of the multiple diagnosis sclerosis in A validated method UHPLC-MS for tryptophan metabolites:Application Journal of Pharmaceutical and Biomedical Analysis

(1)

ContentslistsavailableatScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

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

A validated UHPLC-MS method for tryptophan metabolites:

Application in the diagnosis of multiple sclerosis

Ferenc Tömösi

^a

, Gábor Kecskeméti

^a

, Edina Katalin Cseh

^b

, Elza Szabó

^b

, Cecília Rajda

^b

, Róbert Kormány

^c

, Zoltán Szabó

^a

, László Vécsei

^b,d

, Tamás Janáky

^a,∗

aDepartmentofMedicalChemistry,InterdisciplinaryCentreofExcellence,UniversityofSzeged,Dómtér8,H-6720,Szeged,Hungary

bDepartmentofNeurology,InterdisciplinaryCentreofExcellence,UniversityofSzeged,Semmelweisu.6,H-6725,Szeged,Hungary

cEgisPharmaceuticalsPlc.,Keresztúriút30-38,H-1106,Budapest,Hungary

dHungarianAcademyofSciences,MTA-SZTENeuroscienceResearchGroup,Semmelweisu.6,H-6725,Szeged,Hungary

a r t i c l e i n f o

Articlehistory:

Received12December2019 Receivedinrevisedform6March2020 Accepted7March2020

Availableonline9March2020

Keywords:

Tryptophanmetabolism Derivatization

Liquidchromatography-massspectrometry DryLab^®4

Validation Multiplesclerosis

a b s t r a c t

Thesimultaneousquantitativeestimationoftryptophan(TRP)anditsmetabolitesrepresentsagreat challengebecauseoftheirdiversechemicalproperties,e.g.,presenceofacidic,basic,andnonpolar functionalgroupsandtheirimmenselydifferentconcentrationsinbiologicalmatrices.Ashortultra high-performanceliquidchromatography(UHPLC)–tandemmassspectrometry(MS/MS)methodwas validatedfortargetedanalysisofTRPandits11mostimportantmetabolitesderivedviabothkynure- nine(KYN)and serotonin (SERO) pathwaysin human serumand cerebrospinalﬂuid (CSF):SERO, KYN,3-hydroxyanthranilicacid,5-hydroxyindoleaceticacid,anthranilicacid,kynurenicacid(KYNA),3- hydroxykynurenine(3-HK),xanthurenicacid,melatonin,picolinicacid(PICA),andquinolinicacid(QUIN).

Afterselectingthe“best”reversed-phasecolumnandorganicmodifier,DryLab^®4wasusedtooptimize thegradienttimeandtemperatureinchromatographicseparation.Toachieveabsolutequantification, deuterium-labeledinternalstandardswereused.Amongallcompounds,3wereanalyzedinderivatized (butylester)forms(3-HK,PICA,andQUIN)andtheremaining9inunderivatizedforms.Validationwas performedinaccordancewiththeICHandFDAguidelinestodeterminetheintradayandinterdaypreci- sion,accuracy,sensitivity,andrecovery.TodemonstratetheapplicabilityofthedevelopedUHPLC–MS/MS method,theaforementionedmetaboliteswereanalyzedinserumandCSFsamplesfrompatientswith multiplesclerosis(multiplesclerosisgroup)andthosewithsymptomaticornoninflammatoryneurolog- icaldiseases(controlgroup).TheconcentrationofQUINdramaticallyincreased,whereasthatofKYNA slightlydecreasedinthemultiplesclerosisgroup,resultinginasignificantlyincreasedQUIN/KYNAratio andsignificantlydecreasedPICA/QUINratio.

Abbreviations: 3-HANA, 3-hydroxyanthranilic acid; 3-HK, 3- hydroxykynurenine; 5-HIAA, 5-hydroxyindoleacetic acid; aCSF, artificial cerebrospinal fluid;AGC, automatic gain control; ANA, anthranilicacid; CSF, cerebrospinalfluid;EDSS,expandeddisabilitystatusscale;FA,formicacid;FD, fluorescencedetector;GC,gaschromatography;HPLC,high-performanceliquid chromatography;HQC,high-levelqualitycontrol;KAT,kynurenineaminotransferase;KP,kynureninepathway;KYN,l-kynurenine;KYNA,kynurenicacid;LOD, limitofdetection;LOQ,limitofquantification;LQC,low-levelqualitycontrol;MELA, melatonin;MQC,medium-levelqualitycontrol;MS/MS,tandemmassspectrometry;PFP,pentafluorophenyl;PICA,picolinicacid;PRM,parallelreactionmonitoring;

QC,qualitycontrol;QUIN,quinolinicacid;RPC,reversed-phasechromatography;

RRMS,relapsing-remittingmultiplesclerosis; RSD,relativestandard deviation;

SERO,serotonin;SIL-IS,stableisotope-labeledinternalstandard;TRP,tryptophan;

UHPLC,ultrahigh-performanceliquidchromatography;XA,xanthurenicacid.

∗Correspondingauthor.

E-mailaddress:janaky.tamas@med.u-szeged.hu(T.Janáky).

1. Introduction

The kynurenine(KYN) pathway (KP) is themajor metabolic pathway of the essential amino acid tryptophan (TRP), which leadstotheproductionofneuroprotectiveandneurotoxiccom- pounds.Changesintheconcentrationsofneuroprotectantssuch as kynurenic acid (KYNA) and picolinic acid (PICA) have been describedindifferentdiseases,buttheexcitotoxinquinolinicacid (QUIN)andthefreeradicalgenerator3-hydroxykynurenine(3-HK) areassociatedwithdifferentneurodegenerativediseasesincluding multiplesclerosis[1].KYNAisanendogenousglutamatereceptor antagonistthataffectsallionotropicglutamatereceptorsincluding NMDA, kainate, and AMPA receptors, and it exhibits the high- estafﬁnityfortheNMDAreceptoranddisplaysantioxidantand freeradical-scavengingactivities[2].Conversely,QUINisknown

(2)

foritsbroad-spectrumneurotoxiceffectsincludingitsroleasan NMDA receptor agonist, in addition to its roles in antioxidant depletion,lipidperoxidation,andoxygenintermediategeneration [1].

Multiplesclerosis is a disabling autoimmune, inﬂammatory, neurodegenerative, demyelinating disease affecting the central nervoussystem.Thediagnosisofmultiplesclerosisisbasedonclin- icalsymptomssuggestiveofdemyelinatingdisordersasprovenby MRIandlumbarpuncture.Inadditiontodisseminationofthedis- easeinspaceand time,itisessentialtoexcludeotherdiseases.

AccordingtoMcDonald’scriteria,asrevisedin2017,oligoclonal gammopathyincerebrospinalﬂuid(CSF)orthesimultaneouspres- enceofgadolinium-enhancingandnonenhancinglesionscanverify disseminationintime.Disseminationinspacecanbeprovenby thepresenceofoneormoreT2lesionsintheperiventricular,corti- cal,juxtacortical,orinfratentorialspaceorspinalcord[3].Multiple sclerosisisconsideredactiveifthefollowingﬁndingsarepresent:

relapse,conﬁrmedexpandeddisabilitystatusscale(EDSS)progres- sion,new/enlargedMRIlesions,anddisease-relatedbrainatrophy [4].

Recently, multiple sclerosis subtypes were differentiated according the concentrations of the different TRP metabolites [1,5] because changes in their concentrations appear to serve aspotentialbiomarkers. For example, inacute relapse of relapsing–remittingmultiplesclerosis(RRMS),increasedneuro- toxicQUINconcentration,higherQUIN/KYNandQUIN/KYNAratios [6],andlowerKYNAconcentrationcanbeobservedinCSF,whereas increasedKYNAconcentrationwerefoundinplasma.Theexpres- sionof KYNaminotransferase(KAT) alsoincreasedin redblood cells.DecreasedTRPconcentrationinbothserum andCSFwere describedinpatientswithRRMScomparedwiththoseincontrol subjects.Interestingly,duringremission,lowerKYNAconcentra- tioncanbeobservedinCSF[7],whereashigherKYNAconcentration canbefoundinpatientswiththeprogressive formofmultiple sclerosis[8].

