Ultrahigh-performance supercritical fluid chromatography – mass spectrometry for the qualitative analysis of metabolites covering a large polarity range

The applicability of ultrahigh-performance supercritical fluid chromatography coupled with mass spectrometry (UHPSFC/MS) for the qualitative analysis of metabolites with a wide polarity range (log P: −3.89–18.95) was evaluated using a representative set of 78 standards belonging to nucleosides, biogenic amines, carbohydrates, amino acids, and lipids.

journalhomepage:www.elsevier.com/locate/chroma

Department of Analytical Chemistry, Faculty of Chemical Technology, University of Pardubice, Studentská 573, Pardubice 53210, Czech Republic

a r t i c l e i n f o

Article history:

Received 9 December 2021

Revised 13 January 2022

Accepted 13 January 2022

Available online 15 January 2022

Keywords:

Supercritical fluid chromatography

Metabolites

Amino acids

Human plasma

Mass spectrometry

a b s t r a c t

Theapplicability ofultrahigh-performancesupercriticalfluidchromatographycoupledwithmass spec-trometry (UHPSFC/MS) for the qualitative analysis of metabolites with a wide polarityrange (log P:

−3.89–18.95)wasevaluatedusingarepresentativesetof78standardsbelongingtonucleosides,biogenic amines,carbohydrates,aminoacids,andlipids.Theeffectsofthegradientshapeandthepercentageof water(1,2,and5%)wereinvestigatedontheViridisBEHcolumn.Thescreeningofeightstationaryphases wasperformedforcolumnswithdifferentinteractionsites,suchashydrogenbonding,hydrophobic,π

-π,oranionicexchangetypeinteractions.Thehighestnumberofcompounds(67)ofthesetstudiedwas detectedontheTorusDiolcolumn,whichprovidedaresolutionparameterof39.TheDEAcolumnhad thesecondbestperformancewith58detectedstandardsandtheresolutionparameterof54.The over-allperformanceofotherparameters,suchasselectivity,peakheight,peakarea,retentiontimestability, asymmetryfactor,andmassaccuracy,ledtotheselectionoftheDiolcolumnforthefinalmethod.The comparisonofadditivesshowedthatammoniumacetategaveasuperiorsensitivityoverammonium for-mate.Moreover,theinfluenceoftheionsource ontheionizationefficiency wasstudiedbyemploying atmosphericpressurechemicalionization(APCI)andelectrosprayionization(ESI).Theresultsprovedthe complementarityofbothionizationtechniques,butalsothesuperiorionizationcapacityoftheESIsource

inthenegativeionmode,forwhich53%oftheanalytesweredetectedcomparedtoonly7%fortheAPCI source.Finally,optimized analyticalconditionswereappliedtotheanalysisofapooledhumanplasma sample.44compoundsfromthepreselectedsetweredetectedinhumanplasmausingESI-UHPSFC/MS

inMSEmodeconsideringbothionizationmodes

© 2022TheAuthors.PublishedbyElsevierB.V ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/)

1 Introduction

Metabolomicscanbe describedasthecomprehensivestudyof

smallmolecules,calledmetabolites,intheorganismandthe

asso-ciationofthose withpathophysiologicalstates [1,2].Metabolomic

analysisrequirestheuseofhighlypowerfulanalytical techniques,

such asmass spectrometry (MS) hyphenated to chromatographic

separation techniques, such as liquid chromatography (LC) or

gas chromatography, asa means to simultaneously analyze

com-plex mixtures of metabolites [3] The most widespread

separa-tion modes used for metabolomicsare reversed-phase

ultrahigh-performance liquid chromatography (RP-UHPLC) and hydrophilic

interactionliquidchromatography(HILIC-UHPLC)coupledto

high-∗ Corresponding author

E-mail address: denise.wolrab@upce.cz (D Wolrab)

resolution (HR) MS instruments, such as quadrupole-time-of-flight (QTOF) or Orbitrap [4] The comprehensive analysis of the metabolomeischallengingduetothehighchemicalandstructural diversityofmetabolites InRP-UHPLC,intermolecularhydrophobic interactionsbetweenanalytes,stationaryphase,andmobile phase allow analysis ofa large part of the metabolome of complex bi-ological samples such as urine, plasma, and tissue extracts [5– 7].However, polarand/or ionicspeciesarepoorly retainedin RP-UHPLC[8].TheretentionmechanismintheHILICmodeisbasedon the interactionof polaranalyteswith thepolar stationaryphase, whichallowstheseparationoftheanalytes Therefore,HILIC pro-videscomplementarychromatographicseparationcomparedto RP-UHPLC/MS [9,10] Nonpolar compounds may elute in or close to the voidvolume inHILIC mode.A comprehensivetechniquethat allowsthe separationofpolarandnon-polarmetabolites, suchas lipids,aminoacids,andnucleosides,isdesired

https://doi.org/10.1016/j.chroma.2022.462832

0021-9673/© 2022 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

instruments allow stableandreproducibleresults aswell as

rou-tine hyphenationto massspectrometry[11,12].Mainly, the

back-pressureregulator,injector,andcolumntechnologywereimproved

for thenew generationinstruments, leadingto better acceptance

ofUHPSFC/MS forabroadapplicationrange.Generally,theuseof

supercritical fluid chromatography(SFC) wasfirstdescribedmore

than50yearsago[13],butitsapplicationforlipidomic[14,15]and

metabolomic analysis [12,16,17] represents a rather recent trend

UHPSFC may represent a potential alternative to RP- and

HILIC-UHPLC forthe comprehensiveanalysisof themetabolome by

re-ducing the costs and analysis time Nowadays, UHPSFC mainly

usessupercriticalCO2mixedwithorganicmodifiersasthemobile

phase.The additionoforganicsolvents,typically 2–40%,broadens

the range of analytes that can be separated withUHPSFC Polar

solvents, such as methanol,ethanol, or acetonitrile,are the most

commonlyusedandfacilitatetheelutionofpolarcompounds.The

additionofsmallpercentagesofwater,salts,bases,and/oracid

ad-ditivestothemodifiercanfurtherimprovethepeakshapeandthe

elution of polar andionic compounds [18–20] The low viscosity

and highdiffusionof themobile phase in UHPSFCallow the use

of highflow rateswithout losingseparationefficiencyand

there-fore allow high-throughput analyzes [21,22] Furthermore,almost

all stationary phases used for UHPLC can also be used for

UH-PSFC including modern stationary phasespacked with sub-3 μm

core shell andfully porous particles [23–26] Recently, dedicated

UHPSFC stationary phases were also introduced to the market,

such as the Torus column series from Waters These stationary

phasesare basedon silicamodifiedwithdifferentselectors, such

as propanediol,1-aminoanthracene, diethylamine, 2-picolylamine,

or ethylene-bridged silica, which are potentially suitable for the

separationofmedium topolarmetabolites[18,21,27].Itshouldbe

mentionedthattheanalysisofverypolarcompoundsstillremains

challenging for UHPSFC/MS employingcommon chromatographic

conditions The increase inthe percentageof modifierin CO2 up

to100%duringthegradientincreasestheelutionstrengthfor

po-lar compounds and extends the polarity range of analytes

suit-able for theseparation withUHPSFC/MS instruments These

spe-cialmodesusingincreasedmodifier andCO2 asthemobilephase

arecalledunifiedchromatographyorenhancedfluid

chromatogra-phy[20,28].However,suchmobilephaseconditionsandtheuseof

sub-2 μmparticlecolumnscanlead toexceedingsystempressure

forcing adjustmentofparameters, suchasbackpressure,

tempera-ture,orflowrate.Consequently,theanalyzesmaybeperformedby

increasingtheorganicsolventinthemobilephasegradientand,at

thesametime, decreasingtheflowrates.Followingthisapproach

in metabolomics and using a polar stationary phase, the elution

rangesfromnonpolartopolarcompounds[12]

