Comprehensive comparative compositional study of the vapour phase of cigarette mainstream tobacco smoke and tobacco heating product aerosol

(1)

JournalofChromatographyA,1581–1582(2018)105–115

ContentslistsavailableatScienceDirect

Journal

of

Chromatography

A

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

Comprehensive

comparative

compositional

study

of

the

vapour

phase

of

cigarette

mainstream

tobacco

smoke

and

tobacco

heating

product

aerosol

夽

Benjamin

Savareear

a

_,

_Juan

_{Escobar-Arnanz}

a

_,

_Michał

_Brokl

b

_,

_Malcolm

_J.

_Saxton

b

_,

Chris

Wright

b

_,

_Chuan

_Liu

b

_,

_Jean-Franc¸

_ois

_Focant

a,∗

a_Centre_for_Analytical_Research_and_Technologies_(CART),_University_of_Liege,_Belgium b_Research_and_Development,_British_American_Tobacco,_Southampton,_UK

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received20August2018

Receivedinrevisedform12October2018 Accepted16October2018

Availableonline19October2018 Keywords:

Mainstreamtobaccosmoke Vapourphase

Tobaccoheating(heat-not-burntobacco) product

Thermaldesorption

Comprehensivetwo-dimensionalgas chromatography

Highresolutiontime-of-ﬂightmass spectrometry

a

b

s

t

r

a

c

t

Asimpledirectsamplecollection/dilutionandintroductionmethodwasdevelopedusingquartzwooland Tenax/sulficarbsorbentsforthermaldesorptionandcomprehensivetwo-dimensionalgas chromatogra-phy(TD-GC×GC)analysesofvolatileorganiccompoundsfromvapourphase(VP)fractionsofaerosol producedbytobaccoheatingproducts(THP1.0)and3R4Fmainstreamtobaccosmoke(MTS). Analy-seswerecarriedoutusingflameionisationdetection(FID)forsemi-quantificationandbothlowand highresolutiontime-of-flightmassspectrometry(LR/HR-TOFMS)forqualitativecomparisonandpeak assignment.Qualitativeanalysiswascarriedoutbycombiningidentificationdatabasedonlinear reten-tionindices(LRIs)withamatchwindowof±10indexunits,massspectralforwardandreverselibrary searches(fromLRandHRTOFMSspectra)withamatchfactorthresholdof>700(bothforwardand reverse),andaccuratemassvaluesof±3ppmforincreasedconfidenceinpeakidentification.Usingthis comprehensiveapproachofdatamining,atotalof79outof85compoundsandatotalof198outof202 compoundswereidentifiedinTHP1.0aerosolandin3R4FMTS,respectively.Amongtheidentified ana-lytes,asetof35compoundswasfoundinbothVPsampletypes.Semi-quantitativeanalyseswerecarried outusingachemicalclass-basedexternalcalibrationmethod.Acyclic,alicyclic,aromatichydrocarbons andketonesappearedtobeprominentin3R4FMTSVP,whereaslargeramountsofaldehydes,ketones, heterocyclichydrocarbonsandesterswerepresentinTHP1.0aerosolVP.Theresultsdemontsratethe capabilityandversatilityofthemethodforthecharacterizationandcomparisonofcomplexaerosol samplesandhighlightedtherelativechemicalsimplicityofTHP1.0aerosolincomparisontoMTS.

1. Introduction

Forthepurposeoftobaccoharmreduction,newgenerations

oftobacco heating(heat-not-burn- HnB)products(THPs)were

introducedinthemarket[1,2].Suchnewelectronicallycontrolled

heatingdevicessigniﬁcantlyimpacttheglobalchemical

composi-tionoftheaerosols,comparedtoconventionalcigarettes.Despite

thelargeamountofknowledgeandstandardisedmethodsexisting

fortheanalysisofconventionalcigarettes,verylittleinformation

夽 Selectedpaperfromthe42ndInternationalSymposiumonCapillary Chromatog-raphyand15thGCxGCSymposium,13–18thMay2018,Italy.

∗ CorrespondingAuthorat:UniversityofLiège,ChemistryDepartment–CART, Organic&BiologicalAnalyticalChemistry,Alléedu6AoûtB6c,B-4000,Liège, Belgium.

E-mailaddress:JF.Focant@uliege.be(J.-F.Focant).

is currently available for THPs. In addition, some studies have

reportedthescientiﬁcassessmentofTHPbasedontarget

analy-sesofcompoundstypicallyfoundincombustibleproducts[1,2]

andonlyafewmethodshavebeendevelopedfortheuntargeted

analysisofTHPandcombustiblesamples[3,4].

Themainstreamtobaccosmoke(MTS)thatexits theﬁlterof

a cigarette is an extremely complex chemical mixture [5,6]. It

consistsofliquid/soliddropletscalledtheparticulatephase(PP),

suspended in a mixture of gases and semi-volatiles called the

vapourphase(VP).Apartfromthebulkgases(nitrogen,oxygen,

carbonoxides,nitrogenoxides,ammonia),VPalsoconsistsofVOCs

andtheirimportanceonproductcytotoxicityandcarcinogenicity

hasbeendemonstratedinseveralcellularandanimalsystems[7].

AlthoughitcanbeexpectedthatTHPaerosolislesscomplexthan

MTS,atpresentthechemicalcompositionoftheVPofTHPaerosol

hasnotbeenfullydescribed[8].Developingnewqualitativeand

https://doi.org/10.1016/j.chroma.2018.10.035

(2)

106 B.Savareearetal./J.Chromatogr.A1581–1582(2018)105–115

quantitativeanalyticalmethodologiescapableofVPanalysesfor

bothTHPsandMTSisthereforeimportanttoallowpractical

com-parison.

MTSVPanalysesrequirespeciﬁcsamplingstrategiestoensure

representativecollectionofthewholechemical proﬁleacrossa

verylargedynamicrange[5,9].Variousapproaches,suchasgas

samplingbags[5],solidphasemicroextraction(SPME)[7],

solvent-ﬁlledimpingertrains[10],andcoldtraps[11]havebeenreported

forthispurpose.DirectsamplingofVPonspeciﬁcadsorbentshas

alsobeenreported[12–14].Inthelastfewyears,theuseof

adsor-bents has increasedin signiﬁcance as thermal desorption (TD)

capabilityhasbecomemorewidelyapplicableandrobust.Sucha

solvent-freetechniqueoffersseveraladvantagesoverothersolvent

extractionmethodsthatinvolvemoremanualstepsandoften

suf-ferfromreducedsensitivityduetomultipledilutionstepsbefore

ﬁnalmeasurement[15,16].Amongstotherapplications[17,18]TD

hasrecentlybeensuccessfullyusedfortheanalysisofMTSVPof

differentcigarettetypes[5].

Comprehensive two-dimensional gas chromatography

(GC×GC)coupledtotime-of-ﬂightmassspectrometry(TOFMS)is

theestablishedmethodofchoicefordetailedanalysisofVOCsin

medium-to-highcomplexitysamples[17–19].Advantagesofthe

state-of-artGC×GC-TOFMSinstrumentationhavebeendescribed

in severalreports [17,20,21]. In the context ofidentiﬁcation of

unknowns, an important advantage of GC×GC is the spatial

coordinationofchemicalclassesin2Dchromatogramsthat

pro-videsafurtheridentiﬁcationpointinadditiontolinearretention

indices (LRIs), fragmention and accuratemass MSdata during

thepeakassignmentprocess.Inordertoperformqualitativeand

semi-quantitative untargetedanalyses, the simultaneous useof

a ﬂameionisation detector(FID) and aTOFMS (dualdetection)

hasbeenreported[21,22].Despitethepotentialbeneﬁtsof this

approachintermsofspeedandefﬁciency,relativelyfewreported

applicationsinvolvesuchdualdetectionGC×GCinstrumentation

[23–28]. As GC_×GC can now be more easily coupled to fast

acquisitionHRTOFMS,whichoffersmassaccuracyatthesub-ppm

level consistently on deconvoluted mass spectral signals, an

extradimensionofelementalcompositionismadeavailablefor

compound identiﬁcation [29]. The utility of GC×GC-HRTOFMS

ishoweverconstrainedbythesizeoftheacquireddataﬁlesand

associateddataprocessingrequirements[30,31].Nevertheless,a

GC×GC-HRTOFMSapproachwassuccessfullyusedforcollecting

accuratemassvalues(±15ppm)forincreasedconﬁdenceinpeak

identiﬁcationofdifferenthopessentialoils[32].

Inthisstudy,wedevelopedanoriginalanalyticalmethodfor

the qualitative and semi-quantitative comparison of VOC

con-stituentsintheVPfractionofTHPaerosolandcombustibleMTS.For

thispurpose, TD-GC×GC-TOFMS/FIDand TD-GC×GC-HRTOFMS

wereusedtomaximizetheidentiﬁcationconﬁdenceforthe

non-targetedscreeningapproach.