The simultaneous quantitative estimation of TRP and its metabolitesrepresentsagreatchallengebecauseoftheirdiverse chemicalproperties,namely,presenceofacidic,basic,andnonpo- larfunctionalgroupsandtheirimmenselydifferentconcentrations inbiologicalmatricessuchasserum(plasma)andCSF.Numerous chromatographicmethodshavebeendevelopedforthesuccessful quantitativeestimationofthesemetabolitesusinggaschromatog- raphy(GC),high-performanceliquidchromatography(HPLC),ultra high-performanceliquid chromatography(UHPLC)coupledwith a ﬂuorescence detector (FD), a UV detector, or tandem mass spectrometry(MS/MS)[8–11].TheadvantagesofUHPLC–MS/MS includeitssensitivity,abilitytoquantitatemultiplecompoundsin asinglerun,shortelutiontime,excellentseparationefﬁciency,and requirementofsmallsampleamount.Hénykováetal.[9]developed aUHPLC–MS/MSmethodforthequantitativeestimation ofTRP, KYN,3-HK,KYNA,3-hydroxyanthranilicacid(3-HANA),anthranilic acid(ANA),serotonin(SERO),melatonin(MELA),andotherimpor- tantneuroactivemetabolitesderivedfromTRPinhumanCSFand serum.Fuertigetal.[10]describedthesuccessfulquantitativeesti- mationof13compoundsrelatedtoTRPinthebrainsandplasmaof miceandCSFandplasmaofnonhumanprimates.However,none ofthesestudiesconsideredtheseparationandpeakshapeofearly elutingcomponents.Becauseofthedifferentchemicalcharacteris- ticsofTRPanditsmetabolites,thecombinationofseveralmethods wasusedtomeasuretheirconcentrationsbyGuilleminetal.[8]

includingUHPLCcoupledwithadiodearraydetectorandanFD wereusedtomeasureTRP,KYN,3-HK,3-HANA,andANAconcen- trationsusingisocraticseparation.Inaddition,gradientelutionwas performedtomeasureKYNAconcentrationusinganFD,andGC coupledwithamassspectrometerwasusedtomeasurePICAand QUINconcentrations.

Themainpurposeofthecurrentstudywastodevelopanew, robust UHPLC–MS/MS method to quantify the concentrations of TRP and its 11 most important metabolites (Fig. 1), which were derived via both KP and SERO pathways, including the concentrationsoftherarelymeasuredmetabolitesPICAandQUIN.

Optimization of a chromatographic method can be supported andacceleratedusinginsilicosimulationsoftware[12]todeﬁne themostappropriateconditionsforrapid,sensitive,precise,and reproducibleanalysisas economicallyaspossible.Aftercolumn and organic modiﬁer scouting, the DryLab^®4 method develop- ment/optimizationsoftware[13]wasappliedtopredictretention andresolutionasfunctionsofgradienttime(tG)andtemperature (T).

Tothebestofourknowledge,thisistheﬁrstreportonthesimul- taneousquantitativecharacterizationofendogenousSERO,KYN, 3-HANA,TRP,5-hydroxyindoleaceticacid(5-HIAA),ANA,KYNA,3- HK,xanthurenicacid(XA),MELA,PICA,andQUINinserumandCSF ofpatientswithmultiplesclerosisusingUHPLC–MS/MS.

2. Materialsandmethods 2.1. Reagentsandchemicals

AllreagentsandchemicalswereofanalyticalorLC–MSgrade.

TRPanditsmetabolitesd4-PICAandn-butanolwerepurchased fromSigma-Aldrich(St.Louis,MO,USA).d3-3-HKwasobtained fromBuchemB.V.(Apeldoorn,TheNetherlands).Otherdeuterated internal standards (ISs; d4-SERO, d4-KYN, d3-3-HANA, d5-TRP, d5-5-HIAA,d5-KYNA,d4-XA,d4-MELA,andd3-QUIN)werepur- chasedfromTorontoResearchChemicals(Toronto,ON,Canada).

Acetonitrile(ACN),methanol(MeOH),water,HPLC-gradeammo- niumformate,andammoniumacetatewereobtainedfromVWR Chemicals(Monroeville,PA,USA).Formicacid(FA)waspurchased fromFisherScientiﬁc(Portsmouth,NH,USA)andacetylchloride fromAlfaAesar(Haverhill,MA,USA).

2.2. Preparationofstandard,IS,andqualitycontrol(QC)solutions Stocksolutionswerepreparedindividuallyataﬁnalconcentra- tionof1mg/mL,except3-HKandd3-3-HK(0.5mg/mL).According totheirstabilityandsolubility,severalsolventswereused.SERO, d4-SERO,TRP,d5-TRP,ANA,PICA,andd4-PICAweredissolvedin water–MeOH–FA–ascorbicacid(50:50:0.1:0.02,v/v/v/v).KYN,d4- KYN,5-HIAA,d5-5-HIAA,3-HK,d3-3-HK,MELA,d4-MELA,QUIN, andd3-QUINweredissolvedinMeOHcontaining0.1%(v/v)FAand 0.02%(v/v)ascorbicacid.3-HANA,d3-3-HANA,KYNA,d5-KYNA, XA,andd4-XAweredissolvedindimethylsulfoxide.Allstandard stocksolutionswerepreparedonice,dividedinto200-␮Laliquots, andstoredat−80^◦Cuntilfurtheruse.

Calibration standards were prepared at 12 levels and QC samples at three levels (low-level QC [LQC], eighth point of calibration;middle-level QC [MQC],fourth point ofcalibration;

and high-level QC [HQC], second point of calibration) in arti- ﬁcial CSF (aCSF; containing 127mM NaCl, 1.0mM KCl, 1.2mM KH₂PO₄,26mMNaHCO₃,10mMd-glucose,2.4mMCaCl₂,1.3mM MgCl₂,and 5.26␮Mbovineserumalbumin)[14] forCSFanaly- sis.Calibrationstandardscomprised250␮LaCSF,20␮Lstandard solutionmix(31.25–1000nMSERO,6.25–200nMKYN,0.25–8nM 3-HANA,250–8000nMTRP,12.5–400nM5-HIAA,0.6–20nMANA, 0.25–8nMKYNA,1.25–40nM3-HK,0.1–2nMXA,0.25–8nMMELA, 2–60nMPICA, and 5–160nM QUINin 0.1 %[v/v] aqueousFA), and910␮Lice-coldACNcontaining10␮Lstableisotope-labeled (SIL)-ISmix(800nMd4-SERO,20nMd4-KYN,3nMd3-3-HANA, 4000nM d5-TRP, 200nM d5-5-HIAA, 2nM d5-KYNA, 8nM d3- 3-HK,0.4nMd4-XA, 4nMd4-MELA,15nM d4-PICA,and 25nM d3-QUIN).

(3)

Fig.1.Simpliﬁedpathways oftryptophanmetabolism,indicating theprincipleenzymes,kynurenineaminotransferase(KAT), indoleamine-2,3-dioxygenase(IDO), tryptophan-2,3-dioxygenase(TDO),kynureninase(KYNU),andkynurenine-3-monooxygenase(KMO).

The calibration standards and QC samples for serum analysis were prepared in charcoal-stripped human serum, which waspreparedasdescribedbyMölleretal.[15]. Briefly,1.2gof charcoal-activatedpowder(FisherScientific)wasaddedto20mL serum,rotatedfor 2h, andcentrifugedat14,000 ×g for10min toobtainthesupernatant.Then,“blank”serumwasconfirmedby LC–MS/MStobefreeofTRPanditsmetabolites.Calibrationstan- dardscomprised100␮L“blank” serum,10␮Lstandardsolution mix[156.25–5000nMSERO,312.5–10000nMKYN,7.8–250nM3- HANA,6.25–200␮MTRP,7.8–250nM5-HIAA,6.25–200nMANA, 4.7–150nMKYNA,6.25–200nM3-HK,1.5–50nMXA,0.16–5nM MELA,3.125–100nMPICA,and62.5–2000nMQUINin0.1%(v/v) aqueousFA],and370␮Lice-coldacetone:MeOH(1:1,v/v)contain- ing10␮LSIL-ISmix(1500nMd4-SERO,1000nMd4-KYN,65nM d3-3-HANA,5250nMd5-TRP,200nMd5-5-HIAA,50nMd5-KYNA, 90nMd3-3-HK,25nMd4-XA,4nMd4-MELA,80nMd4-PICA,and 300nM d3-QUIN). Concentrations were selected in accordance withendogenous analyte concentrationsand toreach a proper signaltonoiseratio.