Inrecentyears,moreUHPSFC/MSapplicationshavebeen

inves-tigated, accompanied by unconventionaland innovative

develop-ments regardingtheapplied conditionsandinstrumentalsettings

Analysis ofnaturalproducts, biological samples[29–32],

pharma-ceuticals, nutrients, and environmental samplesare examples for

the application areasof UHPSFC[33].However, onlya few

stud-iesinvestigatedmetabolomicsbyUHPSFC/MS.Thepotentialof

UH-PSFC/MSfortheanalysisofpolarurinarymetaboliteswas

investi-gatedbySenetal.,whoevaluated12differentcolumns,3column

temperatures, and9 differentadditives inmethanol, forthe

sep-aration of60 polar metabolites (log P −7 to 2) [18] Desfontaine

et al described the application of UHPSFC coupled to a

triple-quadrupole MSfortheanalysisofnucleosides,smallbases,lipids,

small organic acids, andsulfated metabolites [12] The analytical

methodwasoptimizedbyinvestigatingseveralparameterssuchas

the kinetic performance, the percentageofcosolvent, the type of

stationary phase, and the composition of the mobile phase

Ad-ditionally, the mixture of 57 compounds was also analyzed by

unifiedchromatographycoupledwithMS Losacco etal.analyzed

49metabolitesinplasmaandurine usingUHPSFC/QTOF-MSwith theevaluationoftheimpact ofthebiologicalmatrix.Mostof se-lectedcompounds werenotaffectedbymatrixinterference(63%), whereby 16% ofcompounds showeda matrix effectinurine and plasmasamples[16].TheUHPSFC/MS analysisoffreeaminoacids wasinvestigatedbyRaimbaultetal.[20].Theseparationof18 na-tive proteinogenic amino acids wasachieved by applying unified chromatographicconditions,startingfrom90%CO2 to100% modi-fier

The aim of this work was to evaluate the suitability of UH-PSFC/MSfortheanalysisof78metabolitesselectedfromthe Hu-manMetabolome Database (HMDB)database based on their rel-evance incancer research Toachieve the elution ofall analytes, enhanced fluid chromatography wasapplied because the analyte set covers a wide polarity range (log P: −3.89 – 18.95) The in-fluence of the percentage of waterin the modifier, the gradient shape,andthe type ofstationaryphase fortheseparation ofthe analyteset wasevaluatedusingUHPSFC/QTOF-MS Theionization efficiencyoftheselected metabolitesemployingelectrospray ion-ization(ESI)andatmosphericpressure chemical ionization(APCI) wascompared.TheMSEmode wasappliedfortheanalysisofthe standardmetabolitemixtureandplasmasamples

2 Materials and methods

2.1 Chemicals

Methanol (CH3OH), acetonitrile (ACN), 2-isopropanol, hexane (all LC/MS gradient grade), water (H2O; LC/MS Ultra, UHPLC/MS grade),andchloroform (LCgrade,stabilizedwith0.5–1% ethanol) were purchased from Honeywell (Charlotte, North Caroline, US) Ammonium acetate, ammonium formate (LC/MS, gradient grade), andformicacid(98–100%,Suprapur)werepurchasedfrom Sigma-AldrichorMerck(St.Louis,MO,U.S.A),respectively.Carbondioxide (CO2)4.5grade(99.995%)wasobtainedfromMesserGroupGmbH (BadSoden,Germany)

2.2 Standards

The standards were purchased from Sigma-Aldrich (see Table1).Standardstocksolutionswerepreparedbydissolvingeach compoundintheappropriatesolventorsolventmixture(TableS1)

toobtainthefinalconcentration of10mgmL−1.Afterwards,nine standardmixtureswerepreparedaccordingtotheanalytecategory anddilutedinMeOH,thatis,mixturesofnucleosides(N°1;100ng

μL−1), biogenic amines (N°2; 20–100 ng μL−1), sugars together withotherorganiccompounds(N°3;100–10ngμL−1),aminoacids (N°4–8;2–100 ngμL−1) andlipids (N°9;10ngμL−1) Analytesin eachmixtureare reportedinTableS1.Standardconcentrationsin the final mixture were established by investigatingthe efficiency and sensitivity of ionization using direct infusion MS for 10 ng

μL−1 and100ngμL−1 aswellastwodifferentsamplesolvent so-lutions,MeOH andMeOH/ACN(50:50,v/v) Theoptimized source parameters,standardconcentration,andsamplesolventwereused forfurtherexperiments.Thesestandardmixtures(100μLofeach) andguanine (10 μL) were mixed, andthe final standard solution wasdilutedwithacetonitriletoobtainafinalsolventcomposition

ofCH3OH/ACN(50:50,v/v)forUHPSFC/MSanalyzes(TableS1)

2.3 Stationary phases

The final standard solution was analyzed using eight differ-ent columns,reported inTable 2 The stationaryphasesdiffer in their column chemistry, providing different interaction sites and

Table 1

List of standard compounds

Nucleosides

Uridine

C 9 H 12 N 2 O 6 244.0695 −2.28

Biogenic amines

N-acetyl-5-hydroxytryptamine C 12 H 14 N 2 O 2 218.1055 0.44

Amino acids

L-Histidine monohydrochloride C 6 H 9 N 3 O 2 155.0695 −1.67

Acetyl- l -carnitine hydrochloride C 9 H 17 NO 4 203.1157 −0.66 D,L-Octanoylcarnitine chloride C 15 H 29 NO 4 287.2096 1.68 D,L-Decanoylcarnitine chloride C 17 H 33 NO 4 315.2409 2.46

D,L-Myristoylcarnitine chloride C 21 H 41 NO 4 371.3035 4.02

D,L-Palmitoylcarnitine chloride

Sugars

Other organic compounds

Polar lipids

( continued on next page )

Table 1 ( continued )

