Simultaneous authentication of species identity and geographical origin of shrimps: Untargeted metabolomics to recurrent biomarker ions

Mandatory disclosure of the species identity, production method, and geographical origin are embedded in the regulations and traceability systems, governing international seafood trade. A high-resolution mass spectrometry-based metabolomics approach could simultaneously authenticate the species identity and geographical origin of commercially important shrimps.

jou rn al h om ep a g e : w w w e l s e v i e r c o m / l o c a t e / c h r o m a

a Institute for Global Food Security, School of Biological Sciences, Queen’s University Belfast, United Kingdom

b ICAR-Central Institute of Fisheries Technology, Cochin, India

c Mass Spectrometry Core Technology Unit, Queen’s University Belfast, United Kingdom

d School of Pharmacy, Queen’s University Belfast, United Kingdom

e University of Chemistry and Technology, Department of Food Analysis and Nutrition, Prague, Czech Republic

a r t i c l e i n f o

Article history:

Received 2 November 2018

Received in revised form 18 February 2019

Accepted 1 April 2019

Available online 3 April 2019

Keywords:

Untargeted metabolomics

Chemometrics

Biomarker identification

Shrimp fraud

Species authentication

a b s t r a c t

Mandatorydisclosureofthespeciesidentity,productionmethod,andgeographicaloriginare embed-dedintheregulationsandtraceabilitysystems,governinginternationalseafoodtrade.Ahigh-resolution massspectrometry-basedmetabolomicsapproachcouldsimultaneouslyauthenticatethespecies iden-tityandgeographicaloriginofcommerciallyimportantshrimps.Thehighlyinnovativeapproachspared theneedformultipletestingmethodswhichareinroutineusecurrently.Arobustchemometricmodel, developedusingthemetabolitefingerprintdataset,couldaccuratelypredictthespeciesidentityofthe shrimpsamples.Subsequently,species-specificbiomarkerswerediscoveredandatandemmass spec-trometrymethodforauthenticationofthespecieswasdeveloped.Twootherchemometricmodelsfrom themetabolomicsexperimentaccuratelypredictedthegeographicaloriginofkingprawnsandtiger prawns.Thestudyhasshownforthefirsttimethatfood-metabolomicsalongwithchemometricscan simultaneouslycheckformultipleseafoodfraudissuesintheglobalseafoodsupply-chain

©2019TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense

(http://creativecommons.org/licenses/by/4.0/)

1 Introduction

Fisheriesandaquacultureproductsareamajorsourceof

liveli-hoodandsustenanceforbillionsofpeopleglobally.TheFoodand

AgricultureOrganisationoftheUnitedNations(FAO)hasestimated

theannual exporttradeof seafoodalone toUS$150billion[1

Unfortunately,seafoodisalsooneofthemostprominentfood

cat-egoriesassociatedwithfoodfraud,underminingthecredibilityof

thewholeseafoodsupplychain[2,3 Seafoodfraudnotonly

threat-enshealthandsafetyofconsumers,alsoputsouroceansandother

waterresourcesatrisk[4 Shrimpsandprawnsareconsideredas

avaluableseafoodproductthataccountsforabout15%ofthetotal

valueofinternationallytradedfisheryandaquacultureproducts

amountingtoaboutUS$43billion[5 Scandals,suchasfraudulent

∗ Corresponding author at: ICAR-Central Institute of Fisheries Technology,

Cochin-28, Kerala, India.

E-mail addresses: niladri icar@hotmail.com , Niladri.Chatterjee@icar.gov.in

(N.S Chatterjee).

labelingoflowvaluespeciesaspremiumspeciesofshrimpsand labelingaquaculturewhitelegshrimp(Litopenaeusvannamei)as premiumwildcaughtshrimphavebeenreportedbyinternational environmentalnonprofit groups[6 Disturbingreports ofslave labourin Thailandshrimpindustry havesurfacedinthemedia, promptingthefoodindustrygiantNestletoinitiatestringent mea-surestopreventhumanrightsabusesintheseafoodsupplychain

inThailand[7 Internationallawsentitleconsumerstoknowthecommercial identityofthespecies,productionmethod(wildcaughtorfarmed) andgeographicaloriginforallcategoriesoffisheryandaquaculture products.Together,thesethreepiecesofinformation constitute the“Traceability”ofaseafoodproduct[8,9 DNAprofilingbased techniquesareconsideredgoldstandard forspecies authentica-tionofseafood.However,successoftheapproachheavilyrelies

ontheavailabilityofcomprehensivereferencesequencelibraries Authenticationofgeographicaloriginandproductionmethodusing DNAbasedtechniquesishighlychallengingandreportsofsuch application are very few [10–12] Similarly proteomics strate-giesaremostlysuitableforauthenticationofspeciesidentityand https://doi.org/10.1016/j.chroma.2019.04.001

0021-9673/© 2019 The Authors Published by Elsevier B.V This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ).

detectingadulterationof ingredients [13–15] Among the

tech-niquesusedtodeterminethegeographicaloriginandproduction

methodofseafood,multi-elementandstableisotoperatioanalysis

approachesarethemostsuccessfullyemployed[5,16].However,

thesedifferentanalyticaltechniquesoftenrequirecomplex

sam-plepreparation,differenttypesofanalyticalplatformsandoften

longassayruntimes.Thisisasignificantdisadvantagewhiletrying

tomanagetraceabilityinafast-moving,complexsupplychainof

perishableseafood

The complete set of metabolites synthesized in a biological

systemconstituteits ‘metabolome’ and is directlylinked toan

organism’sgeneticmake-up,foodintakeandchangesinthe

envi-ronment, it lives in [17] The targeted metabolomics approach

emphasizeson detectionand quantification of a few classes of

compounds,mostly using a unit resolution massspectrometer

Unitresolutionmassspectrometerwithtriplequadrupolemass

analyzer has the advantage of using selective reaction

mon-itoring (SRM) or multiple reaction monitoring (MRM) mode,

detectingthetargetcompoundsattracelevelincomplex

matri-ces.Untargetedmetabolomicswhichintendstostudytheglobal

‘metabolite fingerprint’ of a sample have several advantages

over targetedapproach such ascombining targeted and

untar-geted screening, novel biomarker discovery and retrospective

data analysis Over the last five years, high-resolution mass

spectrometer(HRMS)hasestablisheditselfasthepreferred

ana-lytical choice in untargeted metabolomics research, driven by

increasedaffordability, unsurpassedsensitivityand high

resolu-tion[18–20].HRMSbaseduntargetedmetabolomicscoupledwith

thepowerof chemometricsdataanalysiscanpotentially

inves-tigate multiple authenticity issues within a single experiment

[21,22]