2.3. CollectionofhumanCSFandserumsamples

Thefemalesincludedinthisstudy,whounderwentbothlumbar punctureandbloodsamplecollection,wereenrolledattheDepart- mentofNeurology,UniversityofSzeged.Approvalforthehuman studywasgrantedbythelocalEthicalCommitteeoftheUniver- sityof Szeged(46/2014 and 143/2015), and thestudyprotocol adheredtothetenetsof themostrecentrevisionof theDecla-

Table1

Detaileddemographicandclinicaldataofthestudypopulation.

Subject Multiplesclerosisgroup Controlgroup Ageinyears(mean±SD) 33.84±9.14 37.57±10.09 Ageatonset(mean±SD) 32.47±9.00 –

EDSS(median,IQR) 0.5(0,1.5) –

Dataarepresentedasthemean±SDormedianandinterquartilerangeforEDSS.

EDSS,expandeddisabilitystatusscale;IQR,interquartilerange;SD,standarddevi- ation.

rationofHelsinkiforexperimentsinvolvinghumans.Allenrolled subjectsprovidedvoluntarysignedinformedconsentforparticipa- tion.Inclusioncriteriaforthemultiplesclerosisgroupwereolder than18years, femalesex,diagnosisofRRMS,clinicalfollow-up foratleast2years,andavailabilityofserumandCSFsamplesat thebiobankatthetimeofdiagnosticlumbarpuncture(n=20).The demographicandclinicaldataofthecohortarepresentedinTable1.

DisabilitywasquantitativelyestimatedusingEDSS.EDSSvalues rangedbetween0and5.5 (median=0.5).Age-matchedpatients (n=14)withsymptomaticneurologicaldisorder(e.g.,headache) ornoninﬂammatoryneurologicaldisease(e.g.,benignintracranial hypertension)whohadbeenfollowedupforatleast2yearsserved asthecontrolgroup.

CSFsampleswerecentrifugedimmediatelyafterlumbarpunc- tureat3500rpmfor10min,andaliquotsof500␮Lwerestored at−80^◦Cuntilfurtheruse.Whole-bloodsampleswerecollected in Vacutainertubes, centrifuged,and thesupernatant storedas mentionedpreviously.SerumandCSFsampleswerecollectedand

(4)

Fig.2.FlowchartofthesamplepreparationprocessofhumanCSFandserumsamples.

CSF,cerebrospinalﬂuid;FA,formicacid;ACN-SIL-IS,acetonitrilecontainingstableisotope-labeledinternalstandard;n-BuOH,n-butanol;UHPLC–MS/MS,ultrahigh- performanceliquidchromatography-tandemmassspectrometry.

storedaccordingtothestandardizedinternationalbiobankingcon- sensusprotocoloftheBIOMS-Eunetwork[16].

2.4. PreparationofhumanCSFandserumsamplesforanalysis PriortoproﬁlingtheKPandSEROpathways,sampleswererela- beled;therefore,ablindstudywasconducted.To250␮Lofeach CSFsample,20␮L0.1%(v/v)ofaqueousFAand910␮Lice-cold ACNcontaining10␮LSIL-ISmix(thesameasusedintheprepa- rationofthecalibrationstandards)wereadded,andthemixture wasvortexedfor30s.Itwasallowedtorestfor15minat−20^◦C, vortexedfor60stosupportproteinprecipitation,andincubated foranother15minat−20^◦C.Thesupernatantwasobtainedvia centrifugationofthemixturefor15minat12,000×gat4^◦C.The supernatant(1120␮L)wastransferredtoanewtube,centrifuged for15s,andsplitintotwoequalparts.Afterconcentrationunder vacuum(SavantSC110ASpeedVacPlus,Savant,USA),halfofthe samplewastreatedwith70␮L derivatizingreagent(n-butanol- acetylchloride,9:1,v/v)andincubatedfor1hat60^◦C.Themixture wasdriedundernitrogenbeforereconstitution.Bothpartsofthe sampleweredissolvedin100␮Lstarting eluent,vortexed,centrifuged,andcombined(Fig.2).

Serumsampleswerepreparedinthesamemanner,exceptthat to100␮Lofserumsampleweremixedwith10␮L0.1%(v/v)aque- ousFAand370␮Lice-cold acetone–MeOH(1:1,v/v)containing 10␮LSIL-ISmix,and400␮Lsupernatantwasprocessedfurther (Fig.2).

2.5. InstrumentationandUHPLC–MS/MSanalysis 2.5.1. DevelopmentandoptimizationofUHPLCseparation

UHPLCseparationofTRPand itsmetaboliteswasperformed on an ACQUITY I-Class UPLC^TM liquid chromatography system (Waters, Manchester, UK) comprising Binary Solvent Manager,

SampleManager-FL,andColumnManager.Thedwellvolumeof thesystemwas100␮L.

To select the most appropriate column for analyte separation,thefollowingreversed-phasecolumnsweretested:BEHC18 2.1×50mm(1.7␮m)andCortecsC182.1×50mm(1.6␮m)from Waters(Milford,MA,USA)andKinetexC182.1×100mm(2.6␮m);

KinetexEVOC182.1×150mm(5.0␮m);LunaC182.0×100mm (2.5␮m); Luna Omega Polar C18 2.1×50mm, (1.6␮m); Luna Omega PS C18 2.1×50mm (1.6␮m), Kinetex C8 2.1×150mm (2.6␮m), Jupiter Proteo C12 2.0×150mm (4.0␮m), and Luna Phenyl-Hexyl2.1×100mm(2.6␮m)fromPhenomenex(Torrance, CA,US).

ChromatogramsweredevelopedwithagenericMS-compatible gradientfrom2%to80%Bin7minusing0.1%FAassolventA and0.1%FAinMeOHassolventB.Similargradientelutionwas performedusingACNinsolventBasanorganicmodiﬁer.Foreach measurement,theﬂowratewassetat300␮L/minandTat25^◦C.

Intotal,20␮Lofastandardmixturecontaining2␮g/mLofeach analyte wasinjectedinto thecolumns.The UHPLCsystem was controlledusingMassLynx4.1SCN901(Waters).

Tooptimizeseparation onthe“best”column, a2Dretention modelwasbuiltonthebasisoffourchromatographicrunsafter injectionofamixtureof allanalytes.The t_G valuesweresetat 5and15mintolinearlychangetheeluentcompositionfrom10

%to90%MeOH containing0.1%FA.Chromatographywasper- formedat25^◦C and50^◦Cataﬂowrateof300␮L/min.Peaksin differentchromatogramswereidentiﬁedbydetectingthemolec- ularionsofeveryanalyteonamassspectrometer.Theresulting chromatographicdata(retentiontime,peakwidth,andarea)were enteredintoDryLab^®4(Molnár-Institute,Berlin,Germany)tocre- atea2Dresolutionmapdisplayingthecriticalresolutionsofthe peaksseparatedagainstt_GandT.

ChromatographicseparationforquantitativeestimationofTRP and its 11 metabolites in CSF and serum was performed at

(5)

25^◦Conapentaﬂuorophenyl(PFP)column(Phenomenex;100Å, 100mm×2.1mm,particlesize2.6␮m;“best”column)protected byaPFPguardcolumn(Phenomenex)using0.1%(v/v)aqueousFA assolventAandMeOHcontaining0.1%(v/v)FAassolventB.The ﬁnalgradientwassetasfollows:0.0min,10%B;1.0min,30%B;

3.0min,50%B;3.5min,90%B;5.0min,90%B;5.1min,10%B;

and7min10%B.Foreachmeasurement,theﬂowratewassetat 300␮L/min.SampleTwasmaintainedat5^◦C.Finally,20␮Lofthe samplewasinjectedintotheUHPLC–MS/MSsystem.