2. Materialsandmethods

2.1. Analyticalreagentsandsupplies

Saturatedalkanestandardsolutions(C5–C30)werepurchased

fromSigmaAldrich(Diegem,Belgium). Acustom standard mix

made of benzene, toluene, 2-hexanone, p-xylene, styrene,

bro-mobenzeneando-cresol,waspreparedatconcentrationsof0.25

and0.5␮g/␮LfortheoptimizationofthesplitratiobetweenFID

andTOFMSandthetuningoftherecollectionprocess.Acustom

standardmix(SupplementaryTableS1)waspreparedfor

build-ing thecalibration curves for semi-quantiﬁcationpurposes. All

standardswerepurchasedfromSigma-Aldrich(Diegem,Belgium),

purity was >99.5%. All solutions were prepared volumetrically

usingmethanol(Sigma-Aldrich)andstoredat4◦C.Quartzwool

andTenax/sulﬁcarbstainlesssteelTDtubeswerepurchasedfrom

Markes(Pontyclun,UK).

2.2. Blankanalysisprocedure

Blank analysiswere performedtoensure analytical systems

were free from contaminations and that possible interferences

wereundercontrol.AllTDtubeswereconditionedpriortheiruse

accordingtomanufacturer’sinstructions.Ablanktestwasalways

performedbeforeuseforeachTDtubetocheckforpossible

carry-over.Thesmokingmachineswerelocatedinseparateroomsto

avoidanyinterferencesontheVOCproﬁles.Priortosampling,air

blankanalyseswereconductedoneachsmokingmachine.Forthisa

blankTDtubewasplacedinthesmokingmachineandthesmoking

procedurewasrun withouttobacco consumables.For each

rec-ollectionprocess,ablankrunwasalwaysperformedusingblank

TDtubes intheUnity-2and recollectionunit tocheckthatthe

instrumentwasfreeofcontamination.Afterthisachromatographic

runsequencewasmadeasfollows:instrumentalblank,smoking

machineairblank,recollectionprocedureblank,andan

instrumen-talblankprioranalysesofunknownsamples.Thisprocedurewas

repeatedforeachsampleandreplicateanalyses.

2.3. Samplingprocedure

2.3.1. Samplesandsamplecollection

The THP1.0 (gloTM) device and consumables were provided

byBritishAmericanTobacco(Southampton,UK).Thedescription

ofTHP1.0deviceandsamplingproceduredetailsareillustrated

inFig.S1(SupplementaryInformation).3R4Fresearchreference

cigaretteswereacquiredfromtheUniversityofKentuckyCollege

ofAgriculture(KentuckyTobaccoResearch&DevelopmentCenter,

USA).THPconsumablesandreferencecigaretteswereconditioned

inseparatehumidiﬁerforatleast48hat60%relativeair

humid-ityand22◦C[25].ForTHP1.0samples,VPaerosolsampleswere

generatedusingalinearsyringedrivesystemA14(BorgwaldtKC

GmbH,Germany).Asnostandardpufﬁngregimehasbeendeﬁned

forTHPsofar,allsamplecollectionswereconductedaccordingto

modiﬁedHealthCanadaIntense(HCI)pufﬁngregimeforcigarettes

thatconsistedofbell-shapedpuffs,eachof55mLwithpuff

dura-tionof2sandwith30sintervalsbetweenpuffs[33]withairinlet

zonenotoccluded.Thenumberofpuffscorrespondstotheheating

cycleavailableforTHP1.0usedinthepresentstudy(8puffs).One

THPconsumablewasusedforeachanalysis.Forthe3R4Freference

cigarette,MTSVPfractionwasgeneratedusingaBorgwaldtRM20D

smokingmachine(BorgwaldtKCGmbH,Germany).Smokingwas

conductedaccordingtotherelevantISOstandardsapplyinga35mL

puffof2sdurationtakenevery60swithnoblockingofﬁlter

venti-lationholes[34].Onereferencecigarettewasusedforeachanalysis.

SamplingproceduredetailsareillustratedinsupplementaryFig.S1.

Routineairﬂowandpuffvolumemeasurementswereperformed

priortothesmokerunsforeachsmokingmachines.ForTD

sam-pling,twoprepackedtubes(1stlevel)wereconnectedinseries,

quartzwoolTDtubesfortrappingthetotalparticulatephase(PP)

fractionandTenax/sulﬁcarbTDtubesfortrappingtheVOC

com-ponentofthevapourphase(VP)fraction.TDtubeswereplacedin

betweentheconsumableandthecorrespondingsyringepumpof

thesmokingmachines.Thetotalvolumeofgasdrawnthroughthe

sorbentsforTHPsamplingwas440mLforeachanalysis.Thetotal

volumeofgasdrawnthroughthesorbentsforreferencecigarette

samplingwas280mLforeachanalysis.TDtubeswerecappedwith

DiffLok® caps(MarkesLtd)directlyafterthesamplingprocedure

(3)

B.Savareearetal./J.Chromatogr.A1581–1582(2018)105–115 107

2.3.2. VPfractionrecollection/dilutionprocess

TheVP fractionofaerosol/smoke trappedona TDtube was

splitacrossthree2ndlevelTDtubesﬁlledwiththesamesorbent

priortothechromatographic injectionstoavoid overloadingof

chromatogramsandcontaminationoftheTDunitandGCcolumn.

For thispurpose, a Recollect-10device(SepSolveAnalytical Ltd

(Peterborough,UK)connectedtoUnity-2TDunitthroughaheated

transferlinewasused.Thisinstrumentiscapableofrecollecting

gaseoussamplesacrossuptotenTDtubesat atime. The

dilu-tionproceduredetailsareillustratedinsupplementaryFig.S2.1st

levelTDtubeswereplacedintheUnity-2thermaldesorberfortwo

stagesofdesorptionprocess.Thepre-desorptionstageconsisted

of2minofpre-purgeofthesampletubeatambienttemperature

withaﬂowrateof50mLmin−1.Duringthetubedesorptionstage,

thesampletubewasdesorbedat320◦Cfor 10minwitha ﬂow

rateof60mLmin−1 andtheﬂowequallysplitacrossthreetubes

placedinthemanifold.Themanifold,transferlineandﬂowpath

temperatureweremaintainedat200◦C.TDtubeswerecappedwith

DiffLok®caps(MarkesLtd)directlyaftertherecollectionprocedure

wascompletedandanalysedimmediatelyafterusingTD-GC×

GC-TOFMS/FIDandTD-GC×GC-HRTOFMSinstruments.

2.4. Instrumentalanalysis

TD-GC×GC-TOFMS/FIDanalyseswereperformedusingaLECO

Pegasus 4D (LECO Corp., St. Joseph, MI, USA) GC×GC system

equipped with a quad jet LN2 Cooled Thermal Modulator, a

secondary GC oven, an Agilent Technologies (Santa Clara, CA,

USA)capillaryﬂowtechnologysplitter,athermaldesorptionunit

(TD-100xr,MarkesLtd.),aLECOPegasustime-of-ﬂightmass

spec-trometer(TOFMS)andanAgilentFlameionizationdetector(FID).

TD-GC×GC-HRTOFMS analyses were performed using a LECO

PegasusGC-HRT4D(LECOCorp.)systemequippedwithaquadjet

LN2CooledThermalModulator,asecondaryGCoven,athermal

desorptionunit(TD-100xr,MarkesLtd.)andaLECOPegasushigh

resolutiontime-of-ﬂightmassspectrometer(TOFMS).Aschematic

overviewoftheTD-GC×GC-TOFMS/FIDdualdetectionset-upand

theTD-GC×GC-HRTOFMSset-upisgiveninsupplementary

infor-mation,Fig.S3.

2.4.1. Thermaldesorptionprocedure

SampleTDtubesunderwentthreestagesofdesorptionprocess.

Thepre-desorptionstageconsistedof0.1minofpre-purgeofthe

sampletube atambienttemperaturewitha ﬂow rateof50mL

min−1.Duringthetubedesorptionstage,thesampletubewas

des-orbedat300◦Cfor8minwithaﬂowrateof50mLmin−1.Samples

wererecollectedonacoldtrapcontainingproprietarysorbentfrom

Markes(‘sulphur/labile’carbonnumberC3-C30)atatemperatureof

25◦C.Thecoldtrapplussamplewaspurgedfor2minataﬂowrate

of50mLmin−1toremovepossibletracesofundesirablewaterby

divertingthemtotheventportpriortosampledesorptionintothe

GC×GCinstrument.Thecoldtrapwasdesorbedatthemaximum

heatingrateof24◦Csec−1 from25◦Cto315◦C,atwhichitwas

heldsubsequentlyforanother6minAtransferlineof1.5m

deac-tivatedfusedsilicacolumnwasconnectedbetweentheTD-100xr

andtheGCcolumninletandmaintainedat200◦C.Thesplitratio

forGC×GC-TOFMS/FIDanalysiswassetto1:50andforGC×

GC-HRTOFMSanalysiswassetto1:100.