HRMSbasedmetabolitefingerprintingandchemometricshave

beenusedfor authenticationofseafood/fishonlyontwo

previ-ousinstances.Metabolitefingerprintingonatwodimensionalgas

chromatographyandtimeofflightmassspectrometrywasusedto

distinguishtwobivalvespecies[23].ARapidevaporativeionisation

massspectrometry(REIMS)basedmetabolomicsprofileapproach

wasreportedrecentlyforaccurateidentificationofspeciesidentity

offivedifferentwhitefishfillet[24].Toourknowledge,HRMSbased

untargetedmetabolomicsapproachhasnotbeensofarappliedto

testmultiplefoodfraudissuesinasinglemetabolomics

experi-ment

Inthisstudy,weaimtoexplorethefeasibilityof

authenticat-ingspeciesidentity,geographicalorigin,andproductionmethodof

commerciallyimportantshrimpsandprawnsinasingleuntargeted

metabolomicsexperiment

2 Materials and methods

2.1 Samplescollection

Authenticsampleswerecollectedeither directlyfrom

aqua-culture farms linked to the ICAR-Central Institute of Fisheries

Technologyin India or througha number ofsupermarket

sup-plychainsintheUnitedKingdom.Fivecommerciallyimportant

speciesofshrimps;tigerprawn(Penaeus monodon),kingprawn

(Litopenaeusvannamei),Indianwhiteshrimp(Fenneropenaeus

indi-cus),Indianpinkshrimp(Metapenaeusmonoceros)andArgentinian

red shrimp (Pleoticus muelleri) were considered for developing

chemometricmodelsforspeciesauthentication.Thetigerprawn

samplesincludedwildcaughtprawnsfromIndiaand

Madagas-car;andfarmedprawnsfromVietnamandSriLanka.Farmedking

prawnswereobtainedfromIndia,Thailand,Vietnam,and

Hon-duras.Thesamplesweretransportedtothelaboratorywithin24h

ininsulatedpolystyreneboxeswithdryiceandstoredfrozenat

−80◦Cfollowingremovaloftheheadandoutershell.Thesamples

werefreeze-driedimmediatelyafterwardsandstoredinlabelled polypropylenecontainersat−80◦C.

2.2 Samplespreparation Tenindividualshrimpsampleswerepooledtogethertoobtain onerepresentativesampleforaclassofshrimp,labeledbasedon speciesidentityand countryof origin.Likewise,for aparticular classof shrimpthreerepresentative sampleswereobtained.As pre-treatment,thefreeze-driedsampleswerehomogenizedina planetaryballmill(RetschGmbHPM200,Haan,Germany).Then, 0.05g(±1%)ofpulverizedsampleswasweighedoutona Discov-eryDV215CDAnalyticalBalance(OhausEuropeGmbH,Nanikon, Switzerland)into1.5mLmicro centrifugetubes Next,the sam-pleswereextractedwith1mLofaqueousmethanolcontainingone partultra-purewater(18.2M/cm,MerckMillipore,Billerica,USA) andfourpartsLC–MSgradeChromasolvmethanol(Sigma-Aldrich,

StLouis,MO,USA)bymixingat2500rpmwithDVX-2500 Mul-titubeVortexer(VWRInternational,Lutterworth,UK)for10min, followed by sonication for 30minat maximum frequency in a camSonixC1274waterbathsonicator(Camlab,Cambridge,UK)at roomtemperature.Aftercentrifugationat10,000×gfor10minat

4◦CinaMIKRO200Rcentrifuge(HettichUK,Salford,UK),0.8mL

ofthesupernatant wastransferred intoafreshmicrocentrifuge tubeanddried overnightin amiVac QUP-23050-A00(Genevac, Ipswich, UK)centrifugal sample concentrator The dry extracts werethenreconstitutedin0.8mLofultra-purewaterandfiltered througha0.22␮mCostar® celluloseacetateCentrifugeTube Fil-terbycentrifugationat10,000xg,4◦Cfor10min.Filteredextracts weretransferredintoLCvials(Waters,Manchester,UK)forLC–MS analysis

2.3 UntargetedLC-HRMSanalysis Analyseswerecarried outona WatersAcquity UPLCI-Class system(Milford,MA,USA)coupledtoaWatersXevoG2-SQToF massspectrometer(Manchester,UK)withanelectrospray ionisa-tionsourceoperatinginapositiveornegativemodewithlock-spray interfaceforrealtimeaccuratemasscorrection.Instrument set-tingswereasfollow:sourcetemperaturewassetat120◦C,conegas flowat50Lh−1,desolvationtemperatureat450◦C,anddesolvation gasflowat850Lh−1.Thecapillaryvoltagewassetat1.0kVin pos-itivemodeand0.5kVinnegativemode,respectively.Sourceoffset was80(arbitraryunit).Massspectradatawereacquiredin con-tinuummodeusingMSEfunction(lowenergy:4eV;highenergy: rampfrom20to30eV)overtherangem/z100–1200withascan timeof0.1sAlock-masssolutionofLeucineEnkephalin(1ngmL−1)

inmethanol/watercontaining0.1%formicacid(1:1,v/v)was con-tinuouslyinfusedintotheMSviathelock-sprayataflowrateof

