatographicinformation.MWFPcouldprovidemoreinformationfororiginidentificationthansinglewavelengthHPLCfingerprint,andthemoreaccurateconclusionmaybeobtainedwithMWFPanalysis.Comparedtothespectralwithoutseparation,chromatographicseparationreducesthecomplexityofthespectrum,andtherelationshipbetweenchromatographicpeaksandchemicalcompositionsismoredirect.Inrecentstudies,MWFPmethodwasappliedforthequalityassessmentofthesingleherbs18andproprietaryChinesemedicinewithmorecomplexcomponentsconsistingofavarietyofmedicines19.ThesestudiesbasedontheMWFPmethodprovidenovelandcomprehensivestrategyforTCMqualityassessment.Inthispresentstudy,aMulti-wavelengthFusionProfiling(MWFP)methodisdevelopedtodistinguishtheP.notoginsengsamplesfromtwoproducingareawithvariousgeographicalconditions.MaterialsandmethodsChemicalandmaterialsAcetonitrileandmethanol(HPLC-grade)wereobtainedfromSigma-Aldrich(Steinheim,Germen).Formicacid(HPLC-grade)wassuppliedfromthelocaldealer.Ultra-purewaterwasmadebythepurewatermachine(Millipore,France,18.2 MΩcm@25 °C).ReferencestandardsofGinsenosideRb1(G-Rb1),GinsenosideRg1(G-Rg1),GinsenosideRd(G-Rd),NotoginsenosideR1(NG-R1),NotoginsenosideR2(NG-R2)wereacquiredfromChengduGlipBiotechnologyCo.,Ltd.,Chengdu,Sichuan.ThestructuresofthefivesaponinswerepresentedinFig. 1.Figure1Chemicalstructuresoffivereferencestandards.FullsizeimageSamplesourcesTheP.notoginsengsamplesS1–S15werecollectedinSuiningCity,SichuanProvince;S16–S19andS23–S26werecollectedinWenshanZhuangandMiaoAutonomousPrefecture,YunnanProvince,S20–S22werecollectedinZhaotong,YunnanProvince,andS27–S32werecollectedinHani-YiAutonomousPrefectureofHonghe,YunnanProvince,whicharethemainproducingareasofP.notoginseng.SamplepreparationandHPLCanalysisTheP.notoginsengsamplesweresmashedbythepulverizer.About0.500 gsmashedsamplewasaccuratelyweighedintoa50 mLcentrifugaltube.20 mLmethanol/water(60:40,v/v)wasaddedinandthecentrifugaltubewasleaveinanultrasonicwaterbathtoextractionfor30 min.Afterwardthetubewasputintoa60 °Cwaterbathfor2 handcentrifugedat3000 rpmfor5 min.Thesupernatantwasfilteredthrougha0.22 μmfiltermembraneandcollectedasthesamplesolution.Thestandardssolutionwaspreparedbyaccuratelyweighedamountof1.0 mganddissolvedwithmethanolinthe10 mLvolumetricflask.Allsolutionswerestoredat4 °Cuntilanalysis.ChromatographicseparationwasoperatewithShimadzuLC-20ADequippedwithaPoroshell120EC-C18column(100 × 4.6 mm,2.7 μm).Themobilephaseconsistsofacetonitrile(A)andwater(B).Separationwasachievedusingthefollowinglineargradientprogram:0–10 min,95%B–80%B;10–15 min,80%B–60%B;15–25 min,60%B;25–50 min,60%B–0%B;50–60 min,0%B;60–60.1 min,0%–95%B,70 min,stop.Thecolumntemperaturewasmaintainedat40 °C.A10 μLaliquotofeachsamplewasinjectedintotheHPLC–UVsystem.Theflowratewassetat0.3 mL/min.Thedetectionwavelengthsweresetat203 nm,270 nmand325 nm.Thefivesaponinsconcentrationofthesamplescalculatedwiththechromatographyof203 nm.SimilarityanalysiswithALQFMThefusionfingerprintofeachP.notoginsengsampleandtheirsimilaritywerecalculatedwiththesoftware“DigitizedEvaluationSystemforsuper-informationCharacteristicsofTCMChromatographicFingerprints4.0”(developedbyGuoxiangSunetal.,SoftwarecertificatedNO.0407573,China).TheAverageLinearQuantitativeFingerprintmethod(ALQFM)wasadoptedwiththesoftwaretoanalyzethesimilarity20,21.Therewerethreeparametersinvolved.Averagelinearqualitativesimilarity(Sm),averagelinearquantitativesimilarity(Pm)andthefingerprintvariationcoefficient(α).Thesimplifiedparameter“Grade”wascombinedwithSm,Pmandαtoevaluatethesimilarityofsamples(Table1).Generally,sampleswiththegradevalue≤ 