2.4.2. GC×GC-TOFMS/FIDanalyses

For all analyses a non-polar, 5% diphenyl 95% dimethyl

polysiloxanephase(60m×0.25mmi.d.×0.5␮mdf)(Rtx-5SilMS,

RestekCorp.,Bellefonte,PA,USA)wasusedastheﬁrstdimension

(1_D)_column._Two₃₀_m_columns_were_connected_to_make_a₆₀_m

column.Theseconddimension(2_D)_was_a_midpolar_Crossbond®

silarylene phase column exhibiting similar selectivity to 50%

phenyl/50%dimethylpolysiloxane (1m×0.25mm i.d.×0.25␮m

df)(Rxi®-17SilMS,RestekCorp.).Allcolumnconnections,

includ-ing thetransferlineof TDto the1_D_column _and _two ₃₀_m1_D

columnsand2_D_column_were_made_using_SilTiteTM_␮-Unions_(SGE

International,Victoria,Australia).Theefﬂuentof the2_D_column

wasdirectedtoanAgilentthree-waycapillary ﬂowsplitterand

theefﬂuentwassplitviarestrictorsR1andR2totheTOFMSand

FID.Aconstantsplitpressureof3.8PSIwasusedforcalculating

the1:1splitbetweentheTOFMS/FIDsetup,corresponding

restric-tors(deactivatedfusedsilica)wereR1:2.78m×0.15mmi.d.and

R2:0.53m×0.15mmi.d.Thecarriergaswasheliumatacorrected

constantﬂowrateof1.2mLmin−1.Themainoventemperature

programstartedwithanisothermalperiodat40◦Cfor5min,then

arampof5◦Cmin−1upto300◦C,aﬁnalisothermalperiodat300◦C

for3min.Thesecondaryovenwasprogrammedwitha5◦Coffset

abovetheprimaryoventemperature.Themodulationparameters

consistedofa3smodulationperiod(PM)(0.7shotpulseand0.8s

coldpulsetime)andatemperatureoffsetof15◦Cabovethe

sec-ondaryoventemperature.TheFIDtemperaturewas350◦C,using

anairﬂowof400mLmin−1,hydrogenﬂowof40mLmin−1 and

a N2 make upgasﬂow of 30mLmin−1. TheTOFMS was

oper-atedinelectronimpactmodeusing70eV,detectorvoltagewas

at1800V,ionsourcetemperaturewassetat230◦C,transferline

temperaturewassetat250◦Candmassscanrangem/z35–500at

theacquisitionrateof100spectras−1.Dailymasscalibrationand

autotuningwereperformedusingperﬂuorotributylamine(PFTBA).

SampleswereacquiredusingChromaTOF®(LECOCorp.)software

version4.50.8.

2.4.3. GC×GC-HRTOFMSanalyses

1_D _and 2_D _columns _were _similar _to _TD-GC_×_GC-TOFMS/FID

setupexceptthe2_D_column_was_1.5_m_long_in _the_TD-GC_×

GC-HRTOFMSsetup.Sincetherewasnosplitterconﬁgurationinthis

instrument,the2_D_column_efﬂuent_was_directed_to_the_mass

spec-trometer.The mainoven temperatureprogram startedwithan

isothermalperiodat40◦Cfor5min,thenarampof2.2◦Cmin−1up

to130◦C,followedbyarampof2.8◦Cmin−1to300◦Candaﬁnal

isothermalperiodat300◦Cfor3min.Massspectrawereacquired

intherangem/z35–500attheacquisitionrateof100spectras−1,

inahighresolutionmodewitharesolvingpower>25,000FWHM

form/z218.9851.SampleswereacquiredusingLECOChromaTOF®

-HRT(LECOCorp.)softwareversion1.91.Allotherparameterswere

thesameasforGC×GC-TOFMS/FIDanalyses.

2.5. Dataprocessing

Theschemeofdataproductionandprocessingisillustratedin

supplementaryinformation,Fig.S4.TOFMS/FIDdatawereexported

as.pegand.csvﬁlesforbothTOFMSandFID,respectively.These

datawereseparatelyprocessedandsummarised(retentiontimes,

peak areas, library comparison etc.) using the pixel-based GC

ImageTM _(ZOEX_Corp.,_Houston, _TX,_USA)_software_package

ver-sionR2.5.FortheanalysisofVPandthecomparisonofsamples,

chromatogramswerealignedfollowingaprocedurebasedonthe

creationofatemplatechromatogramthatrecordspeakpatterns

andcarryingoutresamplingofthedatatomatchretentiontimes

usingGCProjectTM,partoftheGCImageTMsoftwarepackage

(Sup-plementary Fig.S5). The following blob detection criteria were

used:area22,volume300,000andSN>200wereappliedoneach

individualchromatogramforTOFMSandFID,basedona

compro-misebetweenthenumberofdetectedsignalsandthequalityofthe

recordedsignals,thelaterdependingstronglyonpeakshapesover

thelargedynamicrange.ForHRTOFMSdata,peakswerematched

usingmassspectralandGC×GCstructuredchromatogrampattern

(4)

Fig.1. RepeatabilityoftherecollectionprocessbasedonaveragepeakareaobtainedfromdirectTDinjection(Black),recollectionoveroneTDtube(Grey),andthreetubes (White)(Forinterpretationofthereferencestocolourinthisﬁgurelegend,thereaderisreferredtothewebversionofthisarticle).

Qualitative analysis was carried out using linear retention

indices C5-C25 (LRI window ±10) and mass spectral

similar-ity/reverse match (similarity/reverse > 700 MS spectra match)

againstlibraryspectrafromLRTOFMSandHRTOFMSinstruments.

Inaddition,anaccuratemassvalueof<±3ppmwasappliedto

giveincreasedconﬁdencefortentativelyidentiﬁedpeaksassuch

mass accuracy values allowed to univocally attribute a

molec-ular formula to most compounds of molecular weights below

200Da [35]. Librarysearchesof blobs/compounds(both TOFMS

andHRTOFMS)wereperformedusingNIST/EPA/NIHMass

Spec-tralLibrary(NIST14)andWileyRegistryofMassSpectralData(9th

edition)withamatchfactorthresholdof>700.Furtheranalysesof

GC×GC–TOFMS/FIDresultsusinginteractiveLRIﬁlters(±10range)

wereperformedusingNIST/EPA/NIHMassSpectralLibrary(NIST

14)database.IfnotavailableintheNIST14database,WEBbasedRI

collections(PubChem)informationwasused[36].

3. Resultsanddiscussion

3.1. Methoddevelopment

3.1.1. Thermaldesorptionsampling

Based ona previous study[5] thermal desorption(TD) was

usedasa uniqueapproachtoaccommodatethechemical

com-plexityandthelargedynamic rangecovered bycomponentsof

MTSand THPsamples.Despitethefact thatsample production

fromTHPandcombustibleproductsrequiresdifferentapparatus,

wholesmoke/aerosolTDsamplingwascarriedoutbyplacingaset

ofTDtubesintheoutputstream(SupplementaryFig.S1).Atﬁrst,

asingleTDtube(Tenax/sulﬁcarb)wasusedbutbreakthroughwas

observed(datanotshown),evenonasinglepuffbasis.Basedon

recentresearch[5],weincludedglassﬁberﬁlterpads(Cambridge

ﬁlterpads,CFP)whichresolvedthebreakthroughissue

(Supple-mentaryInformation,Fig.S7).TosimplifytheprocesstheCFPwas

replacedbyaTDtubecontainingquartzwool,creatingasample

trainthatcomprisedaﬁrstleveltube(Tube1)containingquartz

woolforparticulatephase(PP)trappingandanotherﬁrstleveltube

(Tube2)containingTenax/sulﬁcarbforVPtrapping.Theimpacton

backpressureandairﬂowoftwoTDtubesbetweentheproduct

andthepufﬁngmachinewasminimized(<10mL/min)bylimiting

thequantityofpackedquartzwoolto700mg/tube.Theefﬁciency

ofsamplingwasevaluatedbymonitoringpotentialbreakthrough

ofcompoundsusingathirdtubeplacedaftertheTube2.Underthe

abovesetuptherewasnobreakthorughforbothTHP1.0VPand

3R4FMTSVPsamples,whichisasigniﬁcantfactortoconsiderfor

comparisonofthesamples(SupplementaryInformation,Fig.S8).