2.5.2. Optimizationofmassspectrometricanalysis

Allmass spectrometricmeasurementswereconductedusing theQExactive^TMPlusHybridQuadrupole-OrbitrapMassSpectrom- eter(ThermoFisherScientific,SanJose,CA,USA)connectedonline tothe UHPLCinstrument. The instrument was operatedin the positive-ionmodeusingtheequippedHESI-IIsourcewiththefol- lowingparameters:capillaryT,256^◦C;sprayvoltage,3.5kV;aux gasheaterT,406^◦C;sheathgasflow,48;auxgasflow,11;sweep gasflow,2;andS-lensRFlevel,50.0(sourceauto-defaults).Fullscan wasconductedwithamassrangeof50–300m/zandresolutionof 17,500.Theautomaticgaincontrol(AGC)settingwasdefinedas 3×10⁶charges,andthemaximuminjectiontimewassetto60ms.

For quantitative mass spectrometric analysis of TRP and its metabolitesusingMS/MS,theparallelreactionmonitoring(PRM) data acquisition modewas selected. Toreach the bestprecur- sor/producttransitionforquantitationandmaximizesensitivity, the optimalfragmentation conditions and collision energies of each analyte were identiﬁed. This optimization procedurewas performedforeachindividualstandardbydirectlyinfusingeach solutionintotheionsourceusingaHamiltonsyringeataﬂowrate of20␮L/min.

TheconcentrationsofTRPanditsmetabolitesinbiologicalsam- ples were measured by monitoringthe appropriate transitions (Table 2)determined previously.AGC target wassetat 5×10⁶ chargesfor TRPand d5-TRP becauseof theirhigherconcentra- tionsand2×10⁵chargesfortheremaininganalytes.Themaximum injectiontimewassetat60msandresolutionat17,500.Adivert valveplacedaftertheanalyticalcolumnwasprogrammedtoswitch ﬂowontothemassspectrometeronlywhenanalytesofinterest elutedfromthecolumn(1.4–5.0min)topreventexcessivecontam- inationoftheionsourceandionoptics.Thewashingprocedures oftheautosamplerbeforeandafterinjectingsampleswerepro- grammedtoavoidcarryoverofanalytes.

Controlof themassspectrometer,dataacquisition,anddata processing wasconductedusing Xcalibur^TM 4.1 (ThermoFisher Scientiﬁc).

2.6. Methodvalidation

BiologicalmatricesdevoidofTRPanditsmetabolitesarenot readily available; hence, surrogate matrices (aCSF and “blank”

serum)wereusedtodemonstratetheefﬁciencyofthemethod.In thisstudy,SIL-ISanalogswereusedtovalidatetheUHPLC–MS/MS method. For validation, the linearity, limits of detection (LOD) andquantiﬁcation(LOQ),precision,accuracy,andrecoverywere assessedfollowingtheICHandFDAguidelines[17,18].

2.6.1. Linearity,LOD,andLOQ

Thecalibrationcurvesofthe12analyteswereconstructedfrom thepeakarearatiosofthecompoundtoSIL-ISat11levelsusing mixedworkingstandardsolutions.Accordingtotheacceptancecri- teria,thecalibrationcurveshouldhaveacorrelationcoefﬁcient(r²) of0.99orbetter.TheLODandLOQofeachanalytewerecalculated basedonthestandarderroroftheintercept.LODandLOQwerecal- culatedusingtheformulas3.3×␣/Sand10×␣/S,respectively,

where␣isthestandarderrorofthey-interceptandSistheslope ofthecalibrationcurve(SupplementaryTables1–2).

2.6.2. Precisionandaccuracy

Theaccuracyandintradayandinterdayprecisionswerecalcu- latedbydeterminingﬁvereplicatesofLQC,MQC,andHQCsamples over3consecutivedays(n=45).TheconcentrationvaluesofQCs arepresentedinSupplementaryTables3–6.

Regardingprecision,theacceptancecriterionwas±15%with respecttorelativestandarddeviation(RSD).Regardingaccuracy, thelimitwasthesamerelativetodeﬁniteconcentrations.

2.6.3. Recovery

Todeterminerecovery,twobatcheswerepreparedasdescribed previously [19]. The ﬁrst batchwas spikedbeforeprotein precipitation, whereas the second batch was spiked after protein precipitation.Toachievereliableresults,ﬁveLQC,MQC,andHQC replicateswereprepared(seeChapter2.2).TherecoveryofTRPand itsmetaboliteswasestimatedbycomparingthecalculatedconcen- trationratiosofthetwobatchesatthreeQClevels(LQC,MQC,and HQC;n=45).

2.7. Statisticalanalysis

The calculation of peak arearatios and the calibration and quantitationofanalyteswereperformedfromcollectedrawdata usingXcalibur^TMQuanBrowser(ThermoFisherScientific).Thepro- cesseddataincludingpeakarea,peakarearatio,retentiontime, andconcentrationwereexportedintoMicrosoftExceltocreatean appropriatefileforinputintheRsoftware[20].Thenormalityof thevariableswascheckedusingtheKolmogorov–Smirnovtestand visuallycheckedusingquantile–quantileplots,andtheequalityof varianceswasexaminedusingWelch’sF-test.Outlierswereiden- tifiedusingGrubbs’stest.Tocomparedatabetweenthecontrol andmultiplesclerosisgroups,boxplotswithSDsweregenerated.

Comparisonsbetweenthetwogroupswereconductedusingan independentsamplest-testortwo-sampleWilcoxontestinR.A p-valueof<0.05wasconsideredstatisticallysigniﬁcant.

3. Resultsanddiscussion

3.1. Optimizationofsamplepreparation

TheanalysisofsmallmoleculesusingLC–MSusuallyrequires cleanupprocedurestoeliminatetheworstinterferingcompounds and to concentrate the sample if the analyte is present at extremelylowconcentrations.Biologicalsamplescontainproteins thatadverselyaffecttheanalysisofsmallmolecules.Toimprove sensitivity,proteinsinsampleshadtobeeliminated.Inourpre- liminaryexperiments,fourdifferentprecipitationsolvents were comparedwiththeir3×and5×volumesinhumanserumsamples [i.e., MeOH,ACN, acetone–MeOH(3:7, v/v),and acetone–MeOH (1:1,v/v)]andtwodifferentsolventsinCSFsamples(i.e.,MeOH andACN)tomonitortherecoveryofeachTRPmetabolite.3-HANA couldnotbedetectedusingMeOH(datanotshown).Thebestresult wasachievedusing3×volumeofACNtoprecipitatethemajority ofproteinsinCSFsamplesand3×volumeofacetone–MeOH(1:1, v/v)forserumsamples(seeChapter3.3.3).

Inacomplexquantitativebioanalyticalmethod,theuseofSIL-IS canhelpcontrolthevariabilityofthemethod.Becauseitispro- cessedalongwiththeanalyte,SIL-ISshouldbothhelpcorrectfor variabilityinsamplepreparationduringextractionandchemical derivatizationandcompensateforvariabilityinMSdetection.In thepresentstudy,11deuteratedequivalentsofTRPmetabolites wereaddedatthebeginningofanalysistocontroleverystepof theanalytical procedure. Unfortunately, thedeuteratedformof

(6)

Table2

Exactmassoftheprecursor,quantiﬁer,andqualiﬁerionsandtheoptimalcollisionenergiesoftheanalytesinthePRMmethod;theirexperimental(texp)andsimulated(tsim) retentiontimes,andtheirratio(texp/tsim)inoptimizedultrahigh-performanceliquidchromatographicseparation.