Mono-Sulfo-GalCer(d18:1/24:1) C 48 H 91 NO 11 S 889.6313 13.94

Nonpolar lipids

Table 2

Columns screened in this study

ACQUITY UPC 2 Torus Diol Fully porous hybrid silica Propanediol 100 × 3.0 1.7

ACQUITY UPC 2 Torus 2-PIC Fully porous hybrid silica 2-Picolyl-amine 100 × 3.0 1.7

ACQUITY UPC 2 Torus 1AA Fully porous hybrid silica 1-Amino-anthracene 100 × 3.0 1.7

ACQUITY UPC 2 Torus DEA Fully porous hybrid silica Diethylamine 100 × 3.0 1.7

ACQUITY UPC 2 HSS C18 SB Fully porous silica Octadecyl, nonendcapped 100 × 3.0 1.8

therefore may show different selectivities The Viridis BEH

col-umn(100× 3.0mmI.D,1.7μm)wasselectedforthepreliminary

study In addition, the Torus columns, namely, Diol, 2-PIC, 1-AA,

andDEA(100× 3.0mmI.D,1.7μm),wereevaluated.Thedioland

BEH columns representthe most polar stationary phases among

the selected ones, characterized by propandiol bonded silica

support andfree silanols, respectively.Furthermore, twocolumns

packedwithmodifiedC18silicawereincludedforcolumn

screen-ing,namely,HSSC18SBcolumn(100× 3.0mm,1.8μm)andHSS

T3(100× 2.1I.D;1.8μm).Additionally,thechiralstationaryphase

Trefoil CEL1(150 × 3.0mm I.D;2.5μm) wastested Allcolumns

werepurchasedfromWaters(Milford,MA,USA)

2.4 UHPSFC/MS/MS instrumentation

UHPSFC/MS/MS analysis was performed on the Acquity UPC2

(Waters, Milford, MA, USA) hyphenated with the Synapt G2-Si

(Waters) QTOF mass spectrometer The UHPSFC instrument was

coupled to the MS using the commercial interface kit (Waters)

Gradient mode was used for screening the stationary phase

se-lected forthe separation of the metabolite mixture Supercritical

carbondioxide(scCO2)wasusedasthemobilephaseA,andMeOH

with30 mmolL−1 ammonium acetate and1,2,or 5% ofH2Oor

MeOHwith30mMammoniumformateand2%ofH2Owas

inves-tigatedasmobilephaseB(modifier).Thegradientstartedwith5%

ofB,wasincreasedto70%Bin8.5min,thento100%Bin10.5min,

kept constant for2.5min,andfinally returnedto theinitial

con-dition within 1 min and re-equilibrated for 1 min, with a total

runtimeof15min.Furthermore,aflowgradientwasemployedto

avoidinstrumentoverpressureat100%ofthemodifier;thestarting

flow wasset to2.0mLmin−1,decreasedto0.8mLmin−1 within

14 min,and backto the initial flowin 1min The ABPRwas set

at 1800 psi and the column temperatureat 60 °C The injection

volumewas1μL,andtheinjectionneedlewaswashedaftereach

injectionwithhexane/2-isopropanol/H2O(2:2:1,v/v/v).MeOHwith

0.1% of formic acid and5% of H2Owas used as a make-up

sol-vent to improve the ionization Furthermore, a flow gradient for

themake-up solventwasused: 0min– 0.6mLmin−1,8.5min–

0.2mL min−1,13 min – 0.2mL min−1, 14min – 0.6 mLmin−1,

15min– 0.6mLmin−1 Thefollowing settingswere usedforQTOFmeasurements:HR mode,amassrangeofm/z50–950,andthecontinuummodewith

ascantimeof0.1 wereapplied.Leucineenkephalinwasusedas thelock mass,inwhich thelock masswasacquired withascan timeof0.1 in10 intervals,butnoautomaticlockmass correc-tionwasapplied.ESIandAPCIinpositiveandnegativeionmodes wereinvestigated.ThefollowingparameterswereusedfortheESI mode:capillaryvoltage2.50kV,samplingcone20V,sourceoffset

90V,sourcetemperature150°C,desolvation temperature350°C, conegasflow50L/h,desolvationgasflow1000L/handnebulizer gasflow3.5bar.ThefollowingparameterswereusedfortheAPCI mode: corona current1.0μA, samplingcone 10V, conegas flow

50L/h,nebulizergasflow3.5bar,sourceoffset50V,source tem-perature 150°C, probe temperature600 °C andlockspray capil-lary voltage 3.0kV Column screeningwas performedin positive andnegativeionmodeusingESIandfullscanspectraacquisition Furthermore,theMSEmodewasappliedtodetecttheMSspectra andthecorrespondingfragmentspectraofeachcompoundinone run The MSE methodis characterized by two stages In stage 1, allions aretransmittedfromtheionsourcethroughthe collision cell,whereinlowcollisionenergyisappliedsothatno fragmenta-tioncan beobservedinthemass analyzer,andions arerecorded

astheprecursorions.Instage2,allionsaretransmittedfromthe ionsourcethrough thecollision cell,andacollision energyramp

is applied togenerate andrecord fragment ionsin the mass an-alyzer Hence,thesoftware isabletogeneratetwo spectraatthe sametime; thefirst one showsthe precursorions withno colli-sion energy,and the second one generates fragment ionsdue to theappliedcollisionenergy.Thetrapandtransfercollisionenergy

ofthe lowenergyfunction waskept off andtheramp trap colli-sionenergyofthehighenergyfunctionwassetfrom5to30V.– Theobtainedfragmentswerecomparedwithonlinedatabases,i.e., HMDBandMzCloudforfurtherconfirmation