The emissions from a whole consumable were collected in

ordertomanageknownvariationsbetweenindividualpuffs.This

resulted in highly concentrated samples that required dilution

before TD-GC×GC-TOFMS/FID analysis to prevent overloading

of instruments and carry-over between samples. The

recollec-tion/dilutionprocedure(SupplementaryInformation,Fig.S2)was

alsooptimizedtopreventbreakthroughontherecollectiontubes

andtoallowquantitativedilutionofthesamples.Asillustratedin

Fig.1forasetofrepresentativecompounds,acceptablelosseswere

observedwhenaTDtubecontainingasamplewasrecollectedonto

oneorthreeTDtubes.Theaveragelossbasedonpeakareaswas

9%andrangedbetween4%(Benzene)and19%(2-Hexanone),with

RSDsrangingbetween4%(Toluene)and13%(o-Cresol).

The potential impact on sample integrity of the storage of

TD tubes prior to analysis was evaluated by monitoring peak

areasofasetofselectedanalytes(2-Methylbutane,2-Butanone,

Benzene, Dimethylfurane, 2-Hexanone, Pyridine, Ethylbenzene,

Styrene, Limonene) over time. Peak areas did not signiﬁcantly

vary(<20%)overaperiodofﬁvedays.Nevertheless,sampleswere

alwaysanalysedwithin48hofcollectiononTDtubes.Tube

des-orptionconditionswereoptimizedtoensurethecompletetransfer

ofallanalytesfromtheTDtubetothe1_D_GC_column._This_was

sys-tematicallycontrolledbymonitoringblanklevelsofTDtubesand

ofthethermaldesorbercoldtrap.

3.1.2. Chromatography

TheTD-GC×GC-TOFMS/FIDinstrumentalset-up

(Supplemen-taryinformation,Fig.S3)wasoptimizedbyusingaconstantsplit

pressureapproach. Thedesiredsplitratio(1:1)wasmaintained

during the whole temperature program by calculating column

ﬂowsofbothrestrictors.Thesplitratiowasvalidatedbyreplicate

(n=6)sequentialanalysisofatestmixturebysingle(FID)anddual

detection(FIDandTOFMS).Ratioofaveragepeakareasofthe

sin-gleanddualdetectionanalysisdemonstratedtheefﬁcient1:1split

(5)

repro-B.Savareearetal./J.Chromatogr.A1581–1582(2018)105–115 109

Fig.2.Apexplotofvapourphasesamplefrom3R4FMTSandTHP1.0aerosolobtainedfromFID(top),LRTOFMS(middle)andHRTOFMS(bottom).

duciblity(averageratioof0,496±0,03;RSDvaluesfrom4,1%to

8,7%).

The chromatograms (three individual chromatograms from

three second level TD tubes) wereprocessed for both FID and

LRTOFMSseparately(Supplementary Information,Fig.S5). After

manuallycleaningtheimagesfromcolumnbleed,artifactpeaks

andmutliplehitsduetopeaktailingeffectsforhighlyabundant

compounds, theywere manually matched on a

compound-by-compoundbasisusingﬁrstandseconddimensionretentiontime

(1_t

Rand2tR)matching.Underthisapproach,thetotalnumbersof

peaksfoundtobepresentintheVPfractionofTHP1.0andMTS

were88and207,respectively.Fig.2showsthereconstructed

chro-matograms(apexplots)obtainedforTD-GC_×GC-TOFMS/FIDand

usedfordataprocessing.

InordertoalsocombinehighaccuracyMSdatatocompound

identiﬁcation,theTD-GC×GC-TOFMS/FIDmethodwastransposed

toTD-GC×GC-HRTOFMS.Althoughitmightbeconsidereda

triv-ialstep,thetranspositionofchromatographicmethodsbetween

vacuumoutletinstrumentsfromonethatincludessplittingofthe

ﬂowtoanotherthatusesasigniﬁcantlylongertransferline

(Sup-plementaryInformation,Fig.S3)wasnotstraightforward.Despite

allefforts(variationsofﬂows,columndimensionsandcoating,and

otherGC×GCoperationalconditions)itappearedtobe

impossi-bletoobtainchromatogramsthatalignretentiontimesbetween

LRTOFMSandHRTOFMS.

TheTD-GC×GC-HRTOFMSapexplotsshowninFig.2forTHP1.0

and 3R4F MTS represent the best compromise that could be

obtainedin terms ofrun time andchromatographic resolution.

EachoftheHRTOFMSsignalswasmatchedwiththecorresponding

FID/LRTOFMSsignal. Thisrequired signiﬁcantmanual

interven-tiontolocate thecorrespondingpeak basedonﬁrst dimension

linearretentionindex,positionofthepeakinthesecond

dimen-sionretentionplaneandmassspectraldataforeachcompound.

Despitethetechnicalchallengesexplainedabove,theFID,LRTOFMS

andHRTOFMSsignalsof88and207compoundswerematchedfor

THP1.0and3R4FMTS,respectively.

The developed analytical method was capable of analysing

molecules in the range of C4-C25. However, a limited number

ofunresolved(SupplementaryInformation,Fig.S10)compounds

elutingbeforetheC5regionwerenotconsideredfurtherbecauseof

thelackoflinearretentionindexinformationandalackoffragment

ionsbelowm/z35givinginaccuratepeakassignmentagainstlibrary

match.Adedicatedanalyticalmethodologycoupledtotheuseof

standardswould berequiredtoconsideranalytesbelowtheC5

regionbutwasoutofthescopeofthepresentstudy.When

(6)

110 B. Savareear et al. / J. Chromatogr. A 1581–1582 (2018) 105–115 Table1

Qualitativeandsemi-quantitativeresultsofvapourphasefractionofTHP1.0aerosol.

FID LRTOFMS HRTOFMS

PeakNo. Compoundname Chemical

formula

LRI LRIlib. Amount

(␮g/stick) a_Mean_±_SD MS for-ward match MS reverse match MS for-ward match MS reverse match Mass accuracy (ppm) Chemical class 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 2-Propanone*

Aceticacid,methylester

Carbondisulﬁde Propanal,2-methyl-* 2-Propenal,2-methyl-* 2,3-Butanedione* 2-Butanone*_$ Furan,2-methyl-* Furan,3-methyl-* Unknown

Propanoicacid,methylester

1,3-Pentadiene,3-methyl-,(E)-*

Butanal,3-methyl-* Butanal, 2-methyl-1-Penten-3-one 2,3-Pentanedione* Furan,2,5-dimethyl-* 2-Butanone,