Analyte Precursorion[M+H]⁺ Quantiﬁerion[M+H]⁺ Qualiﬁerion[M+H]⁺ CE(eV) texp(min) tsim(min) texp/tsim

SERO 177.1022 115.0548 132.0810 32 1.53 1.50 1.02

d4-SERO 181.1273 119.0704 136.1060 32 1.53 1.50 1.02

KYN 209.0919 94.0660 146.0603 14 1.61 1.59 1.01

d4-KYN 213.1169 98.0912 150.0856 14 1.61 1.59 1.01

3-HANA 154.0499 136.0393 108.0453 10 2.26 2.24 1.01

d3-3-HANA 157.0688 139.0582 111.0812 10 2.26 2.24 1.01

TRP 205.0973 118.0652 188.0706 18 2.42 2.38 1.01

d5-TRP 210.1285 123.0968 193.1018 18 2.42 2.38 1.01

5-HIAA 192.0659 146.0600 117.0570 18 2.69 2.65 1.01

d5-5-HIAA 197.0972 150.0852 122.0685 18 2.69 2.65 1.01

ANA 138.0547 92.0503 120.0449 20 3.06 3.03 1.01

KYNA 190.0502 116.0503 162.0552 38 3.54 3.55 1.00

d5-KYNA 195.0806 121.0816 167.0867 38 3.54 3.55 1.00

XA 206.0450 178.0508 132.0443 21 3.90 3.91 1.00

d4-XA 210.0695 182.0761 136.0691 21 3.90 3.91 1.00

3-HK 281.1496 152.0709 110.0606 15 3.99 4.03 0.99

d3-3-HK 284.1684 155.0897 113.1082 15 3.99 4.03 0.99

MELA 233.1284 130.0654 115.0546 54 4.11 4.16 0.99

d4-MELA 237.1536 134.0907 119.0801 54 4.11 4.16 0.99

PICA 180.1019 96.0453 124.0396 28 4.38 4.43 0.99

d4-PICA 184.1270 100.0702 128.0647 28 4.38 4.43 0.99

QUIN 280.1543 96.0453 124.0396 37 4.63 4.61 1.00

d3-QUIN 283.1732 99.0641 128.0647 37 4.63 4.61 1.00

CE,collisionenergy;SERO,serotonin;KYN,l-kynurenine;3-HANA,3-hydroxyanthranilicacid;TRP,tryptophan;5-HIAA,5-hydroxyindoleaceticacid;ANA,anthranilicacid;

KYNA,kynurenicacid;XA,xanthurenicacid;3-HK,3-hydroxykynurenine;MELA,melatonin;PICA,picolinicacid;QUIN,quinolinicacid.

Fig.3. Extractedionchromatogramsoftheunderivatized(left)andbutyl-esteriﬁed(right)formsof3-hydroxykynurenine(3-HK,greenpeaks),picolinicacid(PICA,orange peaks),andquinolinicacid(QUIN,redpeaks).

ANAwasunavailable; therefore,for itsquantitative estimation, d5-KYNA,whichhasasimilarchromatographicelutiontime,was used.

SomeTRPmetabolites(3-HK,PICA,andQUIN)haveunfavor- ablechromatographicproperties,makingtheirquantitativeLC–MS estimationunreliable.Thechromatographicpropertiesofanalytes canbemodifiedviaderivatization.Changesinthestructureofan analyteusually affectitsphysicaland chemical properties(e.g., polarity,solubility,stability,andionizationefficiencyinmassspec- trometry),which can alter theseparation characteristicsof the analyte(e.g.,improvedpeakshape,elutiontime,peaksymmetry, efficiency,andplatecount)[21]. Ifthesechangesare favorable, theymightleadtoimprovementintheseparationoftheanalyte.

Regardingprecolumnderivatization,theselectedreactionmustbe quantitativeor,atleast,reproducibleandfreeofbyproducts.

Atthesametime,chemicalmodiﬁcationmayservetominimize matrixinterferencebymovingthecompoundtoapositiononthe chromatogramwhereinterferencewiththematrixcomponentis minimal.

ExceptSEROandMELA,TRPanditsmetaboliteshavecarboxyl group(s).Themostfrequentlyusedapproachforderivatizingcar-

boxylicacidsisesterificationwithashort-chainaliphaticalcoholin thepresenceofanacidasacatalyst[22].Althoughthisestablished esterificationmethod(Fischeresterification)leadstoequilibrium, thereactioncanbeshiftedtowardtheproductsbyapplyingexcess amountofalcohol.EsterificationofTRPanditsmetaboliteswas performedusingMeOH,ethanol,n-propanol,orn-butanol.Because theyhavethelongesthydrophobicaliphaticchains,thebutylated products exhibited the highest retention when reversed-phase columnswereused,andtherewasnocoelutionoftheesterified andnonderivatizedcomponents(datanotshown).Butylesterfor- mation changed thepolarity of the molecules,resulting in the formationof well-retainedpeakswithexcellentpeak shapesin casesof3-HK,PICA,and QUIN(Fig.3).Esterificationof analytes withthemixtureofn-butanolandacetylchlorideisacommonly usedproceduretoderivatizeaminoacidsfornewbornaminoacid screeningviamassspectrometry[22].

Thederivatizationmethodwasoptimizedbyassessingtheeffect ofreactiontime(0,20,30,40,50,and60min;n=3;Fig.4),andthe maximumconversionofcarboxylgroup(s)toesterswasachieved after60-minreaction.Althoughesteriﬁcationwasnotcomplete after60minforallcomponents(74%–95%),themethodcouldbe

(7)

Table3

Physicochemicalproperties(hydrogenbonddonor/acceptorgroups,pKa,logP,logD,andisoelectricpoint)oftryptophanmetabolites(calculatedusingChemAxon’sChemi- calizesoftware[24]).

Analyte Molecularmass(g/mol) pKa1 pKa2 pKa3 pI logP logD(pH3) logD(pH6) logD(pH9) Hydrogenbond donorcount

Hydrogenbond acceptorcount

SERO 212.68 9.31 10.00 9.78 0.48 −1.85 −1.76 0.32 3 2

KYN 306.29 1.19 8.96 6.11 −1.91 −2.35 −1.91 −2.20 3 5

3-HANA 153.14 1.94 10.37 4.82 3.03 1.15 0.75 −0.42 −2.36 3 4

TRP 204.23 2.54 9.40 5.97 −1.09 −1.19 −1.09 −1.22 3 3

5-HIAA 191.18 4.22 9.56 – 1.41 3.80 −0.38 −2.16 3 3

ANA 137.14 4.89 1.95 3.34 1.45 1.27 0.17 −2.00 2 3

KYNA 189.17 2.47 2.31 1.87 2.40 −0.47 −1.68 2 4

XA 205.17 2.17 9.06 14.22 3.25 −0.17 −0.17 −0.65 −2.22 3 5

3-HK 224.21 0.99 9.86 3.378.90 6.11 −2.21 −2.73 −2.21 −2.57 4 6

MELA 232.28 −1.57 15.90 7.08 1.15 1.15 1.15 1.15 2 2

PICA 123.11 1.00 5.52 3.04 −0.65 −0.65 −1.25 −2.71 1 3

QUIN 167.12 0.31 4.16 6.67 2.27 −1.2 −1.23 −3.11 −6.39 2 5

SERO,serotonin;KYN,l-kynurenine;3-HANA,3-hydroxyanthranilicacid;TRP,tryptophan;5-HIAA,5-hydroxyindoleaceticacid;ANA,anthranilicacid;KYNA,kynurenicacid;

XA,xanthurenicacid;3-HK,3-hydroxykynurenine;MELA,melatonin;PICA,picolinicacid;QUIN,quinolinicacid.

Fig.4.Efﬁciencyofesteriﬁcationinpercentagesatdifferentreactiontimes.Quino- linicacid,red;3-hydroxykynurenine,green;picolinicacid,orange.

reliablyusedbecauseofthepresenceofSIL-IS,whichhasnearly identicalchemical andphysicalpropertiesasthetargetanalyte.

Althoughtheabsoluteresponse maybeaffected,theanalyte/IS peakarearatioshouldbeunaffectedandthemethodshouldbe accurate,precise,andrugged.

Duringsamplepreparation,allcarboxylgroup-containingcom- poundswereesteriﬁedin50%ofthesamplesandremixedwith theuntreatedfractiontoanalyzebutylated3-HK,PICA,andQUIN in the same chromatographic run with SERO, MELA, and the non-esteriﬁedcarboxyl group-containingTRPand itsremaining metabolites.

3.2. DevelopmentandoptimizationoftheUHPLCmethod

LC–MS is one of the most prominent analytical techniques becauseofitsinherentselectivityandsensitivity.Mostcurrently used HPLC/UHPLC separation techniques are performed using reversed-phasechromatography(RPC).RPChasbecomethestan- dardtechniqueforanalyzingawiderange ofsmall(evenlarge) compoundsrangingfromneutral polarand nonpolarsolutesto acidic,basic,andamphotericcompounds.Stationaryphasesused inRPCtypically comprisevarying lengthsofhydrocarbonssuch asC18,C8,andC4,forwhichanalyteretentionismainlydrivenby hydrophobicandvanderWaalsinteractions.Asthemobilephase,a mixtureofwaterwithamiscible,polarorganicsolventsuchasACN andMeOHisused,usuallysupplementedwithdifferentadditives.