2.5 Biological sample preparation

Humanplasmacollectedfromdifferenthealthyvolunteerswas

pooled, worked up, and analyzed with UHPSFC/MS under

opti-mized conditions All subjects signed an informed consent An

additional step, namely, limiteddigestion with proteinase K was

added.Beforeproteinprecipitation,2μLofProteinaseKand2μL

of250mM CaCl2 were addedto100 μLofplasmasampleto

ob-tain a final concentration of 5 mM and sonicated for15 min at

40°C.Forproteinprecipitation,1mLofCH3OH/EtOH(1:1,v/v)was

addedtothepooledplasmasamplesonicatedfor15minatroom

temperature(25°C)andstoredfor1hat−20°C.Thesamplewas

centrifuged for15minat10,000rpm,thesupernatantwas

trans-ferred to a glassvialandevaporated undernitrogen The residue

wasdissolvedwith35μLofACN/CH3OH/H2O(4:4:2,v/v/v)+0.1%

formicacidanddiluted1:10withthesamesolventmixture.1μL

wasinjectedforthesubsequentUHPSFC/MSanalysis

2.6 Data processing

Data were acquiredwiththeMassLynxsoftware(Waters).The

WatersCompressionToolwasusedfornoisereduction,whichalso

minimized the raw data file size facilitating data handling The

Accurate MassMeasure tool in MassLynxwasused to apply lock

masscorrectionforbettermassaccuracyandfortheconversionof

datafromcontinuumtocentroidmode,whichfurtherreducedthe

file size Finally, TargetLynx was used to extract retention times,

peak areas,peak height, peak widths(Pkwidth), andasymmetry

factors by providing exact masses and expected retention times

ofall compounds.The resultingtableswere exportedas.csv files

and further processed with Microsoft Excel, i.e., to calculate the

numberofidentifiedstandardsandtherelativestandarddeviation

(RSD%)oftheretentiontime,peakareaandpeakheight

MarkerL-ynxwasusedtogeneratefeaturelistsofm/zwiththe

correspond-ing retentiontime,whichallowed thecalculationofthemass

ac-curacy The complete summary tablesfromMarkerLynx were

ex-portedas.csvfilesandmanually checkedforexperimentalm/z of

eachstandardortargetanalyte

Furthermore,MZmine2.53software[34] softwarewasusedto

assess the influence of the data processing procedure The

fol-lowing settings were applied: the targeted peak detection

mod-ule was used to search the list of compounds with a precursor

mass tolerance of 0.002 m/z or5 ppm and a retention time

tol-eranceof0.1min.Peakintegrationwascheckedandmanually

cor-rected when needed Retention times, peak areas, peak heights,

peak widths (FWHM), asymmetry factors, and experimental m/z

were exported as.csv filesand furtherprocessed with Microsoft

Excel For thechromatographic evaluation, theresolution and

se-lectivity (α)were calculated.Theresultsoftwodifferentsoftware

solutionswerecompared(Tables3andS11)

MSDIAL ver 4.70, was used for metabolite identification in

realhumanplasmausingthedatabasesMSMS-Public-pos-VS15for

positive and MSMS-Public-Neg-VS15 for negative ion mode The

databasesarecomposedofmetabolitesinplasma,whichwere

de-tected bytheMSDIALcommunity,whereby13,303entriesare

in-cluded for positiveion mode and12,879entries for negativeion

mode

3 Results and discussion

3.1 Selection of standard compounds

Theselectedanalytes(Tables1andS1)wereselectedfromthe

Human MetabolomeDatabase based ontheir biological relevance

with a special focus on the metabolites involved in cancer

pro-gression [12,16] The final metabolite mixture varied in

molecu-Fig 1 Relation of partition coefficients (log 10 P) and molecular weights for the in- vestigated analytes and the number of metabolites categorized by compound class: nucleosides (red), biogenic amine (yellow), amino acids (green), sugars (orange), others (dark green), polar lipids (light blue), and nonpolar lipids (blue)

lar weights from 89 to 900 Da and log10P values from −3.89

to 18.95 (Fig 1) The substantial variety in the structural com-position andchemical characteristics ofselected compounds, i.e., polar amino acids to hydrophobic lipids such as triacylglycerols, leadstohighlydiverseretentiontime behavioranddifferent ion-ization efficiencies, altering sensitivity The list includes metabo-litesinvolved invariousbiological pathways,e.g.,metabolites de-rived from the tryptophan pathway Tryptophan is an essential aminoacid,abuildingblockforproteinbiosynthesisandfunctions

asaprecursor forthe conversiontoseveralother metabolites in-cludedinourlist,i.e.,5-hydroxytryptophan,tryptamine,serotonin, melatonin,N-acetyl-5-hydroxytryptamine, l-kynurenine,l-alanine, andglutamic acid Furthermore,clinical studies have shownthat tryptophanmetabolismpromotestumorprogressionthrough mul-tiple mechanisms [35], andits metabolic derivative l-kynurenine

is involved in Alzheimer’s disease and the early stages of Hunt-ington’s disease The catecholamines dopamine, adrenaline, and noradrenaline are derived from the tyrosine pathway [36] with

an implication inthe treatment ofdopamine-responsive dystonia and Parkinson’s disease In general, biogenic amines and amino acids were chosen for their importance in several types of can-cer,namely,ovarian,breast,pancreatic,colon,andoralcancers,and neurodegenerativediseases.Similarly,sugarswereincludedinthis optimizationdue totheir large consumption by tumor cells[37] Other two important biological classes of compounds are nucle-osides andlipids, forwhich evident changes havebeen observed

in cancer patients [38,39] The involvement of lipids in different typesof tumorssuch aspancreatic,gastric, liver,lung, colorectal, andthyroidcancer wasshown[40] Aschematicoverviewofthe biosynthesisreactionsispresentedinthesupplementary informa-tion(Figs.S1andS2), clearlyillustratinghowthevariousanalytes areinterrelated.Intotal64fromthe78analytesshown,the miss-ing14analytesmainlyincludemetabolites,whichare uptakenby dietarymeanssuchasessentialaminoacids(6),caffeineandfolic acid,andconsequentlynobiosynthesiscanbeshown.The remain-inganalyteswereincludedintheanalytesetformechanistic ques-tions,i.e.,isomers.Theimportanceandconnectionofaminoacids forthe biosynthesisof biogenic amines,glucose aswell as lipids canbe seen.Further,itiscommonlyassumedthat the dysregula-tionorabsenceofsomemetabolitesmayharmthewell-being

Tofacilitatethe elutionofnon-polar, polar,andionic metabo-lites,the addition ofmethanol, includingadditives,to scCO2 was necessary.Asmallamountofwaterwasaddedtothemobilephase

to improve the peak shape and solvation ofpolar analytes As a consequenceof thelimitedmaximum upperpressure ofthe UH-PSFC system,it wasnot possibletomaintainhighflow ratesand reach 100% of the modifier for the elution of polar compounds

Table 3

Summary tables of mass accuracy, selectivity, resolution, and the repeatability of the retention time, signal area, and signal height calculated by the average RSD% for ammonium acetate and ammonium formate in positive and negative ionization mode using TargetLynx and MZmine as data processing software

Repeatability (RSD%)