10␮Lmin−1.Thechromatographicseparationwasconductedona WatersCortecsT3column(100mm×2.1mm,1.6␮m).Thecolumn oventemperaturewassetat45◦C,injectionvolumeat3.5␮Land flowrateat0.4mLmin−1.Themobilephaseconsistedof(A)water with0.1%formicacidand(B)methanolwith0.1%formicacid.The gradientwassetasfollows:2.0minof99%(A)followedbyalinear increasefrom1%to99%(B)over16min,isocraticcleaningstepat 99%(B)for0.5min,thenreturnedtoinitialconditions99%(A)over 0.1minandcolumnequilibrationstepat99%(A)for1.4min.Each samplewasinjectedthreetimesinordertoassurereproducibility

Atthebeginningoftheexperiment,10pooledconditioningsamples (QCs)wereinjected.Forqualitycontrol,QCswerealsoinjectedat intervalsofevery10samplesthroughouttheentireexperimentto determinethechromatographicreproducibilityofretentiontimes andpeakintensities[25]

2.4 Chemometricdataanalysis

RawdatageneratedwereimportedtoProgenesisQI2.0software

(Waters,Newcastle,UK).Afterdataconversiontotheappropriate

formatusingafiltersetat1.5,datawerealignedtothebestQC

sampleselectedandpeakpickingfrom0.6to18minwascarried

outwith sensitivitysetat absoluteion intensityof 1000

(arbi-traryunit) and chromatographic peak widthto0.08minA data

matrix of detected metabolitefeatures and corresponding

nor-malisedabundancewasgenerated and thenexportedtoSIMCA

14(Umetrics,Malmo,Sweden)formultivariateanalysis.Toassess

thegeneralqualityoftheacquiredspectraldata(univariate/Pareto

scaled)principalcomponentanalysis(PCA)andmodelassessment

were performed Next, data were mean centred, either Pareto

orunivariatescaledandgroupedintorespectiveclassespriorto

orthogonalpartialleastsquare discriminantanalysis(OPLS-DA)

Thevariableimportanceinprojection(VIP)plotsofthemetabolite

featuresassociatedwithOPLS-DA,andsubsequentcross-checking

of the features in Progenesis QIfor peak quality and intensity

ensuredselectionofreliablemetabolitefeatures.R2(cumulative),

Q2(cumulative)valuesandRMSECVwereusedtodeterminethe

validityofthemodels,withR2(cum)employedasanindicatorof

thevariationdescribedbyallcomponentsinthemodelandQ2&

RMSECVasmeasuresofhowaccuratelythemodelcanpredictclass

membership

2.5 BiomarkerdiscoveryandmethodtransfertoLC–MS/MS

The discovery of characteristic biomarkers for each shrimp

specieswasachievedbygeneratingindividualOPLS-DAmodels

duringbinaryspeciescomparison.AssociatedS-plotsandvariable

importanceinprojection(VIP)plotsenabledidentificationofaset

ofmostpromisingionsinbothionisationmodesresponsiblefor

classseparationamongallspecies.Selectedionswerethoroughly

investigatedinboth therawdataandProgenesis QI2.0forthe

peakqualityandintensityaswellasselectivitybetweenassessed

species.Accuratemassofthebiomarkerionswassearchedagainst

themetabolitedatabasesChemSpider,LipidBlast,Metlin,Human

MetabolomeDatabaseandFooDBtorevealputativeidentities

Biomarkers,whichwereselectivetothespeciesidentitywere

furtherinvestigated bytargeted analysis.Theretention time of

biomarkerswasconfirmedonaXevoTQ-SLC–MS/MSinselected

ionmonitoringmodeapplyingthesamechromatographic

condi-tions(describedinSection2.3.)asintheuntargetedanalysis.Next,

thefragmentationspectrum obtainedin daughterion scanwas

comparedtorespectivespectrumacquiredonQ-ToF toconfirm

markerschemicalidentity.Afteroptimisingcollisionenergiesby

repeatedon-columninjections,themostprominentfragment(s)of

eachbiomarker’sprecursorwereselectedforitsrespectiveMRM

window and chromatographic conditions adjusted to decrease

analysistime.Noneoftheseselectedbiomarkersweresuccessfully

identified;nevertheless, theminimal requirementsofreporting

forunknownmetabolites(retentiontime,prominentionand

frag-ment ion) specified by the Chemical Analysis Working Group

within Metabolomics Standards Initiative (MSI) have been

ful-filled[26].Toassurecorrectidentificationuptothree fragment

ionswereincludedinthefinalMSmethod,however,someofthe

selectedmarkersonlyyieldedonefragmention,thusdecreasing

theirreliabilityduetolackofpossibilityofionratiosmonitoring

[27]

2.6 TargetedLC–MS/MSanalysis

TheanalysiswasperformedonanAcquityUHPLCI-Classsystem

(Waters,Milford,MA,USA)coupledtoXevoTQ-Striplequadrupole

massanalyser(Waters,Manchester,UK)operatinginpositive

elec-trospray ionisation mode The following settings wereapplied: capillaryvoltagewassetat1.0kV,thedesolvationandsource tem-peraturesweresetat450and130◦C,respectively,whilenitrogen cone and desolvation flow rates were setto145 and 1000L/h Argonwasemployedasacollisiongas,withaflowof0.15mL/min, yieldingacollisioncellpressureof2.4×10−3mBar.Inter-scanand -channeldelayswerebothsetto3mswhiledwelltimesranged from20to163ms

Analytes’separationwasperformedonaWatersCortecsT3 col-umn(100mm×2.1mm,1.6␮m),maintainedat45◦Cwitha5␮L injectionofa sampleextract.Thepumpwasoperatedataflow rateof0.4mL/minwithmobilephasesconsistingofA,0.1%formic acidinwaterandB,0.1%formicacidinmethanol.Thefinal gradi-entwasisocratic0–1.0min80%A,linear1.0–2.0min75%A,linear 2.0–4.0min20%A,linear4.0–6.5min10%A,linear6.5–7.0min1%

Aforcolumnflush,goingbacktoinitialconditionsduring0.1min andfinishingwithisocraticcolumnequilibration7.1–9.0min.After each injection,theneedlewaswashed with0.1%formic acidin