TRPandsomeofitsmetabolitesareratherhydrophobicbecause oftheindoleorphenylring;however,theattachedpyridinering andhydroxy,amine,orcarboxygroupsincreasetheirhydrophilic- itywhileprovidingionicpropertiestomolecules.Developingan

LC MSmethodfor thesimultaneousquantitative estimationof thesetypesoftargetcompoundsisextremelychallengingbecause of the great diversity in theirphysicochemical properties (e.g., pKa,isoelectricpoint,hydrophobicity,numberofhydrogenbond donor/acceptorgroups,andsolubility;Table3).

Thefirststageofchromatographicmethoddevelopmentisiden- tifyingthemostpromisingcolumnchemistry,organicmodifier,and pHofthemobilephaseforanalyteseparation.Toscoutthemost appropriatecolumnandorganicsolvent,weperformedtwoini- tialexperimentsusinganumberofcolumnswithelutionusinga MS-compatiblegenericgradient(seeSection2.5.1).The“best”col- umnandorganicmodifierwereselectedviavisualcomparisonof theresultantchromatograms,takingintoaccounttheoverallpeak shapes,retentionofhighlypolarcompounds,andbaselinesepara- tionofanalytesintheshortesttime.Then,DryLab^®4wasusedto optimizeseparationontheselectedcolumn.

AnoctadecylsilanecolumnisthefirstchoiceforseparatingTRP anditsmetabolites;therefore,wetestedsevenC18columns.Sep- arationwasalsoperformedonreversed-phasecolumnscontaining otherhydrocarbongroups (C8,C12,and phenyl-hexyl).Because of the different characteristics of the tested stationary phases, theyexhibiteddistinctretentionandelutionprofiles.PICAwasthe firstelutinganalyteonallcolumns,butsatisfactoryretentionwas observedonlyusingtheBEHC18andLunaOmegaPolarcolumns.

TheearlyelutingQUINcouldnotbedetectedusingtheKinetexC18 andCortecsC18columns,whereasitappearedasawidepeakwhen theBEHC18andLunaC18columnswereused.Itselutionproﬁle wasunacceptablywideontheothercolumns.3-HKexhibitedlow retention,anditsometimeselutedasdoublepeaks(LunaC18,BEH C18,PolarC18,andCortecsC18columns).ThepeakofSEROhad frontingandtailingontheKinetexC18,LunaC18,BEHC18,and CortecsC18columns.ThepeakshapeofXAwasunacceptableon theLunaC18,KinetexC18,CortecsC18,andphenyl-hexylcolumns.

TRP,5-HIAA,ANA,KYNA,andXAgroupedatthesecondpartofchro- matogramsin0.6–1.3-minwindowsusuallyasunseparatedpeaks.

Stationaryphaseswithshorteralkylchainsprovedtobeevenless retentivefortheseanalytes;therefore,wehadtosearchforother typesofreversed-phasecolumns.

Retentiontimeswereshorterandpeaksweresomewhatsharper usingACNasanorganicmodiﬁer;however,noimportantselectiv- ityorsensitivitydifferenceswereobserved.

Fluorinatedstationaryphases,particularlythoseinvolvingPFP moieties,havebecomepopularalternativestothetraditionalalkyl phasesbecauseofthedifferencesinselectivityandretentionthat theyprovide.PFPphasesusemultipleretention mechanismsto separatesmall,highlypolararomaticcompounds[23],andthey appeartobeidealchoicesforanalyzingTRPmetabolites.Theelec-

(8)

tronegativeﬂuorineatomsproduceanelectron-deﬁcientphenyl ring,which permitsthePFPphase toact asa Lewisacidor an electronacceptor.␲–␲interactions canoccurwithsolutesrich in electrons(Lewis bases)suchas amino and hydroxyl groups.

Carbon–ﬂuorinebondsareextremelypolar,thusenablinganalytes toalsoberetainedbydipole–dipoleinteractions,H-bonding,and sometimesionicinteractions,resultinginincreasedanalytereten- tion.Inthepresenceofarigidaromaticring,thesoluteshapecan alsodictateselectivity.Thepredominanceofeachretentionmecha- nismisinﬂuencedbythephysicochemicalpropertiesandstructure ofanalytesandthechromatographicconditionsused.

Core-shellKinetexF5(2.1×150mm,2.6␮m)andtwoKinetex PFP(2.1×100mm,2.6␮m;Phenomenex)columnswithdifferent lotnumbersweretestedusingtheabovementionedgenericgra- dients.Similarly,asobservedfortheC18stationaryphases,PICA, QUIN,and3-HKelutedinwide,skewedpeaksfromeveryfluori- natedphase.Despitethedifferentstructures(distinctfunctional groupsonbenzene,pyridine,orindolerings),thepKavaluesand H-bondingcapabilitiesofKYN–SEROand5-HIAA–KYNA–XApairs couldnotberesolvedontheKinetexF5column,asobservedfor 3-HANA–KYNandXA–TRPpairsontheKinetexPFPcolumn(from newerseries).TheolderKinetexPFPcolumncouldseparatethe abovementionedanalytepairsusingMeOHasanorganicmodifier, anditgenerallyproduceda“good-looking”chromatogram(except PICA,QUIN,and3-HK).ChoosingACNinsteadofMeOHresultedin differencesinselectivity;however,3-HANAandTRPwerecoeluted, followedby deteriorated quantification of3-HANA becausethe concentrationofTRPin serumand CSFareatleastthreeorders ofmagnitudehigherthanthatof3-HANA.Similarly,ANA,KYNA, andXAcouldnotberesolvedusingACN.

BecauseofthedifferentionizablefunctionalgroupsofTRPand itsmetabolites,thepHofthemobilephasecouldinﬂuencethe ionizationstateofthesecompounds.However,because10ofour analytespossesscarboxylgroup(s),theirionizationwassuppressed atpH<3,resultinginmorehydrophobicspecies.Themajorityof thesemetabolitesarezwitterionicmolecules,and7ofthemhave isoelectricpointsbelow6.0,indicatingthatthecarboxylgroupis intheanionicformaroundpH6. AthigherpHvalues,more of thesecompoundshaveadditionaldeprotonizedfunctionalgroups, resultinginanionicspecies.Usually,detectionofanionicanalytes viamassspectrometryinthenegativemoderesultsinlowersen- sitivity.Moreover,almostallofourtestedcolumnsarereportedby theirrespectivemanufacturerstohavepoorstabilityabovepH8.

ThesameconclusioncanbedrawnfromthecalculatedlogP andlogDvaluespresentedinTable3[24].Inreversed-phaseHPLC, retentiontimeiscloselyrelatedtothehydrophobicityofanalytes, whichcanbeestimatedusinglogP(octanol/water)values.Whenan analyteisdissociatedinthemobilephase,logD(distributioncoefﬁ- cient,whichreﬂectsthecontributionofallionicspeciestothetotal hydrophobicityofananalyte)isoftenusedinsteadoflogP.Increas- ingpHresultedinlowerlogDvaluesforalmostalltestedanalytes, whichagainresultedinlowerretentiontime.

Basedontheresultsofcolumn-andorganicmodiﬁer-scouting experimentsand thetheoreticalconsideration ofthe pHofthe mobilephase,separationofTRPanditsmetaboliteswasperformed onaKinetexPFPcolumnunderacidicconditionsusing0.1%(v/v) aqueousFA(pH2.7)andMeOHasanorganicmodiﬁer.