3-hydroxy-Thiocyanicacid,methylester*

2,4-Dimethylfuran*

2-Vinylfuran*

Propanoicacid,2-oxo-,methylester

1-Butanol, 3-methyl-pyrazine 2-Pentanone,4-methyl-* 1H-Pyrrole,1-methyl-* Unknown Pyridine*$ Disulﬁde,dimethyl* 1H-Pyrrole 2-Pentenal, (E)-1,4-Dioxin, 2,3-dihydro-Toluene*_$ Thiophene,3-methyl-* 3-Hexanone 2-Hexanone*$ Cyclopentanone* Hexanal 2,4-Pentanedione 3(2H)-Furanone, dihydro-2-methyl-Furan,2-ethyl-5-methyl-* 1H-Pyrrole,1-ethyl-* 2,5-Furandione 2-Vinyl-5-methylfuran* 2-Furancarboxaldehyde C3H6O C3H6O2 CS2 C4H8O C4H6O C4H6O2 C4H8O C5H6O C5H6O – C4H8O2 C6H10 C5H10O C5H10O C5H8O C5H8O2 C6H8O C4H8O2 C2H3NS C6H8O C6H6O C4H6O3 C5H12O C4H4N2 C6H12O C5H7N – C5H5N C2H6S2 C4H5N C5H8O C4H6O2 C7H8 C5H6S C6H12O C6H12O C5H8O C6H12O C6H12O C5H8O2 C7H10O C6H9N C5H6O2 C7H8O C5H4O2 519 545 555 565 573 588 595 600 609 618 626 638 652 662 685 694 706 706 713 715 725 727 734 735 739 741 743 746 748 752 757 767 771 776 785 790 794 800 804 808 810 816 831 832 835 509 536 549 561 567 595 598 606 614 – 627 643 652 662 681 698 707 706 711 711 725 722 736 736 735 743 – 746 746 755 754 680# 770 776 784 790 791 800 795 809 802 821 830 826 833 13.3±0.15 3.9±0.28 0.3±0.04 8.4±0.13 5.4±0.24 15.7±0.21 1.2±0.08 3.1±0.09 0.7±0.02 0.1±0.00 0.3±0.03 0.6±0.05 7.4±0.02 5.9±0.24 0.3±0.03 5.2±0.09 trace 1.2±0.14 1.6±0.25 0.2±0.01 0.3±0.01 0.6±0.06 0.1±0.01 0.2±0.01 0.1±0.00 1.0±0.01 trace 1.8±0.25 2.1±0.04 2.3±0.31 1.8±0.06 0.2±0.01 0.4±0.02 0.2±0.01 0.3±0.02 0.1±0.01 0.1±0.01 0.9±0.14 0.1±0.00 1.5±0.17 0.2±0.01 0.2±0.01 0.6±0.09 0.5±0.06 15.7±0.23 964 952 899 967 962 885 966 929 762 – 856 865 952 953 899 945 941 885 907 838 886 927 826 958 726 948 – 959 960 926 900 748 917 708 935 783 971 905 880 947 773 892 726 909 957 964 960 944 967 962 891 966 934 785 – 856 878 952 953 899 948 962 892 907 856 893 959 879 958 766 948 – 963 961 949 900 757 927 814 935 805 971 905 915 947 783 892 906 959 957 886 963 960 933 940 963 832 960 799 – 849 744 942 944 720 953 959 973 959 952 928 958 903 967 – 936 – 969 967 948 875 806 949 801 950 903 935 810 757 954 777 906 737 883 950 899 963 989 933 940 974 929 960 884 – 897 864 942 947 764 958 959 980 967 952 930 958 911 967 – 940 – 969 967 954 875 815 958 818 952 903 945 837 768 954 789 906 879 924 950 1.4 −1.1 −1.5 −1.0 −0.4 −1.4 −1.0 −1.0 −0.7 – −0.3 −0.3 −0.4 −0.6 0.0 −0.5 −0.4 −0.2 −0.7 −0.8 −0.1 −0.1 1.5ˆ −0.3 1.2 0.3 – −0.8 −2.5 −1.4 −0.6 −0.6 −0.2 −0.4 −0.2 1.0 0.2 2.4ˆ −0.3 −0.4 −0.8 −0.1 1.2ˆ −0.5 −1.8 8 9 11 7 7 8 8 3 3 13 9 1 7 7 8 8 3 8 9 3 3 9 6 3 8 3 13 3 11 3 7 12 4 3 8 8 8 7 8 9 3 3 9 3 7

(7)

B. Savareear et al. / J. Chromatogr. A 1581–1582 (2018) 105–115 111 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 Vinylcrotonate 1,3-Cyclopentadiene, 5-(1,1dimethylethyl)-2,5-Diethylfuran 2-Hexenal, (E)-p-Xylene*$ 1H-Pyrrole, 2,5-dimethyl-2(3H)-Furanone, 5-methyl-o-Xylene* Bicyclo[3.1.0]hexan-2-one Styrene*_$ Heptanal 2-Methyl-5-isopropenylfuran 6,8-Dioxabicyclo[3.2.1]octane Oxepine, 2,7-dimethyl-1H-Pyrrole, 1-butyl-2-Heptanone, 6-methyl-Benzaldehyde* Dimethyltrisulﬁde* beta-Myrcene Furan, 2-pentyl-Unknown Octanal Unknown delta-3-Carene* o-Cymene* Benzene, 1,2,3-trimethyl-Benzylamine Limonene* Cyclohexanone, 2,2,6-trimethyl-trans-beta-Ocimene* Benzeneacetaldehyde 4-tert-Butyltoluene Unknown Nonanal 1-Methoxyadamantane Decanal Unknown Trimethyl-tetrahydronaphthalene Triacetin Naphthalene,1,3-dimethyl-* C6H8O2 C9H14 C8H12O C6H10O C8H10 C6H9N C5H6O2 C8H10 C6H8O C8H8 C7H14O C8H10O C6H10O2 C8H10O C8H13N C8H16O C7H6O C2H6S3 C10H16 C9H14O – C8H16O – C10H16 C10H14 C9H12 C7H9N C10H16 C9H16O C10H16 C8H8O C11H16 – C9H18O C11H18O C10H20O – C13H18 C9H14O6 C12H12 836 849 855 857 869 870 873 877 897 900 904 938 939 949 953 957 978 986 989 992 997 1008 1012 1017 1035 1035 1036 1039 1048 1052 1061 1067 1079 1113 1198 1216 1247 1279 1343 1385 783# 839 888# 854 865 867 873 881 793# 893 901 933 839# 944 975# 953 971 984 991 993 – 1003 – 1011 1026 1033 1035 1030 1047 1049 1060 1066 – 1109 1179# 1214 – 1250# 1344 1385 0.3±0.04 1.0±0.01 0.1±0.01 trace trace 0.2±0.02 0.6±0.04 0.3±0.00 0.1±0.01 0.1±0.01 0.1±0.01 0.2±0.00 0.2±0.02 0.2±0.00 0.1±0.00 0.1±0.00 0.1±0.10 0.4±0.03 0.4±0.01 0.2±0.01 trace trace trace 0.9±0.05 trace 0.1±0.00 trace 1.0±0.01 0.1±0.01 0.4±0.06 1.6±0.29 0.1±0.01 trace 0.1±0.01 0.1±0.00 0.1±0.00 trace 0.9±0.02 0.1±0.00 0.7±0.01 755 777 854 864 758 772 907 936 763 797 730 854 834 786 855 753 936 941 929 924 – 831 – 850 874 776 753 860 759 812 808 802 – 766 786 – – 906 785 – 815 845 854 969 889 828 933 947 778 911 782 854 850 786 867 832 936 941 965 924 – 851 – 891 952 871 829 885 848 891 929 816 – 779 796 – – 940 902 – 765 863 777 918 880 864 931 953 831 889 892 847 796 782 812 931 898 919 873 909 – 894 – 847 950 923 855 880 815 – 952 – – 888 766 844 – 832 915 899 850 864 856 923 880 867 935 953 835 919 911 921 796 782 849 933 919 924 877 912 – 905 – 851 950 923 855 880 905 – 956 – – 888 786 875 – 845 915 899 0.1 −0.5 −0.0 0.2 0.3 −0.0 −0.1 −0.4 0.1 −0.4 −0.5ˆ −0.8 −0.2 −0.6 −0.7 1.1ˆ −0.3 −0.2 0.7ˆ −0.4 – 0.8ˆ – −0.5 0.0 −0.0 −0.5 −0.2 0.1 2.6 −0.7 0.9 – 1.4ˆ 1.3 0.4ˆ – 1.0 −1.0ˆ −1.0 9 2 3 7 4 3 9 4 8 4 7 3 12 3 3 8 7 11 1 3 13 7 13 2 4 4 4 2 8 1 7 4 13 7 12 7 13 5 9 5

a_sample_mean_(n₌₄₎_along_with_standard_deviation_of_the_{measurements,}_*_compounds_found_in_reference_cigarette_smoke_VP_sample,_compounds_found_in_Hoffmann_list,_$_compounds_were_positively_identiﬁed_using_their

respectivestandard,-informationnotavailable,trace-concentrationbelow0.1␮g/stick,#estimatednon-polarretentionindexfromNISTdatabase,butsemi-standardcolumnretentionindexinformationnotavailablefrom

(8)

qualitativeandquantitativeinvestigationinTHP1.0VPand3R4F

MTSVP,respectively.

3.2. ComparisonsofVPaerosolfractionfromTHP1.0and3R4F

MTS

3.2.1. QualitativeaspectsusinglowandhighresolutionTOFMS

QualitativeanalysisoftheVPfractionofTHP1.0aerosolandMTS

wascarriedoutusingseveralidentiﬁcationcriteriasuchas

reten-tiontimes(1_t

Rand2tR)ofavailableauthenticreferencestandards,

1_tR_linear_retention_index_matches₍_±10_window),_and_mass

spec-tralforwardandreversesimilaritymatches(>700)againstspectral

librariesforbothTOFMSandHRTOFMSdata.Inaddition,accurate

massinformation within±3ppm window based oneither

par-entionsorabundantfragments(limitedcases)wasemployedto

improvetheconﬁdenceofidentiﬁcationofcompounds.Asseenin

Tables1andS2,insomelimitedcases,eitherLRIdatawerenot

availableortheMSmatchfactorwasnotobtainedfromeitherthe

loworhighresolutionmassanalyserbecauseofinterferingions,

butpeakassignmentwasalwaysbackedupbyatleastgoodMS

match(>700)ontheotherMSanalyserandgoodmassaccuracy

(<_±3ppm)values.Inaddition,thespatialcoordinationof

homolo-gousseriesofcompoundsinGC×GCchromatogramsaidedfurther

conﬁrmationofidentities.