H2O/MeCN/MeOH(2:1:1)andpurgedmobilephaseA

LC–MS/MSmethodassessmentwasperformedbyanalysingtest samples,representingeachsampleclass,onthreedifferentdays

Aninitialrunconsistedof33samples-threesamplesperclass whilelasttworunsconsistedof55samples–fivesamplesperclass Sampleswererandomizedineachrunwith‘solvent-QC-solvent’ sequenceinjectedevery8samplestomonitorforpossible carry-over,ionratiosalterationorsensitivitylossthroughouttherun.The initialtestsampleswerepartoftheauthenticsamplescollected butwerenot usedin developingthechemometricmodels.Test samples,purchasedfromlocalsupermarketsweretestedusingthe targetedassayandauthenticatedagainstthelabelclaim

Therelativeabundanceofthespeciesspecificmarkersofking prawn and tiger prawn, varying with geographical origin, was evaluated.RawdatawereprocessedbyTargetLynxv.4.(Waters, Milford,MA,USA) while statisticalanalysis(one-way ANNOVA, basedonabsoluteresponse)andassociatedgraphswereprepared

inGraphPadPrism5.01(GraphPadSoftware,Inc.,LaJolla,USA)

Alineardiscriminantanalysis(LDA)modelforpredictionof geo-graphicalorigin,usingtheUnscramblerXsoftware(Camo),was developedforthespecificmarkersofkingprawns.Themodelwas evaluatedtopredictthegeographicaloriginof60testsamplesof kingprawnswhichwerepartoftheauthenticsamplesgatheredfor theproject

3 Results and discussion

3.1 Highresolutionmassspectrometrydataprocessingand qualityvaluation

LiquidchromatographyhyphenatedtoaQToFmass spectrom-eter,withdataacquiredincontinuummode,isapowerfultoolfor unbiasedrecordofaccuratemassdataofeverysingledetectable metabolite in complex biological samples However, themajor bottleneck in such “global metabolomics” experiments is the unmanageableamountofdatageneratedwhichnecessitatesthe employmentofrobusttoolsfordatavisualization,pre-processing andmetaboliteidentificationtoensurequalityand reproducibil-ityofthedata[28,29].Theuntargetedmetabolomicsexperiments,

in positive and negativeionisation mode, generated morethan

1200gigabytesofrawdataeach.Thedatawasimportedintothe ProgenesisQIsoftwareandwascheckedforqualitybyreviewing theretentiontimealignmentofeachsampleagainstarandomly selectedQCsample.Alignmentscoresofthesamplesrangedfrom 82.9to 98.2% and 88.5to 98.0% in positive and negative ioni-sationmoderespectively;indicating excellentreproducibilityof thedatafortheentiredurationofthemassspectrometry

exper-Fig 1. Eleven different types of shrimps and pool QC samples clustering on PCA scores plot in ESI + (A) and ESI (B) mode.

iment.Various“minimumintensity”valuesforthe“absoluteion

intensity”filterwere triedto optimisepeak peaking sensitivity

oftheProgenesis QIsoftware.Thisoptimisation wasimportant

toensurethatthedetectedmolecularfeaturesareauthenticand

atthesametimerelevantfeaturesarenotmissed.Foreach

set-ting,thedatamatrixforallelevendifferenttypesofshrimpswas

importedinSIMCA14softwareforPCAanalysisandsubsequent

evaluation.Aminimumintensityvalueof1000,bothinpositive

and negativeionisation mode wasfoundto beoptimum

Nev-ertheless,a total of24,411molecularfeatures weredetectedin

positivieionisationmodewhereasinnegativeionisationmodea

totalof4921molecularfeaturesweredetected.AllQCswerefound

tobetightlyclusteredwithinthecentreofrespectivePCAscores

plots.PCAscoresplotin ESI+ andESI modeshowedclear

indi-cation of separation betweendifferent classes ofshrimp, using

second&fourthandsecond&fifthprincipalcomponents

respec-tively(Fig.1A, B).The firstsix principalcomponents explained

80and81%ofvariationinpositiveandnegativeionisationmode

respectively

Though,farmedkingprawnsamplesfromdifferent

geograph-ical origin grouped close by, there were clear indication of

discriminationbasedongeographicalorigin.Similarly,therewas

clearindicationofseparationbetweenfarmedandwildcaughttiger

prawns.TherecordedhighvaluesofR2X(cum)andQ2(cum)for

thePCAmodels(Table1;No.1and2)intheESI+andESI mode,

indicatewellexplainedcumulativevariationofthedatabythe

prin-cipalcomponentsandexcellentpredictioncapabilityofthemodels

[30].Representativetotalionchromatograms(TICs)ofQCsamples

inESI+ andESI mode, presentedintheFig.2.,showtheextent

ofcomplex and rich information obtainedfrom theuntargeted

metabolomicsexperiment

3.2 Chemometricmodelsforshrimpspeciesauthentication Forbuildingthemodelsforshrimpspeciesauthentication, sam-plesofthespeciesoriginatingfromdifferentgeographicalorigin and harvestingmethod wasgrouped as onespecies class.This resultedintotalfivespeciesclassesofshrimps.Thedatamatrixwas exportedtoSIMCA14andexploratoryPCAandOPLS-DAmodels werebuiltusingeitherunivariateorParetoscaling.TheVIPvalues

ofthemolecularfeaturesinanOPLS-DAplotwasusedtoselect themostrelevantfeatures.MetaboliteionswithaVIP>1generally representthosefeaturescarryingthemostrelevantinformation forclassdiscrimination[31].Hence,themolecularfeatureswith VIPscoreofmorethan1.5wastaggedandre-importedto Progen-esisQIsoftwaretocrosscheckthepeakqualityandintensityofthe features.Anumberofroundsofsuchfilteringofthedetected molec-ularfeaturesensuredselectionofmostrelevantmetabolitefeatures thatcontributeinclassificationofdifferentclassesofshrimp Finally,PCA (No.3& 4)modelsweregenerated considering