Derivatization of the carboxyl group(s) of TRP and 9 of its metabolitesviaesterificationwithfourcarbonatomscontaining n-butanolincreasedthehydrophobicityofanalytes,resultingin higherretentiononreversed-phasechromatographiccolumns.The peakshapeofthemostpolarcompounds3-HK,PICA,andQUIN wasoptimalafterbutylation(Fig.3),andtheirretentionincreased significantly.QUINhastwocarboxylgroups,andtheiresterifica- tionmadeQUINthemosthydrophobicTRPmetabolite.Using a genericgradient(seeSection2.5.1),theleastretentiveesterified

3-HKelutedimmediatelybeforethemostretentivenonesterified metaboliteMELA.Esterified5-HIAAandKYNcoeluted,whereas5 butylatedmetabolites(TRP,KYNA,3-HANA,XA,andANA)grouped betweenesterifiedPICAandQUINwithoutdisturbingthedetermi- nationofotheranalytes.

Considering the chromatographic behavior of esterified and nonesterifiedTRPanditsmetabolites,7compoundswereselected for their quantitative estimation in the nonderivatized forms, whereas3metaboliteswerequantitativelyestimatedintheirbuty- latedforms.SEROandMELAwerealsoanalyzed(nonesterifiable), buttheirinvivoconcentrationsprovedtobelowerthantheirLOD inthecaseofCSFsamples.However,SEROcouldbequantifiedin serumsamples.

Apartfrom the type and physical characteristics of thesta- tionaryphase, natureof organicmodifier, andcomposition and pHofthemobilephase,otherchromatographicparameters(e.g., length and profile of the gradient, T, ionic strength of the mobilephase,initialandfinalorganicconcentrationsofeluents, and column dimensions) canalso influence analyte separation.

Improvingthe analytical performance testing of these parame- terswouldbetediousandtime-consuming,butnotcost-effective or “green.” However, computer-assisted HPLC method devel- opment/optimization software can help deﬁne the appropriate conditionsforrobust,precise,andreproducibleanalysisandcan simultaneously save resources. DryLab^®4 is one such software thathelpsthechromatographerdevelopbetterandmorereliable HPLC/UHPLCmethodsinashortertime.Thiscanbeachievedby insilicomodelingofretentionandresolutionbasedonalimited number(2–12)ofinitialexperiments[13].

TooptimizeseparationofTRPandits11metabolitesonaKine- texPFPcolumnusingMeOHasanorganicmodiﬁer,weperformed fourinitiallineargradientchromatographicrunstotesttheeffects ofgradientsteepness/t_G(t_G1=5andt_G2=15min)andT(T₁=25^◦C andT2=50^◦C)onretentiontimeandresolution.Basedonthese chromatograms,thesoftwaresimulatedchromatogramsandcre- ateda2Dcolor-codedresolutionmap,plottingcriticalresolution asa functionoftGandT.Redregionsonthisplotrepresentthe optimalchromatographicconditionswitharesolutionof>1.3.We selectedaworkingpoint(t_G1=5min,T=25^◦C),andusingthegradi- enteditor,wesimulatedtheresolutionofthetwomostproblematic analytepairs:SERO–KYNand3-HK–XA(Fig.5).Thebestseparation wasachievedusingagradientwithfourlinearsegments:0–1min, 10%–30%B;1–3min,30%–50%B;3–3.5min,50%–90%B;and 3.5–5min,90%B(Fig.6).Retentiontimepredictionwasexperi- mentallyveriﬁedusingtheselectedparameters.Theexperimental (texp)andsimulated(t_sim)retentiontimesforeachanalyte,together withtheirratio(r=texp/t_sim),aresummarizedinTable2.Allratios rangedfrom0.99to1.02,indicatinganexcellentmatchbetween thesimulatedandactualseparations.

DryLab^®4 combines over 30 years of HPLC expertise with thelatestsoftware technologiesand enableschromatographers involvedinpharmaceuticalandchemicalindustries,environmen- talprotection,andresearchindustriestoeasilycreatefast,robust, high-qualitymethods[21].Tothebestofourknowledge,thisisthe ﬁrstreportontheuseofDryLab^®4tooptimizeLC/MSanalysisof endogenousmultianalyte-containingbiologicalsamples[25].

3.3. Methodvalidation 3.3.1. Linearity,LOD,andLOQ

TheLOD,LOQ,retentiontime,andlinearityofthecalibration curveswithr²valuesareshowninSupplementaryTable1forCSF andSupplementaryTable2forserum.Forallanalytes,r²exceeded 0.99.LODwaslessthan7.80nM,andinmostcases,LOQwaslower than8.26nMforallmetabolites,exceptinSEROandQUIN(23.65 and20.59nM,respectively).

(9)

Fig.5.DryLab^®4resolutionmapwithworkingpoint(tG=5min,T=25^◦C).

Fig.6. Extractedionchromatogramofquantitativelyestimatedanalytes:serotonin(1),l-kynurenine(2),3-hydroxyanthranilicacid(3),tryptophan(4),5-hydroxyindoleacetic acid(5),anthranilicacid(6),kynurenicacid(7),xanthurenicacid(8),3-hydroxykynurenine(9*),melatonin(10),picolinicacid(11*),quinolinicacid(12*).*,indicates derivatizedanalytes.

3.3.2. Precisionandaccuracy

Theintradayandinterdayprecisions(RSD)ofthemethodfor thetwomatriceswereobtainedbyanalyzingﬁvereplicatesofthe threeQClevelson3consecutivedays(SupplementaryTables3–6).

Accordingtotheresults,thedevelopedmethodwasfoundtohave areliableprecision.

Theaccuracyrangedfrom86.7-112.0%ofintradayandinterday measurements(SupplementaryTables3and4)inaCSFandfrom 89.1-107.7%inthe“blank”serum(SupplementaryTables5and6), whichareinlinewithrecommendeddata[17,18].

3.3.3. Recovery

Tomeasureanalyteconcentrations,asimpleproteinprecipita- tionmethodwasselectedinsteadofliquid–liquidorsolid-phase extraction.Ingeneral,proteinprecipitationiseasier,morerapid, andreasonablethan theabovementioned methods.Theprocess ﬁrstwasconductedusingACNandMeOHforaCSFandwithsup- plementationwithacetone–MeOH(3:7,v/v)andacetone–MeOH (1:1, v/v) for the “blank” serum. Finally, for protein precipitation, ACN was used for aCSF and acetone–MeOH (1:1, v/v) for the “blank” serum. Analyte recoveries were determined at threedifferentconcentrationstoprovethattherecoveryofeach

analytewasconcentrationindependent,reproducible,andconsis- tent.

Analyterecoveriesrangedfrom93.8-105.3%foraCSF(Supple- mentaryTable7) andfrom 84.7-109.4% forthe“blank” serum (SupplementaryTable8).Ourvaluesarewithintherangerecom- mendedbytheICHandFDAguidelines[17,18].

3.4. TRPmetaboliteproﬁlechangesinmultiplesclerosis

In mostcases, thequantitative estimation of many analytes withdifferent polarities is challenging. Hence,derivatization is commonlyusedtoimprovetheselectivityandsensitivityofthe methodbecauseitchangesthechemicalpropertiesofcompounds and increasesthemassof analytes,which provideanexcellent strategyforavoidingtheinterferenceofmatrixcomponents[26].

Therefore,toachievesuccessfulsimultaneousquantitationofthe 12analytes,aderivatizationstepwasincludedtoimprovethepeak shapeandretentionof3-HK,PICA,andQUIN.Althoughtheconcen- trationsofbothSEROandMELAwerebelowtheirLODinCSFand thatofMELAwasalsobelowitsLODinserum,themethodwas validatedaccordingtotheICHandFDAguidelines[17,18].Appli- cabilityofthemethodwasprovenbyquantifyingthemetabolitesof

(10)

Fig.7.Boxplotsofsignificantchangesbetweenthecontrolandmultiplesclerosisgroupsincerebrospinalfluid,intheconcentrationsofkynurenicacid(KYNA),3- hydroxykynurenine(3-HK),picolinicacid(PICA),quinolinicacid(QUIN),KYNA/l-kynurenine(KYN),QUIN/KYNA,andPICA/QUIN.Significancewasevaluatedusingan independentsamplest-testoratwo-sampleWilcoxontestaftertheF-test:*p<0.05;**p<0.01;***p<0.001.

Table4

ConcentrationsofTRPanditsmetabolitesinhumancerebrospinalﬂuidandseruminthecontrolandmultiplesclerosisgroups.