MZmine

Repeatability (RSD%)

at thesame time Therefore,a decreasing flow rategradient was

applied simultaneouslywiththeorganicmodifiergradient,

allow-ingustoreach100%ofthemodifier.Thisapproach,calledunified

chromatography, wasintroduced by Chester [28] Theadjustment

oftheeluentstrengthofthemodifierupto100%enabledtowiden

thepolarityrangeoftheanalytesetsuitableforUHPSFC/MS

mea-surements

3.2 Screening of water percentage and gradient evaluation with BEH

column

The chromatographic performance of the Viridis BEH column

was evaluated forthree differentpercentages (1%, 2% and5%) of

waterinthemethanolic modifier,whilethe concentrationof

am-moniumacetatewaskeptconstantat30mmolL−1.Theadditionof

2–7%ofwatertothemodifier[12,16,20,31]tofacilitatetheelution

of polarcompounds andtoimprovepeak shape,probablycaused

bytheimprovedsolubility,iscommonlyreportedintherecent

lit-erature Six consecutive injections of the standard mixture were

performedinpositiveandnegativeionmodestotesttheinfluence

of1,2,and5%H2Ointhemodifieronthechromatographic

perfor-mance Retentiontimesof78selectedmetabolitesarereportedin

Table S2.Fig.2showstheoverlayofchromatogramsobtainedfor

threedifferentamountsofwaterinthemodifierforvarious

com-pounds Generally, the retention time increased with increasing

amountofwaterinthemodifier.Forsomeanalytes(Fig.2A,C),the

peak shape worsened,andpeak tailing occurredusing5% of

wa-ter inthemodifier.Furthermore,agradualincrease inthesystem

pressure was observed using 5% of waterin the modifier, which

regularly causedoverpressure of the instrument The experiment

was repeatedafter severalweeksto ensure that this observation

wasnotan artefact.Thesametrendofthegradualincreaseofthe

systempressurewasobserved,whichcouldbecausedbysolubility

issuesoftheadditiveinthemobilephaseleadingtoprecipitation

and columnblockage.A goodcompromise was obtainedwith2%

or1% ofwaterinthemodifier; all compoundswere eluted

with-out anyimpairmentoftheGaussianpeakshape(Fig.2).However,

nonpolar triacylglycerols and diacylglycerols were eluted closeto

thevoidvolumewith1%ofwaterinthemodifier.Consequently,2%

ofwaterinthemodifierwasassessedasthemostsuitableamount

of waterin the modifier to achieve the bestbalance in terms of

retentiontimeandpeakshape

The nextstep ofthestudywasto improvethegradientshape

toobtainagoodseparationoftheentirestandardmixtureaswell

as good peak shapes Two gradients were evaluated, gradient A

(usedfortheevaluationofthepercentageofwaterinthemodifier)

startingfrom5% modifierto75% in10min,andgradientB start-ingfrom5% modifierto 70%modifier in8.5min.Analyteseluted faster andshowed better peak shapeswith gradient B (Fig S3B) compared to gradient A (Fig S3A) Consequently, gradient B was furtherusedforcolumnscreening

3.3 Column screening and performance evaluation

The choice ofstationaryphase chemistry, columndimensions, and mobile phase composition is crucial for successful separa-tion Subsequently, various stationary phases were evaluated for the separation performance of the standard metabolite mixture, such asDiol, 2-PIC, 1-AA, DEA, BEH,CEL 1,HSSC18 SBandHSS T3(Table2).AllcolumnsareclassifiedasUHPSFCcolumns(except HSST3), areproduced by the samemanufacturer (Waters, Torus, andViridiscolumns)andmosthadthesamesub-2μmparticle di-mensionsaswellascolumnlength anddiameter,forbetter com-parability (100 × 3.0 mm I.D, 1.7μm, fully porous hybridsilica) [18,21,31]

The eight stationary phases screened were dedicated UHPSFC columnsfromthesamemanufacturerwithsub-2μmparticles.The majorityofthestationaryphasesarecomposedofthesame back-bone (bridged ethylene hybrid particles) ensuring that the non-selective interactions are comparable and the different selectors bonded on the silica support cause differencesin the chromato-graphic performance The different selectorsbonded to the silica supportallowdifferentselectivitiesfortheseparationofthe stud-iedmetabolitesasaresultofthedifferentinteractioncapabilities

of the analyte andthe stationary phase The simplest stationary phaseregardingtheselectorstructureistheBEHcolumnwithfree silanolsonthesurface,allowingH-bonding andhydrophilic inter-actions.FortheDiolcolumn, propandiolislinkedtothemodified silicasupport.Consequently,thehydroxylgroupsallowH-bonding andhydrophilic interactions, but the hydrocarbon chains provide hydrophobicinteractionsaswell.ThesilicaparticlesofTorus2-PIC, Torus1-AA,andTorusDEA aremodifiedwith2-picolyl-amine, 1-amino-anthracene,anddiethylamine,respectively.Thesestructures allowmultipleinteractionsoftheselectorwiththeanalyte,suchas stericinteractions, hydrogenbonding, Vander Waalsinteractions, dipole-dipoleinteractions,anionic exchangetype, orπ-π interac-tions.ThecolumnsHSSC18SB(100× 3mmI.D;1.8μm)andHSS T3(100× 2.1mmI.D;1.8μm)columnsarepackedwithsilica par-ticlesmodifiedwithoctadecylbondedligandson thesurface, en-ablinghydrophobicinteractions.Thecolumnsdifferintheir resid-ualactivityofthesilanolgroup,asHSSC18T3isend-capped com-paredtoHSSC18 SB.Thefree residualsilanolgroupsadditionally

Fig 2 Effect of water percentage (1% - green, 2% - red, and 5% - blue) in the mobile phase on the retention behavior of selected metabolites: A) l -tryptophan, B) N-acetyl-

5-OH-tryptamine (I °) and serotonin (II °), C) d -glucose (I °) and myo-inositol (II °), and D) LPC (18:1) (I °) and PC (36:2) (II °) Analytical conditions: BEH (100 × 3.0 mm, 1.7 μm) column; 60 °C; 1800 psi (ABPR); mobile phase: CH 3 OH + 30 mmol L −1 ammonium acetate, and 1%, 2%, and 5%; composition of the make-up solvent: CH 3 OH + 0.1% formic acid and 5% of H 2 O