4914and900molecularfeaturesinpositiveandnegativeionisation moderespectively(Table1).ThePCAscoresplots(Fig.3A,B),both

inpositiveandnegativeionisationmode,showedclearseparation basedonthespeciesidentityoftheshrimpsamples.Thespecies classof“Blacktigerprawn”and“Kingprawn”representsamples

ofdifferentoriginorproductionmethod.Despitethisvariability

oforigin/productionmethod,thesamplesofkingprawnandblack tigerprawnformeddistinctiveclustersinthePCAplots.Whilstfirst fourcomponentsaccountedforaround80%cumulativevariationin thedata,separationwasmostlyachievedalongPC1andPC2.The bettervaluesofR2XandQ2forthePCAmodels(No.3&4)as com-paredtothePCAmodelsNo1and2(Table1 indicatedreliable selectionofmostrelevantfeatures

Fig 2.Representative total ion chromatograms (TICs) of pool QC samples in ESI+ (A)and ESI¯(B) mode showing extent of metabolite signature obtained from the shrimp samples.

Fig 3.Five different shrimp species and pool QC samples clustering on PCA scores plot in ESI+ (A) and ESI¯(B) mode; OPLS-DA scores plot of species discrimination of shrimp

Table 1

Values of different statistical parameters for developed chemometric models from detected metabolite features in positive and negative ionisation mode, where “A” is number

of multivariate component, “N” is number of samples, “R2X” is the fraction of the variation of the X variables explained by the model, “R2Y” is the fraction of the variation of the Y variables explained by the model, Q2 is the fraction of the variation of the X and Y variables that denotes the prediction ability of the model.

5 OPLS-DA 4 + 1+0 99 0.911 0.996 0.995 Positive mode; Univariate scaling; Discrimination of Species

6 OPLS-DA 5 + 2+0 99 0.927 0.989 0.988 Negative mode; Univariate scaling; Discrimination of Species

7 OPLS-DA 3 + 1+0 36 0.925 0.998 0.998 Positive mode; Univariate scaling; Discrimination of Tiger Prawn origin

8 OPLS-DA 3 + 1+0 36 0.975 0.994 0.993 Negative mode; Pareto scaling; Discrimination of Tiger Prawn origin

9 OPLS-DA 3 + 1+0 36 0.966 0.999 0.999 Positive mode; Univariate scaling; Discrimination of King Prawn origin

10 OPLS-DA 3 + 1+0 36 0.989 0.997 0.996 Negative mode; Pareto scaling; Discrimination of King Prawn origin

OPLS-DAmodel(No6)withfivepredictiveX-Ycomponentsand two

orthogonalcomponentswasgeneratedwithresultingR2X=92.7%,

R2Y=98.9%, Q2=98.8% and RMSECV of 5.2% for the ESI data

(Table1).AhighvalueforR2YandQ2(closerto1)indicatehigh

explainedvariation andpredictiveability ofan OPLS-DAmodel

respectively.Whereas,alowervalueof RMSECVindicatebetter

predictiveabilityofanOPLS-DAmodel[25].Thedifferentclassesof

shrimpspeciesappearedasbetterseparatedandtightlygrouped

clustersintheOPLS-DAscoresplotsinESI+ ascomparedtoESI

mode(Fig.3C,D),inaccordancetothebetterstatisticalparameters

obtainedinESI+mode

3.3 Chemometricmodelsforshrimporiginauthentication

OPLS-DAmodelsforthedatasetinESI+thanESI modewere

developed for prediction of geographical origin of the “Tiger

prawns”and the“Kingprawns”.Asimilarstrategy, followedfor

speciesauthenticationofshrimpsamples,wasadoptedfor

selec-tionof themostrelevant molecularfeatures that contributeto

discriminationof geographical origin of tiger prawns and king

prawns The OPLS-DA models for tiger prawns in ESI+ than

ESImode were finally built on datasets containing 1602 and

2081molecularfeaturesrespectively.TheOPLS-DAmodelsinESI+

modewasautofittedwiththreepredictiveX-Ycomponentsand

oneorthogonal component resulting in R2X (cum)=92.5%,R2Y

(cum)=99.8%,Q2(cum)=99.8%andRMSECV=2.5%.TheOPLSDA

model in ESI mode, fitted with similar numbers of

predic-tiveandorthogonalcomponentsresultedinequallygoodvalues

of R2X (cum)=97.5%, R2Y (cum)=99.4%, Q2 (cum)=99.3% and

RMSECV=4%(Table1,No7&8).Interestingly,thesamplesofwild

caughttigerprawnsfromIndiaandMadagascarclusteredcloser

together(shownwithin ellipse)in theOPLS-DA scoresplots in

ESI+andESI mode(Fig.4A,B);stronglyindicatingdiscrimination

basedonharvestingmethod.Samplesofwildcaughtandfarmed

tigerprawnsfromthesamecountryoforiginwerenotavailable

tofurtherexploresuchdiscriminationindetail.Similarly,OPLS-DA

scoresplotsforkingprawnsdisplayedwellseparatedandtightly

groupedclustersofkingprawnsoriginatedfromIndia,Thailand,

VietnamandHonduras(Fig.4C,D).Again,thevaluesofR2Y(cum)

andQ2(cum)wascloseto1establishingstrongpredictive

capabil-itiesofthemodelsinESI+andESImode(Table1.No9&10)