CSF Serum

Analyte Controlgroup Multiplesclerosisgroup p-value Controlgroup Multiplesclerosisgroup p-value

Concentration(nM) Concentration(nM)

SERO <LOD <LOD – 919.9±451.1 725.9±365.4 0.1761

KYN 49.3±16.6 60.4±21.2 0.1691^a 2397±664.4 2365±714.9 0.8949

3-HANA 0.7±0.7 1.1±0.8 0.1533^a 58.0±20.6 55.3±16.8 0.6853

TRP 1847±305.6 1781±367.3 0.9277^a 52043±11057 50948±9325 0.4089

5-HIAA 101.0±36.7 85.8±31.4 0.3487 66.0±21.3 50.7±11.3 0.0248

ANA 7.0±3.9 6.3±3.0 0.7957^a 16.2±6.3 15.9±5.0 0.9047

KYNA 0.9±0.6 0.5±0.2 0.0469 32.2±10.4 27.2±9.5 0.1692

XA 0.037±0.034 0.053±0.062 0.8534^a 13.7±9.0 10.6±5.0 0.5981^a

3-HK 4.7±2.8 6.4±2.6 0.0346^a 92.0±46.3 89.8±27.9 0.8724

MELA <LOD <LOD – <LOD <LOD –

PICA 13.4±4.6 10.0±3.4 0.0224 35.4±8.4 31.6±9.2 0.2586

QUIN 16.4±2.8 24.7±5.1 0.0001 100.5±45.9 158.0±73.5 0.0302

QUIN/KYNA 36.2±37.1 73.2±57.8 0.0015^a 3.7±2.0 6.5±2.9 0.0183

KYN/TRP(×10³) 28.2±11.1 35.1±14.5 0.1575 48.2±15.8 46.8±13.5 0.9855^a

PICA/QUIN 0.8±0.4 0.4±0.2 0.0065 0.4±0.3 0.2±0.2 0.0476^a

KYNA/KYN(×10³) 18.5±12.3 8.9±5.0 0.0041^a 14.2±5.6 11.5±3.3 0.0832^a

CSF,cerebrospinalﬂuid;LOD,limitofdetection;SERO,serotonin;KYN,l-kynurenine;3-HANA,3-hydroxyanthranilicacid;TRP,tryptophan;5-HIAA,5-hydroxyindoleacetic acid;ANA,anthranilicacid;KYNA,kynurenicacid;XA,xanthurenicacid;3-HK,3-hydroxykynurenine;MELA,melatonin;PICA,picolinicacid;QUIN,quinolinicacid.

Signiﬁcantp-valuesarehighlightedinbold.Dataarepresentedasthemean±SD(control,n=14;multiplesclerosis,n=20).p-valuesweredeterminedusinganindependent- samplest-testandtwo-sampleWilcoxontest.

aTwo-sampleWilcoxontest.

interestfrombothhumanCSFandseruminashortstudy,inwhich metaboliteconcentrationswerecomparedbetweenthemultiple sclerosisandcontrolgroups.

Investigationofthemolecularbackgroundofneurodegenera- tivediseasessuchasmultiplesclerosisrequiresintensiveresearch, andsomeanalytesderivedfromTRPhavebeenstudiedcompre- hensively[1,2,5,8].Theneedtomeasureallimportantmetabolites inasinglerunisnecessaryconsideringtheefﬁciency,ﬁnancial,and time-savingaspectsofamethod,whereasrobustnessensuresthe applicabilityofthesamemethodondifferentbiologicalmatrices,

whichiscrucialforbettercomprehensionoftheactualpathome- chanismsofdiseases.

In a recent study [8] of TRP metabolites, changes in KYNA, PICA, QUIN, and 3-HK concentrations were compared between patientswithmultiplesclerosisandcontrols.Ourﬁndingsarein linewithliteraturedata,andQUIN-inducedexcitotoxiceffectscan becounterbalancedbyKYNA.In thecontextofmultiplesclero- sis,QUINconcentrationincreaseddramatically inboth CSF and serum,whereasKYNAconcentrationslightlydecreased (Table4 andFigs.7and8),resultinginasigniﬁcantlyhigherQUIN/KYNA

(11)

Fig.8.Boxplotsofsigniﬁcantchangesbetweenthecontrolandmultiplesclerosisgroupsinserum,intheconcentrationsof5-hydroxyindoleaceticacid(5-HIAA),QUIN, QUIN/KYNA,andPICA/QUIN.Signiﬁcancewasevaluatedusinganindependentsamplest-testoratwo-sampleWilcoxontestaftertheF-test:*p<0.05;**p<0.01;***p<0.001.

ratiointhemultiplesclerosisgroupthaninthecontrolgroup.The QUIN/KYNAratioreﬂectsexcitotoxicitybecauseexcitotoxicityis increasinglyfavoredastheratioincreases.Thehypothesisthatexci- totoxicTRPmetabolitescancauseneurodegenerationinmultiple sclerosisissupportedbythesedata.

TheKYNA/KYNratio(apotentialsurrogatemarkerofKATactiv- ity)decreasedinthemultiplesclerosisgroupcomparedwithinthe controlgroup.Thisoriginatedfromdecreasedmetabolismofthe neuroprotectivebranchofKP,aphenomenondescribedinother disorders[27].

Inaddition,thePICA/QUINratio,whichincreasedinthemulti- plesclerosisgroup,maybearesultoftheabilityofneuroprotective PICAtoantagonizeQUINneurotoxicity[28]ormayberepresenta- tiveoftheinﬂammatoryprocessesinmultiplesclerosis.

InCSFsamples,themeasured3-HKconcentrationsignificantly increased,whichisexpectedbecausethisneurotoxicmetaboliteis knowntopotentiateQUIN-inducedexcitotoxicity[8]. Moreover, QUINis involved inthephosphorylationof neurofilaments,the structuralcomponentsofaxons[29].Theincreasedconcentration ofneurofilamentsinCSFandserumofpatientswithmultiplescle- rosisreflectstheextentofneuroaxonaldamage,asdescribedina previousstudy[30].

4. Conclusion

We validateda UHPLC–MS/MSmethodfor thesimultaneous quantitativeestimationofTRPandits11mostimportantmetabo- litesderivedviabothKPandSEROpathways(SERO,KYN,3-HANA, 5-HIAA,ANA, KYNA,3-HK,XA,MELA,PICA,andQUIN).Theval- idated chromatographic method is accurate and applicable to humanCSFandserum.Although3metabolites(3-HK,PICA,and QUIN) were analyzed in derivatized forms, theywere assessed togetherwith 9 underivatized metabolitesin a singlerun. The

chromatographicmethodwasinsilicooptimizedusingDryLab^®4.

Selectivity,linearity,LOD,LOQ,precision,accuracy,andrecovery valuesuponvalidationwerewithintheacceptablerangesrecom- mendedbytheICHandFDAguidelines[17,18].Theapplicability of the chromatographic method wasproved by comparing the describedconcentrationsobtainedusingbiologicalmatriceswith literaturedata.

Thisstudyprovidesa referencefortheclinicalandscientiﬁc researchofmultiplesclerosis.Ourresultssuggestthatratiosofdif- ferentTRPmetabolitescouldbeputativebiomarkersofthisdisease.

Tothebestofourknowledge,thisistheﬁrststudytoquantitatively estimatederivatized3-HK,PICA,andQUINsimultaneouslywith9 otherTRPmetabolitesinhumanserumandCSF.

DeclarationofCompetingInterest

Theauthorsdeclarethattheyhavenoknowncompetingﬁnan- cialinterestsorpersonalrelationshipsthatcouldhaveappearedto inﬂuencetheworkreportedinthispaper.

CRediTauthorshipcontributionstatement

FerencTömösi:Methodology, Investigation,Formal analysis, Writing-originaldraft,Visualization.GáborKecskeméti:Method- ology,Investigation,Formalanalysis,Writing-originaldraft.Edina Katalin Cseh: Writing - original draft. Elza Szabó: Resources.

Cecília Rajda: Writing - original draft. Róbert Kormány: Data curation,Software.ZoltánSzabó:Formalanalysis,Datacuration.

LászlóVécsei:Conceptualization,Supervision,Fundingacquisition.

TamásJanáky:Conceptualization,Supervision,Fundingacquisi- tion,Writing-review&editing.