allowhydrophilicinteractionsinthecaseoftheHSSC18SB,which

maybeadvantageous fortheanalysisofpolarcompounds.Trefoil

CEL1 (150× 3.0 mm I.D; 2.5μm)is astationaryphase basedon

polysaccharides, inwhichthesilica gelismodifiedwithcellulose

tris-(3,5-dimethylphenylcarbamate), allowingmultiple interactions

suchasstericinteractions,hydrogenbonding,π-πinteractions

Different chromatographic parameters were evaluated to

de-termine the best column for the separation of the analyte

mix-ture, such as the number of compounds not detected and the

peak asymmetry factor (As), which is calculated as the ratio of

the peak widthinthe backhalf andthe peakwidthin thefront

halfat10% ofthepeak height.Forbettercomparability,the same

gradient was applied forthe separation of78 metabolites forall

stationary phases investigated (Fig S3B) The retention times of

each standard for each tested column are reported in Table S2

Fig 3A shows the chromatogramsof guanine and guanosine

de-pending on the employed stationary phase The highest number

of compounds detected, depending on the stationary phase, was

as follows: Diol > BEH > 2-PIC > HSS C18 SB > DEA > Cel

1 > 1-AA > HSS T3 (Fig 3B) This indicates that with

increas-ing hydrophobicity and bulkiness of the stationary phase

selec-tor, lessanalytesare detected However, some hydrophobic

inter-actions favor separationand detectionincomparison to only

hy-drophilicinteractions, asreflected forthe DiolandBEH columns

Eachcompoundwasinjectedseparatelyaswellasinamixturefor

each column Therefore, thecompounds not detected inthe

ana-lytemixturearebelowthedetectionsensitivitybecausetheywere

identifiedwheninjectedseparately,someathigherconcentrations

Thisproves that theanalytesareelutedforeach column,but

be-causeofthebroadpeakshapeandhighasymmetry,theywere

be-low thedetectionsensitivity,notallowingtheir identification.The

mostdifficultcompoundstoidentifyformostofthecolumnswere

metabolites with primary amines in the structure The primary

amines may undergo ionic interactions with the free silanols of

thestationaryphase.Asionicinteractionsaregenerallyslow,broad

peakshapescanbeobserved,whichmayleadtosensitivityissues

Theenhancementofthecationconcentrationinthemobilephase

could lead to an improvement of the peak shape and sensitivity

sincecations functionasdisplacers.Onthe otherhand,increased baseconcentrationsinthemobilephasemayleadtoion suppres-sion The detection sensitivity was diminished especially for the metabolites PS, LPS, PG, LPG, PA and LPA, putrescine, spermine, spermidine,anddopamine,and5-methoxyindole,5-hydroxyindole, D,L-dylurenine,andfolicacid

The performance of the columnwas generally considered ac-ceptablewhentheAsvaluewasintherangeof0.9–1.5.Therefore, foreach column, the percentageof compounds not detected, the analyteswithAs below0.9andwithAsgreaterthan1.5were cal-culated(Fig.S4).Thehighestpercentageofsymmetricalpeakswas observed forBEH> DEA > Diol> Cel 1> 1-AA > 2-PIC > HSS C18>HSST3inpositiveionmodeandBEH>DEA>2-PIC>HSS T3 > Cel 1 > Diol > 1-AA > HSS C18 in negative ion mode (Table S3) Furthermore, the asymmetry values of each detected compoundandtheirtotalaverage fortheeight screenedcolumns are reported in Table S3 Broad chromatographic peaksand tail-ingwerefound forthelipidclassesPS,LPS,PG,LPG,PA,andLPA,

as also known from the literature For biogenic amines, namely spermine,spermidine,andputrescine,abroadanddistortedpeak shapewasobserved.TheresultssuggestthattheDiolcolumn rep-resentsagoodcompromisebetweenthenumberofdetectedpeaks andtheasymmetryvaluescomparedtotheothersevencolumns The performance of the method was further investigated

by determining the mass accuracy, selectivity, resolution, peak area, peak height, retention time stability, and number of total compounds detected in both polarity modes (Tables S4–S9) L tryptophan wasselected as the reference compound forthe cal-culationofresolutionandselectivity,asoneofthelasteluting an-alytes The average time of the first peak in the run from three consecutive blanks (solution of CH3OH/CHCl3, 1:1 v/v) was con-sideredasthe voidtime needed tocalculate the capacityfactors (TableS5) Themedianofselectivityandtheaverageofall deter-mined resolutionvalues foreach columnwere investigated.They were determined from the average values of six consecutive in-jectionsofthestandard mixtureforeach columnbyapplyingthe optimized conditions The median and the average of the over-all mass accuracy, selectivity, and resolution are reported in

Ta-Fig 3 (A) Selected chromatograms for guanine (I °) and guanosine (II °) on the various stationary phases (red: 1-AA, green: BEH, light blue: Diol, violet: 2-PIC, dark blue:

DEA) (B) Bar chart for the number of detected compounds (blue) and non-detected compounds (red) for individual screened columns (C) Median of the selectivity values and (D) Average of the resolution values for all metabolites on the various stationary phases Analytical conditions: mobile phase: CH 3 OH + 30 mmol L −1 ammonium acetate and 2% of water; mobile phase of the make-up pump: CH 3 OH + 0.1% formic acid and 5% of H 2 O; 60 °C, 1800 psi (ABPR), ESI ( + ) and ESI ( −)

bles S4–S6, as well as the average and RSD% of the peak area,

peak height, andretentiontime inTables S7–S9.The median

ob-tainedforallanalyteswasusedtocomparetheoverallselectivity

of the stationary phases, since the overall average may be

influ-enced by analytes eluted close to the void volume The highest

median selectivity was observed for DEA > Cel 1 > 2-PIC =

1-AA > BEH > HSS T3> Diol > HSSC18 in positive ion (Fig 3C)

mode and BEH > 1-AA > 2-PIC = HSS T3 = HSS C18 > Cel

1>Diol>DEAinnegativeionmode(TableS5).Itshouldbe

men-tionedthatmuchfewercompoundsweredetectedanddifferences

in themedianselectivity are negligiblein thenegativeionmode

compared tothepositive ionmode(TablesS2 andS5).The

high-est average resolutionwasobserved forDEA > Diol> BEH >

2-PIC>1-AA>Cel1>HSSC18>HSST3inpositive(Fig.3D)and

DEA > Diol > BEH > 2-PIC > 1-AA > HSSC18 > Cel 1 > HSS

T3 in negative ion mode (Table S6) A small shift in mass

pre-cision wasobserved, whichcorresponds to the retentiontime of

the analyte (Table S4) The ionization efficiency depends on the

gradient shape ofthechromatographic runand, consequently, on

the retention times ofthe analytes Furthermore, the type of

in-teractionsoftheanalyteswiththestationaryphaseinfluencesthe

peak shape, because ionic interactions are slow andmaylead to

broaderpeaksincomparisontofasterinteractionssuchas

interac-tions basedon partitionorsolubility.The peakarea, peakheight,

and averageretention time together withthe retentiontime

sta-bility were investigated The highest average peak area was

ob-served forHSST3>HSSC18>Diol>BEH>Cel1>1-AA>

2-PIC > DEA in positive ion mode and HSS T3 > HSS C18 > Cel

1> > BEH>2-PIC >Diol>1-AA> DEAinnegativeionmode

(Table S7).Thehighestaveragepeakheight wasobservedforHSS

T3 > HSSC18 > 2-PIC > Diol> 1-AA> BEH > Cel1 > DEA in

positive ionmode andDiol> DEA >2-PIC > BEH> 1-AA>Cel

1 > HSSC18> HSST3in negativeionmode (Table S8) The

av-erageretentiontimesonthedifferentstationaryphasesshowthe

distribution of analyteswithin the chromatographic run andthe

extent and type of interactions of the analyte with the selector

The highestaverage retentiontime was observedfor BEH> HSS C18 > Diol > 1-AA > DEA > 2-PIC > Cel 1 > HSS T3 in pos-itive ion mode and BEH > DEA > Diol > 1-AA > 2-PIC > HSS C18>Cel1>HSST3innegativeionmode.Therelativestandard deviationof the retention timesof the analyte for6 consecutive injectionsofthemetabolitemixtureoneachstationaryphasewas investigated, describing the reproducibility of the retention time ThehigheststabilityofretentiontimewasobservedforDiol>Cel