3.4 Validationofthechemometricmodels

Therecognitionabilityforalltheclassificationmodelsbothin

positiveandnegativeionisationmodewas100%aspresentedinthe

corresponding misclassification tables (Supplementary material

FigureS1) A misclassificationtablewhich providesa quantita-tivemeasureoftheperformanceofaclassordiscriminantanalysis modelisasummaryofthenumberofcorrectlyclassified observa-tions,withknownclassbelonging.AlltheOPLS-DAclassification models were then validated using the response permuatation option inSIMCA.TheY-datainthetraining setispermutedby randomlyshufflingtheirpositionwhilethenumericvalueremain same.Thepermutationprocedurecanberepeateda numberof timesbetween20and100.TheR2YandQ2Yvaluesofthederived modelsfromthepermutedY-dataarethencomparedwiththeR2Y andQ2Yvaluesoftherealmodeltocheckthevalidityofthe clas-sificationmodel.ThepermutationplotsfortheOPLS-DAmodels presentedinsupplementarymaterialFigureS2-S4summerizesthe resultofresponsepermutationtesting.Itcanbeobservedthateven after100permutations,theR2YandQ2Yvaluesofthedeveloped OPLS-DAmodelsaresubstantiallyhigherthanthecorresponding permutedvaluesindicatingvalidityofthemodels.TheR2Y inter-ceptbelow0.3 andQ2Yinterceptbelow0.05forthedeveloped OPLS-DAmodelsareanotherstrongindicationofthevalidityof themodels.Tofurtherconfirmthepredictionabilityofthe devel-opedmodels,20percentofthesamplesfromtheoriginaltraining setwastakenoutandusedastestsamplesetinthecorresponding refittedmodel.It canbeobservedintheclassificationlist (Sup-plementaryTableS1-S6)thatallthetestsamplesofknownclass identitywasrecognisedaccurately.Theuntargetedmetabolomics experimentwasrepeatedthreetimeswithfreshsamplesetsand eachtimeruggedchemometricmodelswerederivedestablishing reproducibilityoftheapproach

3.5 Biomarkerdiscoveryandputativeidentification

Atotalof36biomarkerswereputativelyidentifiedinESI+mode while a total of 33 ions were putatively identified in ESI ¯mode basedonvariousaccuratemassdatabasesearch(Supplementary TableS7-S8).Putative identificationof metaboloitecorresponds

toIdentificationLevel3ofmetabolomicsstandardinitiative[26]

Ametabolitemayproduces multipleaccuratemass signals cor-respondingtoisotopes(12C, 13C),adducts ([M+H]+,[M+Na]+), multimers([2M+H]+),charges([M+2H]2+),andneutralloss frag-ments([M+H-H2O]+)inaHRMS.Signalannotationwasperformed

inProgenesisQIsoftwareclusteringisotopes,basedontheisotopic patternofagivenmolecule.Theadductsofamoleculesharingthe sameretentiontimewasclusteredandtranslatedintoonevalue (molecularion).However,whereonlyoneadductisavailablefor

afeaturethem/zvalueofthepseudomolecularionwasreported Theseaccuratemassescanbesearchedagainstdatabase(s)withina definedwindowtoretrievepotentialcandidates.Anin-builtsearch engineinProgenesisQIwasusedtosearchtheaccuratemasses againstvarious databases.A masserrorof lessthan 2ppm and isotope similarityof more than 80% was consideredas criteria forreportingidentity Ameta-librarydeveloped within

Progen-Fig 4. OPLS-DA scores plot for discrimination of tiger prawn originated from India, Madagascar, Srilanka, and Vietnam in ESI+ (A) and ESI¯(B) mode; OPLS-DA scores plot for discrimination of king prawn originated from Honduras, India, Thailand, and Vietnam in ESI+ (C) and ESI¯(D) mode.

esisQIusingtheidentifiedbiomarkerswassuccessfully usedto

identify the metabolitesin shrimp samples during subsequent

metabolomicsexperiments.Asetof34biomarkersexclusivetothe

speciesidentityofashrimpsampleswerediscoveredfollowingthe

workflowpresentedinsupplementarymaterialFigureS5.Presence

oftheseexclusivebiomarkerswerefurtherconfirmedbyanalysing

theindividualrawdatafilesinMasslynx4.1softwareandwas

fur-therinvestigatedinLC–MS/MSfordevelopingatargetedmethod

ofshrimpauthentication.Unfortunatelyidentityoftheseexclusive

biomarkerscouldnotbeestablishedwithreasonableconfidenceby

accuratemassdatabasesearch

3.6 LC–MS/MStargetedmethodforshrimpauthentication

DuringthedevelopmentandassessmentofLC–MS/MStargeted

method34markerswereinitiallyevaluatedforselectivityinthe

first run Overall, 17 biomarkers were deemed species specific

(Table2).Duetolackoffullchemicalidentification,all

biomark-erswereassignedarbitraryIDsstemmingfromthespeciesnames

andaconsecutivenumberinwhichtheywereanalysed.Foreach

oftheanalysedshrimpsspeciesatleastonespecificmoleculewas

selectedi.e.kingprawn(KP1),tigerprawn(TP1,TP4–7),Indian

pinkshrimp(IPS1and4),Indianwhiteshrimp(IWS4,5,7,11)and

Argentinianredshrimp(ARS5and6).Nevertheless,other,less

spe-cificmarkerswerealsoincludedinthemethodduetorelativelylow

cross-talkortoaidgeographicaloriginelucidationi.e.KP2and3for

kingprawnandTP3fortigerprawn(Table2)

The selective markers provided a species specific response

throughoutthreeassessmentruns,wherebysample’sspecieswas

assigned only in the presence of all assigned markers

transi-tionswithcompliantion ratios TheLC–MS/MSmethodproved

toberobustwithnocarry-overorsensitivityloss(QCsresponse

RSD<10%forallmarkers,for100injectionruns).Additionally,ion

ratioswereintherangeof±20%ofthemeanQCsvalueforthe

threeanalyticalruns.RepresentativeXICofselectedspeciesspecific

markershavebeenpresentedinsupplementarymaterialFigureS6

Totrialtheassaywithmarketplacesamples,threeadditionalblind

runsconsistingoftotal76samplesofkingprawns,10samplesof

tigerprawnsand3samplesofeachArgentinianredshrimp,Indian

whiteshrimp,andIndianpinkshrimpwereperformed.Speciesof

thesampleswerecorrectlypredictedforallthespeciesassessed yielding0%falsepositivesandnegativesrate