As expected, reverse library matches (ranged 757–971 and

764–989forLRTOFMSandHRTOFMS,respectively)wereslightly

betterthanforwardmatches(ranged 708–971and720–973 for

LRTOFMS and HRTOFMS, respectively) and HRTOFMS matches

were globally betterthan LRTOFMS matches, even though MS

librariesaremadeofLRMSdatasets.Thiscanbeexplainedpartly

bythehighermassresolutionoftheHRTOFMSinstrument,which

providesnarrowermasswindowsandreducestheimpactof

inter-feringionsfromco-elutingspeciesandbackgroundnoise[37].

BecauseoftheoperationoftheHRTOFMSinelectronimpact

mode(70eV)toprovidethedesiredfragmentationforMSlibrary

searchpurpose,forsomecompoundstheabundanceofthe

molec-ularionwasverylow.Thiswasparticularlythecaseforalkanes,

aldehydes, alcohols and nitriles, for which the most abundant

fragmention wasused to estimatemass accuracy,as reported

earlier[32].Theuseofasofterionisationtechniquesuchas

UV-basedphoto-ionisation(PI)iscurrentlyunderinvestigationinother

projectsandmayimproveMSdataqualityforfragmentionsand

molecularions[38].Inpractice,82%and94%oftheidentiﬁed

ana-lytespresentedausablemolecularionforidentiﬁcationforTHP1.0

and3R4FMTS,respectively.Amongthese,70%oftheassignedpeaks

exhibitedamassaccuracyof±1ppm,forbothsampletypeswhile

15%and25%ofassignedpeaksexhibitedamassaccuracybetween

±1and±3ppmforTHP1.0and3R4FMTS,respectively.Despitethe

highaddedvalueofaccuratemasstoMSlibrarycomparisonsfor

identiﬁcationofcompounds[39],theunequivocalidentiﬁcationof

compoundspresentindifferentisomericformsstillrequiresthe

useofLRIsandideallytheanalysisofpurestandardcompounds,

whichisdifﬁculttoimplementinnon-targetedstudies.Byapplying

thecomprehensivedataminingstrategiesdescribedpreviously79

(outofthe85)(Table1)and198(outofthe202)compoundswere

identiﬁedintheTHP1.0aerosolVPand3R4FMTSVP

(Supplemen-taryTableS2),respectively.Theremainingfewcompoundscould

notbeidentiﬁedatthisstage.Nevertheless,forbothTHP1.0VPand

3R4FMTSVPsamples,itconstitutestheﬁrstcomprehensivelistof

VOCscomposingtheirchemicalproﬁle.Previously,aTD-GC×

GC-TOFMSapproachwasabletotentativelyidentify127compounds

[5].Whencomparedtoourcurrentapproachthatwasableto

iden-tify181compoundsinthecarbonnumberrangeC6-C14,asetof56

compoundswerefoundtobecommontobothstudies(TableS2).

WhenweconsidertheidentiﬁedcompoundsofTHP1.0VPand

3R4FMTSVPsamples,35compoundswerecommontobothTHP1.0

aerosolVPand3R4FMTSVPsamplesandarehighlightedinTable1

andinSupplementaryinformation,TableS2.Theidentiﬁed

com-pounds were further grouped in chemical classes (Table 2)to

highlightmajor chemical differencesbetweenthe two typesof

aerosolVP.Itappearedthatthehigherchemicalcomplexityof3R4F

MTSsamples(202compoundsversus85compoundsforTHP1.0

aerosolVP)wasmainlyduetothelargenumberofacyclic,alicyclic

and monocyclic aromatichydrocarbons.In 3R4F MTS VP,these

threechemicalfamiliesaccountedfor71%ofthe202compounds

found.ForTHP1.0VP,thesethreechemicalfamiliesaccountedfor

16%ofthe85compoundsfound,whileheterocyclics,aldehydesand

ketonesaccountedfor53%.Thesedataillustratethechemical

dif-ferencebetweentheVPproducedbycombustionandpyrolysisof

tobaccoandtheVPofanaerosolproducedbylowertemperature

heatingoftobacco[4].Interestingly,italsoappearedthat,among

theso-called‘Hoffmannlist’of44compoundsofprimeinterestin

termsoftoxicity,sevenandsixanalyteswerepresentinVPsamples

generatedin3R4FMTS(SupplementaryInformation,TableS2)and

THP1.0aerosol(Table1),respectively.Amongthem,therewere5

commoncompounds(2-propanone,2-butanone,pyridine,toluene

andstyrene).

3.2.2. Semi-quantitativeanalysisusingFID

FID/TOFMS dual detection was intended to facilitate

semi-quantitative analysis of the samples. Tobacco-related samples

containhundredsofcompoundsoverseveralordersofmagnitude

ofconcentration[5,40].Thispresentsachallengeformass

analy-sersthathavelineardynamicrangeslimitedto3or4decades.After

robustassignmentofcompoundidentitiesbasedon(HR)TOFMS,

FID signals were used for semi-quantiﬁcation [21,22]. Because

of the non-targeted nature of analysis, a generic, external

cal-ibration wasconstructedusing representativecompoundsfrom

thirteenchemicalclasses(SupplementaryInformation,TableS1).

Theselectionofstandardcompoundswasbasedontheassigned

VOCchemicalclassesinVPfractionsofTHP1.0aerosoland3R4F

MTS samples.The calibrationrange wasbased ondetectability

ofboth FIDand TOFMSdetectors (datanot shown).Thelowest

pointwassetat0.1␮g/␮L, lowerconcentrationswerereported

astraces.Thehighest point(50␮g/␮L)wassetaftermeasuring

THP1.0VPsamples.Threecompoundsin3R4FMTSVP(TableS2)

wereoutofthatrange,thusthesecompoundsconcentrationswere

estimatedfromanextentionofthecalibrationplot.Compounds

weresemi-quantiﬁedusingtheresponsefactorobtainedforthe

representativesubstanceofthesameclassinthecalibration

solu-tionafternormalisation betweensamplesandcalibrationusing

anotherexternalstandardtoaddresspotentialinstrumentaldrift.

Ifnorepresentativecompoundwasincludedincalibration,

non-classiﬁedcompounds(organosulfur,miscellaneousandunknowns)

weresemi-quantiﬁedusingthefactorsobtainedforbromobenzene,

acetonitrileandp-xylene,respectively.

The resultsof the external 9-pointcalibration (n=4)at the

␮g/␮Lleveldemonstratedacceptablereproducibility(RSD%<6%,

range3–6%)forallcompounds(SupplementaryInformation,Table

S1).Semi-quantitativeresultsforchemicalclassesidentiﬁedinthe

vapourphasefractionsofTHP1.0aerosoland3R4FMTSare

pre-sentedinTable2.

Fig. 3 illustrates the differences in chemical composition

and abundancebetweenTHP1.0VPand 3R4F MTS VP samples.

THP1.0VP generally contained lower abundances of the less

volatilecomponentsofchemicalclasses,whichwouldbeexpected

whenconsideringthetemperaturesatwhichtheaerosolswere

formed.The major THP1.0VP classes were aldehydes, ketones,

estersand heterocycliccompounds.In 3R4F MTSVP,

hydrocar-bons,ketones,andheterocycliccompoundsweremoreabundant

(9)

B.Savareearetal./J.Chromatogr.A1581–1582(2018)105–115 113

Table2

Summaryofqualitativeandsemi-quantitativeinformationforVPsamplesofTHP1.0aerosoland3R4Fcigarettesmokewiththeirrespectiverelativepercentage.

THP1.0VP 3R4FMTSVP

Chemical class

Chemicalclassname No.of

analytes Relative %of analyte Amount (␮g/stick) Relative %of amount No.of analytes Relative %of analyte Amount (␮g/stick) Relative %of amount 1 2 3 4 5 6 7 8 9 10 11 12 13 Acyclichydrocarbons Alicyclichydrocarbons Heterocycliccompounds Monocyclicaromatic hydrocarbons Polycyclicaromatic hydrocarbons Alcohols Aldehydes Ketones Esters Nitriles Organosulfurcompounds Miscellaneouscompounds Unknowns 3 3 18 9 2 1 14 14 9 ND 3 3 6 4 4 21 11 2 1 16 16 11 ND 4 4 7 1.4 2.8 11.9 1.1 1.6 0.1 48.3 37.9 9.5 0.0 2.8 0.2 0.1 1 2 10 1 1 0 41 32 8 0 2 0 0 70 41 17 33 3 ND 7 14 1 8 2 2 4 35 20 8 16 1 ND 3 7 0 4 1 1 2 496.8 173.1 69.4 119.0 2.4 0.0 57.6 264.4 2.5 36.8 4.7 1.2 1.0 40 14 6 10 0 0 5 22 0 3 0 0 0 ND-notdetected.