1>DEA> 1-AA> 2-PIC> BEH> HSSC18 >HSST3inpositive ionmodeandDEA >Diol>1-AA>2-PIC> BEH>Cel1> HSS C18>HSST3innegativeionmode

TheDiolcolumndidnotprovidethebestresultsforeach eval-uated parameter,butthe comparisonof thechromatographic pa-rametersmentionedforeachcompound andstationaryphase, in-cluding the total numberof detected peaksin positive and neg-ative ion modes, reveals that the overall best performance was achievedwiththeDiolcolumn,asalsopreviouslyreportedfor uri-narymetabolites[18].67ofthe78structuralandchemicalhighly diversecompounds inthemixtureweredetectedontheDiol col-umn In order to investigate the reason for detected and non-detected compounds dependingon theanalyte structure, the an-alyte set was classified accordingto their functional groups (Ta-ble S14) However, no general trend depending on the presence

offunctionalgroupswasobserved.TheDiolcolumnwasusedfor furtherevaluation withinthis study.The putative explanationfor whichDiolworkedwellfortheseparationofmostoftheanalytes

isthe possiblepolarandhydrophobic interactions of the station-ary phase withthe polar andhydrophobic parts of the analytes Further,thesmallselectorstructureoftheDiolcolumnmayfavor theaccessibilityoftheanalytestointeractwiththeselectorofthe stationaryphase,incontrast tothebiggerselectorstested, which maybeleadingtosterichindrance

Thereliable separationofisomeric and/or isobaricmetabolites

incomplexbiologicalsamplesisimportantinmetabolomics stud-ies Examples of isomeric metabolites included in our selected standardmixtureareleucine,isoleucine,andnorleucine,some

as well asN-acetyl-serotonin and1-methyl-tryptophan.Examples

of isobaric metabolitesinvestigated are asparagine andornithine,

asparticacidand5–hydroxy-indole,glutamine,andlysine,aswell

as glutamic acid and 5–methoxy-indole Only the BEH and the

Diolcolumnyieldedapartial separationofleucineandisoleucine

with respect to norleucine, while their coelution with all other

columns was detected Dopamine and octopamine are also

im-portant examples of isomers The first was below its detection

sensitivity, which did not allow detection in most cases; on the

other hand,octopamine waseasily detected However,each

stan-dard wasinjected individually,allowing goodseparation between

thesetwocompounds.Infact,octopaminewaselutedoneach

col-umnbetween5 and7min,while dopamine waseluted between

8 and 9 min on various stationary phases Finally, the isomers

d-glucose and myo-inositol,as well as N-acetyl-serotonin and

1-methyl-tryptophan, were always separated independently of the

stationary phase All detected isobaric metabolite pairs were as

well separatedonallstationaryphasesinvestigated.Theseresults

have shownthatthe optimizedmethodprovides goodseparation

ofnotonlyverydiversemetabolitesbutalsosomeoftheirisomers

andisobars

18 of the 78 compounds investigated in our study were also

included in the analyte set investigated by Losacco et al of 49

compounds[16]andDesfontaineetal.of57compounds[12],such

assome aminoacids,biogenic amines,nucleosides andlipids.For

comparison reasons, special focus was placed on the evaluation

of theseparation performance ofthose compounds Agood peak

shape and peakasymmetry havebeenreached applyingthe Diol

column, i.e., foradenosine,leucine,andsphingomyelin (1.61,1.45,

and 1.31,respectively; TableS3) Additionally,the% RSD of

reten-tiontimestabilitywasdeterminedforeachcolumn(TableS9).The

overall stabilityofthe retentiontime was0.31% RSDforthe Diol

columnand0.06,0.41and0.07%foradenosine,leucineand

sphin-gomyelin,respectively.Inconclusion,thereportedmethodyielded

comparableresultsfortheDiolcolumnincomparisontothedata

shownbyLosaccoetal.usingthePoroshellHILICcolumn[16].Itis

importanttoemphasize,thatthecolumnscreeningwasperformed

forUHPSFC/MSdedicatedsub2-μmcolumnsfromthesame

man-ufacturer and the larger set ofanalytes in terms of polarity and

mass rangeinthepresentwork.The metabolitesinvestigatedare

mainly interrelated (Figs.S1 andS2), besideschosen analytes

in-cludedtostudymechanisticaspects.Thisenabledacompleteand

exhaustiveanalysisofchromatographicandmassspectrometric

pa-rameters

3.4 Evaluation of ammonium acetate versus ammonium formate as

an additive in the modifier

The influence of the type of additive in the mobile phase

on retention time, peak area, peak height, mass accuracy,

selec-tivity, resolution, and peak asymmetry for the standard mixture

was investigated using the Diol column Six consecutive

injec-tionswereperformedusing30mmolL−1ofammoniumacetatein

CH3OH/H2O(98:2,v/v)or30mmolL−1 ofammoniumformate in

CH3OH/H2O(98:2,v/v)asa modifier.The peakareasandheights

of each detected compound were normalized to the average

in-tensity ofthe lockmass todiminish theinfluence ofthe drift of

the instrumental response over time (Table S10) The processed

dataofthenormalizedareaandnormalizedheightwerecompared

using bar graphs for the positive (Figs S5 andS6) and negative

(Figs S7 andS8) ionization mode Signal responses forall

com-pounds were higher for ammonium acetate compared to

ammo-nium formate (Table S10) As a result, a highernumber of

com-poundsweredetectedwithammoniumacetate(67)thanwith

am-monium formate (65) Table 3 shows a summary of the average

Fig 4 Base peak intensity chromatograms of the standard set of metabolites ob-

tained under optimized conditions (black) and reconstructed ion current chro- matograms for selected compounds: caffeine (red), MG (0:0/18:1/0:0) (blue), Cer (d36:2) (olive), sphinganine (d18:0) (orange), melatonin (wine), adenine (magenta), adenosine (violet), N-acetyl-5-OH-tryptamine (royal), LPC (18:1/0:0) (cyan), palmi- toylcarnitine (dark yellow), acetylcarnitine (dark cyan), taurine (pink), l -tyrosine (light magenta), l -tryptophan (dark gray), 5-OH- l -tryptophan (light orange), and l - arginine (light blue) Analytical conditions: Diol (100 × 3.0 mm; 1.7 μm); mobile phase: CH 3 OH + 30 mmol L −1 ammonium acetate and 2% of water; composition of the make-up solvent: CH 3 OH + 0.1% formic acid and 5% of H 2 O; 60 °C, 1800 psi (ABPR), ESI ( + )

mass accuracy, selectivity, resolution and RSD of the peak area, peak height, andretentiontime Furthermore,detailed valuesfor each compound, depending on the additive applied on the Diol column,arereportedinTablesS3–S10 forammoniumacetateand Tables S10–S11 for ammonium formate The average mass accu-racy,selectivity,andresolutionwasslightlyhigherforammonium formate than ammoniumacetate (Table3) No generaltrendwas observedforthesignal andretentiontimestabilitydependingon theadditive.However,ammoniumacetatewasselectedasthe ad-ditive of choice in the mobile phase for the investigated analyte setbecauseoftheslightlyhighernumberofdetected compounds and the higher signal response Data processing was performed independently with TargetLynx, which was used by default, and MZmine, to assess whether the data processing software has an impactonthe results(TableS11) Thesamechromatographicand method parameters were investigated with MZmine aswith Tar-getLynx,and comparedto each other Data were comparablebut notthesame,whichshowsthatthedataprocessingsoftware em-ployedmayhaveanimpactontheresults