5wererecommendedasthesimilaronestothechosenreferencesample.Table1TCMsimilaritygradescriteriabasedonALQFM.FullsizetableStatisticalanalysisHierarchicalclusteringanalysis(HCA)wasperformedaccordingtothecontentsofthefivesaponinsconcentrationofthe32batchP.notoginsengsamplestoshowtherelatednessofthesamplesbasedonthecompositionindexescontent.Heatmapclusteringandtheprincipalcomponentanalysis(PCA)wereperformedaccordingtothepeakareasof33co-possessingpeaksofthefusionfingerprintswiththesamplestoshowtherelatednessofsamplesbasedonthefusionfingerprints.Theco-possessingpeaks’areasofthefusionfingerprintswiththeP.notoginsengsamplesgroupedaccordingtotheirorigin(YunnanprovinceorSichuanprovince)wereevaluatedbyOrthogonalProjectiontoLatentStructureDiscriminantAnalysis(OPLS-DA)toverifygroupingandtofindthepeaksthatcanrepresentthedifferencebetweenthetwoorigins.HCAwasperformedwiththeSPSSstatisticalsystem(version13.0)usingthebetweengrouplinkagemethodwithsquaredEuclideandistances.HeatmapclusteringwereperformedinRProgrammingLanguageandthesimilaritywasmeasuredwiththebetween-groupslinkagemethodandsquaredEuclidean.PCAandOPLS-DAwereperformedusingSIMCA14.1software.ResultsanddiscussionMethodologyevaluationofHPLCanalysisThecalibrationcurvesoffivesaponinsstandardsolutions,aswellasthelinearrangesandthelimitofdetection(LOD)wereshowedinSupportingInformationTableS1.Allthestandardproductsshowedexcellentlinearity(R2 ≥ 0.999)overthetestedconcentrationranges.Thequantitativeaccuracywasassessedwiththestandardadditionmethod.Theaveragerecoveriesforthefiveinvestigatedcompoundsrangedfrom91.84to105.38%.ThefingerprintstabilitywasevaluatedwithRSDvaluesofeachco-possessingfingerprintpeakareawere,respectively,lessthan3.22%,3.71%and2.16%forthestability,precisionandrepeatabilitytests.Consideringtheseresults,themethodwasaccurateandvalidenough.CompositionindexescontentanalysisAccordingtopreviousstudies,saponinsaretheimportantfunctionalcomponentsofP.notoginseng,andareoftenusedasitsindicatorcomponents.ThesumofG-Rb1,G-Rg1andNG-R1contenthasbeenusedasamarkerforthequalitycontrolofP.notoginsenginChinesePharmacopoeia.InthisstudyfivesaponinsaboundantinP.notoginsengreportedinourpreviousstudy22arechosenascompositionindexes.Thecontentoffivesaponinscontentsin32P.notoginsengsampleswereshowedinTable2.Table2Theconcentrationofthefiveinvestigatedcompoundsinthesamples(mg/g).FullsizetableTheHCAplotbasedonthefivesaponinscontentsisshowninFig.S1.Thefailureoforigindiscriminationbeginswiththeminimumbetween-groupsdistance.TheresultverifiedthatitisimpossibletodistinguishtheoriginofP.notoginsenginYunnanProvince(thegenuineproducingareainplateaumountainarea)andSichuanProvince(thenewintroducedproducingareainhillsinthehinterlandofthebasin)bytheindicatorscontent.HPLCfingerprintsimilarityanalysiswithALQFMTheUVspectraofthefivesaponinswereinvestigated.TheUVspectraofsamplesolutionisdifferentwiththatofthefivesaponinsin203–350 nm(showninFig. 2).InordertofullyreflectthecharacteristicsoftheP.notoginseng,threeabsorptionbands(203 nm,270 nmand325 nm)werechosentoestablishthefusionfingerprint,whichcapableofsynthesizingenhancingsignalresponseandrichfingerprintinformation.Figure2UVspectrumoffiveinvestigatedsaponinsreferencestandardsandP.notoginsengextract.G-Rg1GinsenosideRg1,G-RdGinsenosideRd,G-Rb1GinsenosideRb1,NG-R2NotoginsenosideR2,NG-R1NotoginsenosideR1.FullsizeimageThefingerprintsof32samplesolutionwiththethreeabsorptionbands(203 nm,270 nmand325 nm)wereaccomplished,respectivelyatfirst(showninSupportingInformationFig.S2).Thenthefusionfingerprintofeachsample(showninFig. 