Fig.3.TD-GC×GC-LRTOFMStwo-dimensionalapexbubbleplotillustratingthedifferenceinrelativechemicalcomplexitybetween3R4FMTS(top)andTHP1.0aerosol (bot-tom).Bubblesizesarerelatedtolevelsofanalytes.Fordatavisualisationpurpose,bubblesizeswererescaledbasedonconcentrationintervals:<0.1␮g=1;0.1-0.25␮g=2.5; 0.25-0.5␮g=5;0.5–1␮g=10;1–2.5␮g=15;2.5–5␮g=20;5–10␮g=25;10–25␮g=30;25–50␮g=35;50–100␮g=40;>100␮g=60(seeTables1andS2fordetailedvalues).

(10)

Fig.5.Differencesofchemicalconcentrationof35overlappingcompoundsfoundinboth3R4FMTSVPandTHP1.0VPsamples.Y-axisinsertedaslogarithmicscalefordata visualisationpurpose(actualvaluesareprovidedinTables1andS2).

The total abundance of compounds present in the VP

frac-tionof3R4FMTSwasestimatedtobetentimesgreaterthanin

THP1.0aerosolVP.In3R4FMTSVP,acyclic,alicyclicand

mono-cyclicaromatichydrocarbonsaccountedfor64%intermsofthe

totalestimatedconcentration.ForTHP1.0VP,thesethreechemical

familiesrepresentedlessthan4%ofthetotalestimated

concen-tration.Aldehydes,ketones,andheterocyclicsaccountedfor41%,

32%and10%respectivelyofthetotalestimatedconcentrationof

analytesinTHP1.0VP.Table1(andSupplementaryinformation,

TableS2)providesindividualsemi-quantitativevaluesin␮gper

consumable(stick)foreachassignedcompound.Fig.4illustrates

thechemicaldistributionofthe35compoundspresentinboth

sam-pletypes.Thesummedconcentrationofthese35compoundswas

sixtimeshigherin3R4FMTSVP(457.5␮g/stick)thaninTHP1.0VP

(73.7␮g/stick).Whenconsideringrelativeamounts,ofthe35

com-moncompoundsfoundinTHP1.0and3R4FMTSproducts,higher

concentrationsweresystematicallymeasuredinMTS,onawhole

productbasis,exceptforpyridineanddimethyltrisulﬁde(Fig.5).

Inthesetwocases,thelevelsinTHP1.0appearedtobemarginally

higherbutconﬁrmationofanydifferencewouldrequire

quantita-tiveanalysis.Finally,intermsofsemi-quantiﬁcation,2-propanone

(oneofthe‘Hoffmannlist’toxicants)waspresentatsigniﬁcantly

lowerconcentrationsinTHP1.0VPsamples(13.3␮g/stick)thanin

3R4FMTSVP(152_␮g/stick).Similarly,theconcentrationsofother

Hoffmannlistcompounds(e.g.toluene,2-butanoneandstyrene)

weresigniﬁcantlyreducedinTHP1.0VPcomparedto3R4FMTSVP

(seeTables1andTableS2fordetailedvalues).Resultswerealso

comparedtoLietal.data[3],despitethefactthattheyuseda

tar-getapproachusingdifferentTHP2.2producttocomparetoMTS

VP.Theystudiedthepercentagereductionofanalytes

concentra-tionbetweenTHP2.2andMTSsamples.Thepercentagereduction

rateoftolueneand2-propanonewere97%and87%,respectively.

Almostsimilarresultswerefoundinthepresentworkfortoluene

(99%)and2-propanone(91%).Thereduction rateof2-butnaone

wasfoundtobe41%intheirstudy[3]althougha97%reduction

ratewasobservedforTHP1.0VPsample.Thepresentworkwas

basedonapreliminarysemi-quantitativeapproachthatrequires

furthervalidtionusingspeciﬁccalibrationsolutions,ideallyusing

stableisotopedilution.

4. Conclusions

A novel characterization approach based on the use of

independentandcomplementaryTD-GC×GC-TOFMS/FIDand

TD-GC×GC-HRTOFMSmethodshavebeendevelopedfortheanalysis

oftheVPaerosol fractionofTHP1.0and 3R4FMTS.Compounds

wereidentiﬁedusingLRIs,MSmatchesagainstspectrallibraries

usingbothLRTOFMSandHRTOFMSdata,andaccuratemassvalues.

Thiscomprehensivedataminingapproachpermittedthe

assign-ment of chemical identity for more than 90% of the detected

constituentsforbothsampletypes.Thechemicalcompositionof

theVPofTHP1.0aerosolwasobservedtobemuchlesscomplex

than3R4FMTSVP.TheGC×GC-FIDsemi-quantitativedata

indi-catedthatthetotalabundanceofanalytesintheTHP1.0VPwasten

timeslowerthaninthe3R4FMTSVP.Inconclusion,thepresent

studyprovidesdatathatcontributetoamoreextensive

chemi-calcharacterisationoftheaerosolsgeneratedbytobaccoheating

products.

AppendixA. Supplementarydata

Supplementarymaterialrelatedtothisarticlecanbefound,in

theonlineversion, atdoi:https://doi.org/10.1016/j.chroma.2018.

10.035.

References

[1]J.-P.Schaller,D.Keller,L.Poget,P.Pratte,E.Kaelin,D.McHugh,G.Cudazzo,D. Smart,A.R.Tricker,L.Gautier,M.Yerly,R.R.Pires,S.LeBouhellec,D.Ghosh,I. Hofer,E.Garcia,P.Vanscheeuwijck,S.Maeder,Evaluationofthetobacco heatingsystem2.2.Part2:chemicalcomposition,genotoxicity,cytotoxicity, andphysicalpropertiesoftheaerosol,Regul.Toxicol.Pharmacol.81(2016) S27–S47.

[2]M.Forster,S.Fiebelkorn,C.Yurteri,D.Mariner,C.Liu,C.Wright,K.McAdam,J. Murphy,C.Proctor,AssessmentofnoveltobaccoheatingproductTHP1.0.Part 3:Comprehensivechemicalcharacterisationofharmfulandpotentially harmfulaerosolemissions,Regul.Toxicol.Pharmacol.93(2018)14–33.

[3]X.Li,Y.Luo,X.Jiang,H.Zhang,F.Zhu,S.Hu,H.Hou,Q.Hu,Y.Pang,Chemical analysisandsimulatedpyrolysisoftobaccoheatingsystem2.2comparedto conventionalcigarettes,NicotineTob.Res.(2018)1–8.

[4]B.Savareear,R.Lizak,M.Brokl,C.Wright,C.Liu,J.-F.Focant,Headspace solid-phasemicroextractioncoupledtocomprehensivetwo-dimensionalgas chromatography-time-of-ﬂightmassspectrometryfortheanalysisofaerosol fromtobaccoheatingproduct,J.Chromatogr.A1520(2017)135–142.

(11)

B.Savareearetal./J.Chromatogr.A1581–1582(2018)105–115 115 [5]B.Savareear,M.Brokl,C.Wright,J.-F.Focant,Thermaldesorption

comprehensivetwo-dimensionalgaschromatographycoupledtotimeof ﬂightmassspectrometryforvapourphasemainstreamtobaccosmoke analysis,J.Chromatogr.A1525(2017)126–137.

[6]M.Brokl,L.Bishop,C.G.Wright,C.Liu,K.McAdam,J.F.Focant,Multivariate analysisofmainstreamtobaccosmokeparticulatephasebyheadspace solid-phasemicroextractioncoupledwithcomprehensivetwo-dimensional gaschromatography-time-of-ﬂightmassspectrometry,J.Chromatogr.A1370 (2014)216–229.

[7]Q.Ye,Developmentofsolid-phasemicroextractionfollowedbygas chromatography-massspectrometryforrapidanalysisofvolatileorganic chemicalsinmainstreamcigarettesmoke,J.Chromatogr.A1213(2008) 239–244.

[8]E.Simonavicius,A.McNeill,L.Shahab,L.S.Brose,Heat-not-burntobacco products:asystematicliteraturereview,Tob.Control0(2018)1–13.

[9]J.-Z.Dong,J.N.Glass,S.C.Moldoveanu,AsimpleGC–MStechniqueforthe analysisofvaporphasemainstreamcigarettesmoke,J.MicrocolumnSep.12 (3)(2000)142–152.

[10]T.Miyake,T.Shibamoto,Quantitativeanalysisbygaschromatographyof volatilecarbonylcompoundsincigarettesmoke,J.Chromatogr.A693(1995) 376–381.

[11]K.G.Darrall,J.A.Figgins,R.D.Brown,G.F.Phillips,Determinationofbenzene andassociatedvolatilecompoundsinmainstreamcigarettesmoke,Analyst 123(1998)1095–1101.

[12]G.Barrefors,G.Petersson,Assessmentofambientvolatilehydrocarbonsfrom tobaccosmokeandfromvehicleemissions,J.Chromatogr.A643(1993) 71–76.