Finally,theDiolcolumn(100× 3mmI.D;1.7μm),themodifier

ofMeOHwith30mmolL−1 ammoniumacetate and2%ofwater, themake-upsolventofMeOHwith0.1%formicacidand5%of wa-ter(seeFig.4)wereevaluatedasthebestchoicefortheseparation

oftheinvestigatedanalyteset

3.5 Comparison of ESI and APCI ionization techniques

Theionization efficiencymaychangedependingonthe chem-icalpropertiesandchemicalstructureoftheanalyzedcompounds andtheappliedionsource.ESIisthemostwidelyusedionsource Sensitivitystronglydependsontheflowratesemployed,sinceESI representsaconcentration-dependentionizationtechnique.APCIis

amassflowdependentionizationprocessmoresuitableforhigher flow rates UHPSFC/MS methods generally use flow rates higher thanthoseofUHPLC/MSmethods;therefore,theevaluationofthe ionsource on the ionization efficiencyof target compounds may

be ofinterest.However, themajorityof UHPSFC/MSmethods use

a splitter,whichreduces theflow intothemass spectrometer

fa-voring ESI.A systematic investigation wasconducted to evaluate

the influenceofthe ionizationsourceonthe numberandtype of

detected analytes Thestandard mixturewasanalyzedby ESIand

APCIinbothpolaritymodes.Theoptimizedchromatographic

con-ditionsandoptimizedionsourceparameterswereapplied.The

re-sults showed that ESI in general led to a higher ionization

effi-ciency compared to APCI(Figs S9–S12).However, forsome

ana-lytes,thepeakarea andpeak height(normalizedtothesumarea

and sumheight considering the total compounds in the positive

andnegativeionization mode fortheESIandAPCIsources)were

higher forthe APCIsource than forthe ESIsource, showingthat

ESI andAPCI can be complementary (Table S12) The sensitivity

was higherforseveralamino acids,such asl-tyrosine,ornithine,

phenylalanine, taurine, as well as l-tryptophan, l-arginine, and

l-lysine and two nucleosides (adenine and guanine) using APCI,

and dopamine wasonly detected using the APCIsource On the

other hand, the areas and heights of the l-carnitine derivatives,

LPC(18:0),andmelatoninwereenhancedwithESI.l-carnitineand

acetyl-l-carnitine were only detected using ESI To illustrate the

comparisonofthenormalizedareaandheightforsomeidentified

standardsinpositiveionmodeforbothionsources,bargraphsare

showninFigs.S9–S12.Thecorrespondingvaluesofthenormalized

peak areas,heights,andretentiontimesofeach standardforboth

ionsourcesarereportedinTableS12.Thetotalnumberofdetected

compoundswas67and48forESIandAPCI,respectively,showing

thewider applicationrangeofESIfortheinvestigatedanalyteset

[32].Inthenegativeionizationmode,mostanalyteswere not

de-tectedusingAPCI(Figs.S11andS12).Furthermore,inthenegative

ionization mode, thesignal responsewassignificantlylower than

inthepositiveionization mode,regardlessoftheionsourcetype

ESI provided theoverall best ionization efficiencyforthe analyte

setinbothionmodes

3.6 Application to human plasma

Optimized chromatographic and MS conditions were applied fortheanalysisofpooledhumanplasmasamplestoevaluate the applicability of the method to real samples The protocol used for sample preparation was based on the application of an ad-ditionalstep prior to protein precipitationbased on the addition

ofproteinase K.Thisprocedure allowed the releaseof associated metabolitesthrough relaxationof the tertiary structure of native proteins andconsequently a higherpossibility oftheir identifica-tion [41] In addition,plasma samples obtained by the following protein precipitation were injected and analyzed in MSE mode

MSE modeallows theuntargeted scanningof theMS andMS/MS levels by applyinglow andhigh collisionenergy within one run Thisincreasesthe identificationconfidence ofmetabolites, asthe characteristicfragmentsofmetabolitesprovideadditional informa-tion.First,thestandardmixturewasanalyzedtoobtainclean frag-mentionspectraasreferencewithoutinterferencescausedbythe complexmatrixofarealhumanplasmasampleusingESIandAPCI

No differencesinfragmentationbehavior were observedbetween ESIandAPCI(TableS13).Thedilutedhumanplasmasample(1:10) wasanalyzedusingESIandAPCI.Fig.5AshowstheTICofhuman plasmaobtainedwithESIandAPCI Itcanbeseenthatthe sensi-tivityishigherforESIthanAPCI,alsoforrealhumanplasma sam-ples The extractedionchromatograms (XIC)of selected metabo-litesdetectedinhumanplasmaarepresentedinFig.5BusingESI Thetargeteddata analysisrevealed that44 and5 compounds in-cludedintheanalytesetwerealsodetectedinthedilutedplasma sampleusing ESI andAPCI, respectively, in positive andnegative ionmode(Table S15).The reductionofthesamplecomplexityby optimizingthesamplepreparationprotocol,i.e.,usingsolid phase extraction,mayhelpincreasing thesensitivitytodetectthewhole analyte set in human plasma The untargeted MSE approach al-lowedthe use ofmetabolomicsdatabases to link m/z features to

Fig 5 (A) Impact of the type of ion source on sensitivity Base peak intensity chromatogram of human plasma using red) ESI and blue) APCI (B) Selected extracted ion

chromatograms of human plasma using ESI (green: ornithine, blue: glucose, red: serotonine, black: adenosine) Pie charts of the untargeted m/z feature analysis in human plasma for (C) positive and (D) negative ion mode using MS DIAL 5823 m/z features were detected in positive and 2769 m/z features in negative ion mode The m/z features were categorized according to the compound class: red) analyte set of the study, green) nucleoside and derivatives, orange) amino acids and derivatives, violet) polyphenols, light blue) lipids and dark blue) other metabolites Analytical conditions: Diol (100 × 3.0 mm; 1.7 μm); mobile phase: CH 3 OH + 30 mmol L −1 ammonium acetate and 2% of water; composition of the make-up solvent: CH 3 OH + 0.1% formic acid and 5% of H 2 O; 60 °C, 1800 psi (ABPR), ESI ( + )