Threemarkersforking(KP1,KP2,andKP3)andtigerprawns (TP3,TP4 andTP6)showedsignificant(p<0.001) differencesin responsebetweengeographicaloriginwithintheassessedspecies groups(Fig.5a–f).Thedifferenceinrelativeresponseofthesaid markersbasedongeographicaloriginisa strongindicationthat

itmightbealsopossibletoemploythosemarkersfor geograph-ical origin authentication of shrimp ona unit resolution triple quadrupolemassspectrometerplatform

Overtheyearsthesensitivityofhighresolutionmass spectrom-etershasimprovedtremendously,enablingdetectionofevenlow concentrationcompounds.However,asizable numberof recur-rentmolecularfeaturesinanymetabolomicsexperimentremain unknown.Chromatographicisolation oftheseunknown compo-nentsforidentificationisoftennotpracticalduetolowabundance Guessing theidentity and then synthesizing thecompound for confirmationmightalsofailandisanexpensiveaffair[32] Defin-ingtheserecurrentunidentifiedmetaboliteswithaccuratemass andfragmentions/spectraisapracticalsolutiontotheseproblem

In thisstudy wehave demonstratedthat aclass specific recur-rentunidentifiedbiomarkercanbeusedsuccessfully todevelop foodauthenticationassays.Thesuccessofthisapproachwasalso demonstratedfordetectionofadulterationinoregano[33]

4 Conclusions

Theleveloffraudinfisheriesgloballyisahugeissue.Thereare manymeasuresinplaceandinitiativesbeingdevelopedtotryand lessentheimpactthishasontheintegrityofseafoodthatisa sta-pleinthedietofbillionsofcitizensaroundtheworld.Intermsof thelaboratorytestingmethodsthatsupporttraceabilitysystems, therearequiteanumberoftheseandwhiletheyarefitforpurpose

in termsof uncoveringparticular aspectsoffraudulent practise theycanonlyprovideevidencethatoneparticularformof mal-practicemayhaveoccurred.Herewereportahighlyinnovative approach usinghighresolution mass spectrometryand chemo-metricsthatcandistinguishthespeciesidentityandgeographical originofshrimpinasinglemetabolomicsexperiment.Recurrent speciesspecificexclusivemarkerswereidentifiedfromthe

Table 2

Details of the LC–MS/MS method for the 18 markers employed in species and geographical origin elucidations Where: a cone voltage for all the compounds was set to 20 V, * markers deemed exclusive to the associated species.

King Prawn

398.2

0.063

40

143.1

0.108

20

Tiger Prawn

615.0

0.016

40

131.0

0.016

20

Indian Pink

Shrimp

382.2

0.044

40

Indian White

Shrimp

369.2

0.016

20

Argentinian

Red Shrimp

Fig 5. Box plots representing differences in response of king prawn (a, b, c) and tiger prawn (d, e, f) markers depending on the geographical origin of the samples Box plots with median of the absolute response measured in five samples per each region and whiskers at 5th and 95th percentile Significance levels: ***<0.001 and **<0.01.

Conflicts of interest

Acknowledgements

SB/OS/PDF-011/2015-16

04.001

References

[1] The State of World Fisheries and Aquaculture, 2018 (Accessed July 2018) http://www.fao.org/state-of-fisheries-aquaculture

[2] M.Á Pardo, E Jiménez, B Pérez, Villarreal, Misdescription incidents in seafood sector, Food Control 62 (2016) 277–283, http://dx.doi.org/10.1016/j.foodcont 2015.10.048

[3] S.M van Ruth, P.A Lunging, I.C.J Silvis, Y Yang, W Huisman, Differences in fraud vulnerability in various food supply chains and their tiers, Food Control

84 (2018) 375–381, http://dx.doi.org/10.1016/j.foodcont.2017.08.020 [4] J He, From country-of-origin labelling (COOL) to seafood import monitoring program (SIMP): how far can seafood traceability rules go? Mar Policy 96 (2018) 163–174, http://dx.doi.org/10.1016/j.marpol.2018.08.003 [5] I Ortea, J.M Gallardo, Investigation of production method, geographical origin and species authentication in commercially relevant shrimps using stable isotope ratio and/or multi-element analyses combined with chemometrics: an exploratory analysis, Food Chem 170 (2015) 145–153, http://dx.doi.org/10.1016/j.foodchem.2014.08.049

[6] Shrimp: Oceana Reveals Misrepresentation of America’s Favourite Seafood,

2019 (Accessed January 2017) https://oceana.org/news-media/publications/ reports/shrimpfraud

[7] Thai Seafood Action Plan March, 2016 (accessed November 2017) https:// www.nestle.com/media/news/progress-in-tackling-seafood-supply-chain-abuses

[8] M Bailey, S.R Bush, A Miller, M Kochen, The role of traceability in transforming seafood governance in the global South, Curr Opin Environ Sustain 18 (2016) 25–32, http://dx.doi.org/10.1016/j.cosust.2015.06.004 [9] L Tinacci, D Stratev, I Vashin, I Chiavaccini, F Susini, A Guidi, A Armani,

identification by DNA barcoding: a first survey on the Bulgarian market, Food

Control 90 (2018) 180–188, http://dx.doi.org/10.1016/j.foodcont.2018.03.007

[10] I Ortea, A Pascoal, B Ca ˜ nas, J.M Gallardo, J Barros-Velázquez, P Calo-Mata,

Food authentication of commercially-relevant shrimp and prawn species:

from classical methods to Foodomics, Electrophoresis 33 (2012) 2201–2211,

http://dx.doi.org/10.1002/elps.201100576

[11] D.I Ellis, H Muhamadali, D.P Allen, C.T Elliott, R Goodacre, A flavour of

omics approaches for the detection of food fraud, Curr Opin Food Sci 10

(2016) 7–15, http://dx.doi.org/10.1016/j.cofs.2016.07.002

[12] Y.T Lo, P.C Shaw, DNA-based techniques for authentication of processed food

and food supplements, Food Chem 240 (2018) 767–774, http://dx.doi.org/10.

1016/j.foodchem.2017.08.022

[13] M.F Mazzeo, R.A Siciliano, Proteomics for the authentication of fish species, J.

Proteomics 147 (2016) 119–124, http://dx.doi.org/10.1016/j.jprot.2016.03.

