Retrospective screening of synthetic cannabinoids, synthetic opioids and designer benzodiazepines in data files from forensic post mortem samples analyzed by UHLC-QTOF-MS from 2014 to 2018

(1)

Retrospective screening of synthetic cannabinoids, synthetic opioids and designer benzodiazepines in data ﬁ les from forensic post mortem samples analysed by UHPLC-QTOF-MS from 2014 to 2018

Per Ole M. Gundersen

^a,b,

*, Sebastian Broecker

^c

, Lars Slørdal

^b,a

, Olav Spigset

^a,b

, Martin Josefsson

^d,e

aDepartmentofClinicalPharmacology,St.OlavUniversityHospital,Trondheim,Norway

bDepartmentofClinicalandMolecularMedicine,NorwegianUniversityofScienceandTechnology,Trondheim,Norway

cBroeckersSolutions,Berlin,Germany

dDepartmentofPhysics,ChemistryandBiology,LinköpingUniversity,Linköping,Sweden

eNationalForensicCentre,DrugUnit,Linköping,Sweden

ARTICLE INFO Articlehistory:

Received7January2020

Receivedinrevisedform19March2020 Accepted25March2020

Availableonline3April2020 Keywords:

Retrospectivescreening UHPLC-QTOF-MS Postmortembloodsamples Newpsychoactivesubstances Syntheticcannabinoids Syntheticopioids Designerbenzodiazepines

ABSTRACT

Theintroductionofnewpsychoactivesubstances(NPS)ontheillicitdrugmarkethasledto major challengesfortheanalyticallaboratories.Keepingscreeningmethodsuptodatewithallrelevantdrugsis hardtoachieveandtheriskofmissingimportantﬁndingsinbiologicalsamplesisamatterofconcern.

Aiming for an extended retrospective data analysis, diagnostic fragment ions from synthetic cannabinoids(n=251),syntheticopioids(n=88)anddesignerbenzodiazepines(n=26)notincluded inouroriginalanalyticalmethodwereobtainedfromthecrowdsourceddatabaseHighResNPS.comand convertedtoapersonalizedlibraryinaformatcompatiblewiththeanalyticalinstrumentation.Datafiles fromtheanalysisof1314forensicpostmortemsampleswithanAgilent6540ultrahighpressureliquid chromatography quadrupoletime-of-flight massspectrometry (UHPLC-QTOF-MS)performedinour laboratoryfromJanuary2014toDecember2018wereretrievedandretrospectivelyprocessedwiththe newpersonalizedlibrary.Potentiallypositivefindingsweregroupedintwo:Themostconfidentfindings containedMS/MSdataforlibrarymatch(category1)whereasthelessconfidentfindingslackedsuchdata (category2).Fivenewcategory1findingswereidentified:Flubromazepamintwodatafilesfrom2015 and 2016, respectively, phenibut (4-amino-3-phenylbutyric acid) in one data file from 2015, fluorofentanylinonedatafilefrom2016andcyclopropylfentanylinonedatafilefrom2018.Retention timematcheswithreferencestandardsfurtherstrengthenedthesefindings.Alistof35presumably positivecategory2 findingswas generated.Ofthese,onlyonefindingofphenibutwasconsidered plausible after checking retention times and signal-to-noise ratios. This study shows that new compounds can be detected retrospectively in data files from QTOF-MSusing anupdated library containingdiagnosticfragmentions.Automaticscreeningprocedurescanbeuseful,butamanualre- evaluationofpositivefindingswillalwaysbenecessary.

1.Introduction

In recent years, there has been a continuously increasing numberofnewpsychoactivesubstances(NPS)appearingonthe Europeanillicitdrugmarket[1].Thediversityandhighnumberof new compounds pose challenges for clinical and toxicological laboratories who strive to keep their drug screening methods

updated. Synthetic cannabinoids, i.e. compounds acting as cannabinoid receptor agonists and produced as alternatives to

D

-9-tetrahydrocannabinol(THC)representthelargestandmost structurally diverse group [2]. New synthetic opioids, and in particularthefentanylanalogues,havebeenofmountingconcern because of their formidable toxic potential [3–6]. Designer benzodiazepines is another group in focus due to the high prevalenceofuse,atleastinourcountry[7],comparedtoother groupsofNPS.