[13]S.M.Charles,S.A.Batterman,C.Jia,CompositionandemissionsofVOCsin main-andside-streamsmokeofresearchcigarettes,Atmos.Environ.41 (2007)5371–5384.

[14]S.Uchiyama,T.Tomizawa,Y.Inaba,N.Kunugita,Simultaneousdetermination ofvolatileorganiccompoundsandcarbonylsinmainstreamcigarettesmoke usingasorbentcartridgefollowedbytwo-stepelution,J.Chromatogr.A1314 (2013)31–37.

[15]E.Woolfenden,Sorbent-basedsamplingmethodsforvolatileand

semi-volatileorganiccompoundsinair.Part1:sorbent-basedairmonitoring options,J.Chromatogr.A1217(2010)2674–2684.

[16]P.-H.Stefanuto,K.Perrault,S.Stadler,R.Pesesse,M.Brokl,S.Forbes,J.-F. Focant,Readingcadavericdecompositionchemistrywithanewpairof glasses,ChemPlusChem79(2014)786–789.

[17]S.Stadler,P.-H.Stefanuto,M.Brokl,S.L.Forbes,J.-F.Focant,Characterization ofvolatileorganiccompoundsfromhumananaloguedecompositionusing thermaldesorptioncoupledtocomprehensivetwo-dimensionalgas chromatography–time-of-ﬂightmassspectrometry,Anal.Chem.85(2013) 998–1005.

[18]S.L.Forbes,K.A.Perrault,P.-H.Stefanuto,K.D.Nizio,J.-F.Focant,Comparison ofthedecompositionVOCproﬁleduringwinterandsummerinamoist, mid-latitude(Cfb)climate,PLoSOne9(2014).

[19]M.Brokl,L.Bishop,C.G.Wright,C.Liu,K.McAdam,J.F.Focant,Analysisof mainstreamtobaccosmokeparticulatephaseusingcomprehensive two-dimensionalgaschromatographytime-of-ﬂightmassspectrometry,J. Sep.Sci.36(2013)1037–1044.

[20]J.Dallüge,R.J.J.Vreuls,J.Beens,U.A.Th.Brinkman,Optimizationand characterizationofcomprehensivetwo-dimensionalgaschromatography withtime-of-ﬂightmassspectrometricdetection(GC×GC-TOFMS),J.Sep.Sci. 25(2002)201–214.

[21]R.Shellie,P.Marriott,P.Morrison,Comprehensivetwo-dimensionalgas chromatographywithﬂameionizationandtime-of-ﬂightmassspectrometry detection:qualitativeandquantitativeanalysisofwestAustralian sandalwoodoil,J.Chromatogr.Sci.42(2004)417–422.

[22]Z.L.Cardeal,M.D.R.GomesdaSilva,P.J.Marriott,Comprehensive

two-dimensionalgaschromatography/massspectrometricanalysisofpepper volatiles,RapidCommun.MassSpectrom.20(2006)2823–2836.

[23]T.Dijkmans,M.R.Djokic,K.M.VanGeem,G.B.Marin,Comprehensive compositionalanalysisofsulfurandnitrogencontainingcompoundsinshale oilusingGC×GC–FID/SCD/NCD/TOF-MS,Fuel140(2015)398–406.

[24]L.Nicolotti,C.Cordero,D.Bressanello,C.Cagliero,E.Liberto,F.Magagna,P. Rubiolo,B.Sgorbini,C.Bicchi,Paralleldualsecondarycolumn-dualdetection: afurtherwayofenhancingtheinformativepotentialoftwo-dimensional comprehensivegaschromatography,J.Chromatogr.A1360(2014)264–274.

[25]D.Bressanello,E.Liberto,M.Collino,S.E.Reichenbach,E.Benetti,F.Chiazza,C. Bicchi,C.Cordero,Urinarymetabolicﬁngerprintingofmicewithdiet-induced metabolicderangementsbyparalleldualsecondarycolumn-dualdetection two-dimensionalcomprehensivegaschromatography,J.Chromatogr.A1361 (2014)265–276.

[26]P.Q.Tranchida,S.Salivo,I.Bonaccorsi,A.Rotondo,P.Dugo,L.Mondello, Analysisoftheunsaponifiablefractionoflipidsbelongingtovarious milk-typesbyusingcomprehensivetwo-dimensionalgaschromatography withdualmassspectrometry/flameionizationdetectionandwiththesupport ofhighresolutiontime-of-flightmass,J.Chromatogr.A1313(2013)194–201.

[27]J.Krupˇcík,R.Gorovenko,I. ˇSpánik,P.Sandra,D.W.Armstrong,

Flow-modulatedcomprehensivetwo-dimensionalgaschromatographywith simultaneousﬂameionizationandquadrupolemassspectrometricdetection, J.Chromatogr.A1280(2013)104–111.

[28]D.Turner,G.H.Morgan,B.White,Amatteroftaste....ﬂavourproﬁlingby GCxGC-qMS/FID,Chromatogr.Today(2012)40–43,

flavourprofilingbygcxgc-qmsfid/1136/.(Accessedonlineon12/04/2018)

http://www.chromatographytoday.com/articles/gc-mdgc-gc-ms/32/diane turnerdrg.h.morganbryanwhite/amatteroftaste.

[29]P.-H.Stefanuto,K.A.Perrault,R.M.Lloyd,B.Stuart,T.Rai,S.L.Forbes,J.-F. Focant,Exploringnewdimensionsincadavericdecompositionodour analysis,Anal.Methods7(2015)2287–2294.

[30]Y.Duan,F.Zheng,H.Chen,M.Huang,J.Xie,F.Chen,B.Sun,Analysisof volatilesindezhoubraisedchickenbycomprehensivetwo-dimensionalgas chromatography/highresolution-timeofﬂightmassspectrometry,LWT– FoodSci.Technol.60(2015)1235–1242.

[31]J.D.Byer,K.Siek,K.Jobst,DistinguishingtheC3vsSH4masssplitby comprehensivetwo-dimensionalgaschromatography-highresolution time-of-ﬂightmassspectrometry,Anal.Chem.88(2016)6101–6104.

[32]D.D.Yan,Y.F.Wong,L.Tedone,R.A.Shellie,P.J.Marriott,S.P.Whittock,A. Koutoulis,Chemotypingofnewhop(HumuluslupulusL.)genotypesusing comprehensivetwo-dimensionalgaschromatographywithquadrupole accuratemasstime-of-ﬂightmassspectrometry,J.Chromatogr.A1536 (2017)110–121.

[33]J.Gee,K.Prasad,S.Slayford,A.Gray,K.Nother,A.Cunningham,E.

Mavropoulou,C.Proctor,AssessmentoftobaccoheatingproductTHP1.0.Part 8:studytodeterminepufﬁngtopography,mouthlevelexposureand consumptionamongJapaneseusers,Regul.Toxicol.Pharmacol.93(2018) 84–91.

[34]ISO,3308,RoutineAnalyticalSmokingMachine-DeﬁnitionandStandard Conditions,InternationalOrganisationforStandardisation,Geneva,2000.

[35]T.Kind,O.Fiehn,Metabolomicdatabaseannotationsviaqueryofelemental compositions:Massaccuracyisinsufﬁcientevenatlessthan1ppm,BMC Bioinform.7(2006)1–10.

[36]LinearRetentionIndicesdatabaseonlines:https://www.ncbi.nlm.nih.gov/ pccompound?cmd=search(Accessed20March2018).

[37]P.Ausloos,C.L.Clifton,S.G.Lias,A.I.Mikaya,S.E.Stein,D.V.Tchekhovskoi,O.D. Sparkman,V.Zaikin,D.Zhu,Thecriticalevaluationofacomprehensivemass spectrallibrary,J.Am.Soc.MassSpectrom.10(1999)287–299.

[38]A.Giri,M.Coutriade,A.Racaud,K.Okuda,J.Dane,R.B.Cody,J.-F.Focant, Molecularcharacterizationofvolatilesandpetrochemicalbaseoilsby photo-ionizationGC×GC-TOF-MS,Anal.Chem.89(2017)5395–5403.

[39]E.Rathahao-Paris,S.Alves,C.Junot,J.-C.Tabet,Highresolution massspectrometryforstructuralidentiﬁcationofmetabolitesin metabolomics,Metabolomics12(2016)10.

[40]J.Dallüge,L.L.P.vanStee,X.Xu,J.Williams,J.Beens,R.J.J.Vreuls,U.A.Th. Brinkman,Unravellingthecompositionofverycomplexsamplesby comprehensivegaschromatographycoupledtotime-of-ﬂightmass spectrometry,J.Chromatogr.A974(2002)169–184.