007

[14] I Ortea, G O’Connor, A Maquet, Review on proteomics for food

authentication, J Proteomics 147 (2016) 212–225, http://dx.doi.org/10.1016/

j.jprot.2016.06.033

[15] P Ferranti, The future of analytical chemistry in foodomics, Curr Opin Food

Sci 22 (2018) 102–108, http://dx.doi.org/10.1016/j.cofs.2018.02.005

[16] M Costas-Rodríguez, I Lavilla, C Bendicho, Classification of cultivated

mussels from Galicia (Northwest Spain) with European protected designation

of origin using trace element fingerprint and chemometric analysis, Anal.

Chim Acta 664 (2010) 121–128, http://dx.doi.org/10.1016/j.aca.2010.03.003

[17] O Fiehn, Metabolomics-the link between genotypes and phenotypes, Plant

Mol Biol 48 (2002) 155–171, http://dx.doi.org/10.1023/A:1013713905833

[18] D Cavanna, L Righetti, C Elliott, M Suman, The scientific challenges in

moving from targeted to non-targeted mass spectrometric methods for food

fraud analysis: a proposed validation workflow to bring about a harmonized

approach, Trends Food Sci Technol 80 (2018) 223–241, http://dx.doi.org/10.

1016/j.tifs.2018.08.007

[19] K Böhme, P Calo-Mata, J Barros-Velázquez, I Ortea, Recent applications of

omics-based technologies to main topics in food authentication, Trends

Analyt Chem 110 (2019) 221–232, http://dx.doi.org/10.1016/j.trac.2018.11.

005

[20] M Castro-Puyana, R Pérez-Míguez, L Montero, M Herrero, Application of

mass spectrometry-based metabolomics approaches for food safety, quality

and traceability, Trends Analyt Chem 93 (2017) 102–118, http://dx.doi.org/

10.1016/j.trac.2017.05.004

[21] S Esslinger, J Riedl, C Fauhl-Hassek, Potential and limitations of non-targeted

fingerprinting for authentication of food in official control, Food Res Int 60

(2014) 189–204, http://dx.doi.org/10.1016/j.foodres.2013.10.015

[22] G.P Danezis, A.S Tsagkaris, V Brusic, C.A Georgiou, Food authentication:

state of the art and prospects, Curr Opin Food Sci 10 (2016) 22–31, http://dx.

doi.org/10.1016/j.cofs.2016.07.003

[23] S.M Rocha, R Freitas, P Cardoso, M Santos, R Martins, E Fi gueira, Exploring

the potentialities of comprehensive two-dimensional gas chromatography

coupled to time of flight mass spectrometry to distinguish bivalve species: comparison of two clam species (Venerupi sdecussata and Venerupis philippinarum), J Chromatogr A 1315 (2013) 152–161, http://dx.doi.org/10 1016/j.chroma.2013.09.049

[24] C Black, O.P Chevallier, S.A Haughey, J Balog, S Stead, S.D Pringle, M.V Riina, F Martucci, P.L Acutis, M Morris, D.S Nikolopoulos, Z Takats, C.T Elliott, A real time metabolomic profiling approach to detecting fish fraud using rapid evaporative ionisation mass spectrometry, Metabolomics 13 (2017) 153, http://dx.doi.org/10.1007/s11306-017-1291-y

[25] C Black, S.A Haughey, O.P Chevallier, P Galvin-King, C.T Elliott, A comprehensive strategy to detect the fraudulent adulteration of herbs: the oregano approach, Food Chem 210 (2016) 551–557, http://dx.doi.org/10 1016/j.foodchem.2016.05.004

[26] L.W Sumner, A Amberg, D Barrett, M.H Beale, R Beger, C.A Daykin, T.W.M Fan, O Fiehn, R Goodacre, J.L Griffin, T Hankemeier, N Hardy, J Harnly, R Higashi, J Kopka, A.N Lane, J.C Lindon, P Marriott, A.W Nicholls, M.D Reily, J.J Thaden, M.R Viant, Proposed minimum reporting standards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI), Metabolomics 3 (2007) 211–221, http://dx.doi.org/10.1007/ s11306-007-0082-2

[27] Commission decision of 12 August 2002 implementing council directive 96/23/EC concerning the performance of analytical methods and the interpretation of results (2002/657/EC), Off J Eur Commun 221 (2002) 8–36 [28] T Kind, M Scholz, O Fiehn, How large is the metabolome? A critical analysis

of data exchange practices in chemistry, PLoS One 4 (2009) 440, http://dx.doi org/10.1371/journal.pone.0005440

[29] G.A Theodoridis, H.G Gika, E.J Want, I.D Wilson, Liquid chromatography–mass spectrometry based global metabolite profiling: a review, Anal Chim Acta 711 (2012) 7–16, http://dx.doi.org/10.1016/j.aca 2011.09.042

[30] L Eriksson, P.L Andersson, E Johansson, M Tysklind, Megavariate analysis of environmental QSAR data Part II – investigating very complex problem formulations using hierarchical, non-linear and batch-wise extensions of PCA and PLS, Mol Divers 10 (2006) 187–205, http://dx.doi.org/10.1007/s11030-006-9026-4

[31] H.W Cho, S.B Kim, M.K Jeong, Y Park, N Gletsu, T.R Ziegler, D.P Jones, Discovery of metabolite features for the modelling and analysis of high-resolution NMR spectra, Int J Data Min Bioinform 2 (2008) 176–192 [32] S Stein, Mass spectral reference libraries: an ever-expanding resource for chemical identification, Anal Chem 84 (2012) 7274–7282, http://dx.doi.org/ 10.1021/ac301205z

[33] E Wielogorska, O Chevallier, C Black, P Galvin-King, M Delêtre, C.T Kelleher, S.A Haughey, C.T Elliott, Development of a comprehensive analytical platform for the detection and quantitation of food fraud using a biomarker approach The oregano adulteration case study, Food Chem 239 (2018) 32–39, http://dx.doi.org/10.1016/j.foodchem.2017.06.083