Todevelop,establishandmaintainascreeningmethodcapable of detecting all drugs relevant at any given time is a major challenge.Theuseofhighresolutionmassspectrometry(HR-MS)

* Correspondingauthorat:Dep.ofClinicalPharmacology,St.OlavUniversity Hospital,Postbox3250Torgarden,7006Trondheim,Norway.

E-mailaddress:[email protected](P.O.M. Gundersen).

http://dx.doi.org/10.1016/j.forsciint.2020.110274

ContentslistsavailableatScienceDirect

Forensic Science International

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a te / fo r s c i i n t

(2)

e.g. quadrupole time of ﬂight mass spectrometry (QTOF-MS) instrumentation has proven to be an applicable tool when searching for drugs of abuse in biological samples [8–13]. The detectionofunknowncompoundsistimeconsumingandhardly feasible on a routine basis with a large number of samples.

Consequently,themethodmustencompassascreeningortargeted approach, based on an extensive and comprehensive database containingmultipletypesofdataforidentification.Suchdatacan beretentiontimes(RTs)and fragmentationdatafromcollision- induceddissociation(CID), in silicoorothertheoreticalevalua- tions,inadditiontothemolecularformulaofthesubstance.The database can be created and maintained “in-house” by the laboratory.Thisrequiresaccesstoahighnumberofwell-defined reference compounds. Procurement of reference standards is costly,particularlyifadatabaseshouldbeuptodatewithasmany new and relevant compounds as possible. Databases are also commerciallyavailablefromsuppliersofMSinstruments(e.g.the Forensic Toxicology Personal Compound Database and Library fromAgilent),butusersaredependentonthefrequencyofnew releasesand/oradditionsbeinguptodate.Therearealsoexamples ofcommercialoperatorsofferingfreedatabases(e.g.themzCloud fromThermoFisher).Anotheropportunityiscrowdsourceddata- baseswithinformationsubmittedbyglobalHR-MSusers.Onesuch example is HighResNPS.com [14]. When performing CID on a certain compound, different instrument configurations tend to generate the same diagnostic fragment ions even though the relativeabundancemayvary.Thus,fragmentdataacquiredonone instrument canthen be usedas identification across platforms [14–16].Inprinciple,thesameistrueforacrowdsourceddatabase withdiagnosticfragmentsacquiredbyinstrumentsfromdifferent manufacturers, providing that the added fragment masses are convertedtotheoreticalvalues.

Incontrasttoanalyticalmethodsbasedonsingleionmonitoring ormultiplereactionmonitoring,HR-MSfull-spectrumdataremain availableandpermittheidentiﬁcationofnon-targetcompoundsand retrospectiveanalysis,alsocalledpost-targetanalysis.Fordatafrom HR-MS instrumentation with fragmentation capabilities, e.g.

QTOF-MSorlineariontrapOrbitrap,fragmentationdataarealso available.Inprinciple,allcompoundsareavailableforinvestigation at a certain level, but thedata available arelimited bysample extractionrecovery,chromatographicselectivityandthedegreeof ionization and fragmentation. Depending on which acquisition mode is used, the QTOF-MS data also contain fragment ions originatingfromthemolecularionsgenerated intheionsource.

Based on newknowledge, post-targeted analysis of data cangenerate new findings in a specific toxicological or clinical sample and ultimatelychangetheconclusioninaparticularcase.Aretrospective studyisalsoimportantasaninternalqualitycheckforthelaboratory toassesswhetherthescreeningrepertoireusediscomprehensive andrelevant.Inaddition,newtrendsindrugabusecanbeidentified, as exemplified in the study by Kriikku et al. where the toxic lifespan of U-47700wasexplored[17].