3)andthesimilarityevaluationresults(presentedinTable3)werecalculated.Figure3Fusionfingerprintof32batchesofP.notoginsengsamples.S1sample1,S32sample32.FullsizeimageTable3TheevaluationresultsobtainedfromALQFMof32batchesofP.notoginsengsamples.FullsizetableInALQFM,qualitativeanalysisisperformedfirst,thenquantitativeanalysisisused.AsTable3showedthatallthesampleshadα  0.90,indicatingthatthesamplesweresimilarinthedistributionandnumberofchemicalcompositions.BasedonTable1,allthesamples(α  0.90)shouldbeingrade1–5.Becauseofthedifferencecontentinthesamples,someofthesamplesPm  130%whichmadetheirgradesbetween6and8.IfconsideronlyqualitativesimilaritySm,origindistinguishisnotshowedinanywavelengthfingerprintsnorinthefusionfingerprints.Asfurtherelaboratedinfollows,theadditionofPmmakesthe“Grade”havethefunctionoforigindiscrimination,andthefusionfingerprintsreflectsanadvantageinthisrespect.CorrelationbetweenGradeandorigins“Grade”isasimplifiedparameterforevaluatingqualitativeandquantitativesimilarityofsamplesofALQFMmethod.Thegradesdistributionofsamplesfromdifferentproducingareawerereflectedwiththescatterplots(Fig.S3).AsshowninFig.S3A–Cthesamplesofthetwoproducingareascannotdistinguishwithgradesdistributionatanysinglewavelength,whileasshowninFig.S3Dmostofthesamples(12of15)fromSichuanProvincehadagradebetween1and4,andmostofthesamples(15of17)fromYunnanProvincehadagradebetween5and8calculatedwithfusionfingerprints.ItcanbeseenthatthefusionfingerprintisbetterthanthesinglechannelfingerprintintherespectoftheproducingareasdistinguishcapacityofSichuanandYunnanwithGrade.Thiscanbeinterpretedasthefusionfingerprintscanreflectthesamplecharacteristicsmorecompletelythanthesingle-channelfingerprintsbecauseofthecomprehensiveutilizationofthechromatographicinformationofthethreebandsofultravioletspectrum.TheerrorfrequencyofusingGradetodistinguishtheoriginsisabout5of32.RelationshipbetweenfusionfingerprintsandoriginsTheHCAheat-map(Fig. 4)showstherelatednessof32P.notoginsengsamplesbasedonthefusionfingerprints.ExceptS8,thesamplesareclusteredintotwogroups,onegroupisS1–S15andanothergroupisS16–S32.Theresultisconsistentwiththeorigindistinguish(S1–S15werecollectedfromSichuanprovinceandS16–S32werefromYunnanprovince).ThePCAscoreplots(Fig. 5A)showedtheclassifyresultconsistentwiththatofHCAheat-map.Figure4Hierarchicalclusteringanalysisof32P.notoginsengsamplesfromtwoproducingarea.FullsizeimageFigure5Principalcomponentanalysisscoreplots(A);orthogonalprojectiontolatentstructurediscriminantanalysisscoreplots(B);andVIPplots(C)forP.notoginsengsamples.FullsizeimageThatistosay,whiletheconventionalunsupervisedsimilarityevaluationanalysisofthefusionfingerprintsusedtodistinguishtheoriginsbetweenthetwoprovincestheerrorfrequencyratewouldreduceto1of32.ToanalyzethecharacteristiccompoundsmostreflectedtothedifferenceoftheP.notoginsengsamplesfromthetwoprovinces,OrthogonalProjectiontoLatentStructureDiscriminantAnalysis(OPLS-DA)wereusedtoanalyzethepeakswhichcontributedtoclassificationofthefusionfingerprints(Fig. 5B).Variableinfluenceonprojectionstatistics(VIP)valueswerecalculatedtoreflectthecontributionofeachvariabletotheestablishedmodel.The15peakswithVIPvalue> 1inFig. 5Cwereconsideredtoberelevantfortheclassificationsignificantly.QualitativeanalysisofpeaksrelevantfortheoriginclassificationHighperformanceliquidchromatographycombinedwithhighresolutionmassspectrometry(OrbitrapLC–MS,ThermoScientific)wasinvolvedtoanalysisthesample.DatawereprocessedwiththesoftwareExactive2.8withdatabasesearching(MZVault,MassList,MZCloud).TheresultswereshowninTable4.Table4Qualitativeanalysisofpeaksrelevantforthegeographicaloriginclassification.FullsizetableInTable4ginsenosideRg1,ginsenosideRb1,ginsenosideRdaretheselectedcompoundstoinvestigatetheconcentrationinthisstudy,andthesignificantdifferenceoftheircontentintheP.notoginsengsamplesfromtwoorigins(P