Thenumberofstudiesapplyingsucharetrospectiveapproach inaforensicorclinicaltoxicologysettingarelimited.Nobleetal.

processed2339forensicsamplesretrospectivelywithatargeted screeningmethodtodetect504-anilidopiperidine-relatedfenta- nylanalogues[18].InanothercasestudyU-47700,diclazepamand ﬂubromazepamweredetectedinretrospect[19].Mollerupetal.

applieda post-targetedapproach when developing a screening method for valproate using positive ionisation mode [20].

Retrospectiveanalysisofurine sampleshasbeenusedtodetect metabolitesofpesticides[21].Post-targetedanalysisofdatahas alsobeenusedfordetectionofdrugsandpesticidesinnon-human matricesincludingsewagewater,surfacewaterandfood[22–26].

SinceDecember2013,ourlaboratoryhasutilizedaworkﬂow basedonultra-highperformanceliquidchromatography(UHPLC)

coupledtoa6540QTOF-MSfromAgilent(SantaClara,CA,USA)for therapeutic drugs and drugs of abuse in post mortem blood samples.ThesameUHPLCandMSmethodhasbeenappliedfrom 2014tothepresent.Acommercialdatabasesuppliedwithentries addedmanuallyafteranalysingreferencematerialshasbeenused foridentiﬁcation.However,inordertodetectaneworpreviously unknown drug in a biological sample, additional information connectedtothecaseorsample(e.g.aseizure)hastobeavailable.

In ourexperience, suchinformationis rarelyavailable, andthis mayincreasetheriskofmissingdetectionofNPS.Theconsistency of the screening methodenables retrospective analysissothat new compounds can be found. The use of HighResNPS for identifying compounds in samples analysed on Agilent QTOF- MS has previously been shown, but only ﬁles from data independent acquisition (DIA) could be investigated with this approach [14]. Our method was based on data dependent acquisition(DDA) which, asopposed toDIA,involves acquiring ofMS/MSspectraafterselectionofprecursorionsisolatedbythe quadrupole.Athoroughexplanationof thedifferencesbetween DIAandDDAcanbefounde.g.inthepapersofSundströmetal.[27]

andBroeckeretal.[8].TobeabletouseHighResNPS,diagnostic fragmentinformation fromthedatabasehadtobeconvertedto spectra in the format accepted by the Agilent MassHunter Qualitative searching tool. In Agilent terminology, a library is thesumofcompoundsinadatabasecontainingMS/MSspectra andthesedatabasesandlibrariesarecalledPersonalCompound DatabaseandLibrary(PCDL).

Theaimofthisstudywastore-processdataﬁlesofforensicpost mortemsamplesacquiredfromJanuary2014toDecember2018 in a PCDL-facilitated search for NPS belonging to the sub- groups synthetic cannabinoids, synthetic opioids and designer benzodiazepines.

2.Materialsandmethod 2.1.Chemicalsandreagents

Reference substances used in the experiments to calculate recoveriesandmatrixeffectsandexploreinstrumentsensitivities werepurchasedassolidmaterialorstocksolutionsfromeitherof thefollowingsources:CaymanChemicals (AnnArbor,MI, USA), ChironAS(Trondheim,Norway),SigmaAldrich(St.Louis,MO,USA) andLipomed(Arlesheim,Switzerland).Individualstocksolutions intherangefrom0.2to1.0mg/mLwerepreparedandcombined into working solutions which were spiked into blood. For conﬁrmationoftentativeﬁndings,referencesubstancesoftilidine, phenibut (4-amino-3-phenylbutyric acid) and JWH-167 were purchasedfromSigmaAldrich,ChironASandCaymanChemicals respectively.LC–MSqualityacetonitrile,methanol,LiChrosolve¹ waterandARISTAR¹formic acidwereallpurchasedfromVWR Chemicals(Oslo,Norway).AmmoniumacetateofLC–MSgradewas fromSigmaAldrich(St.Louis,MO,USA).Asolutionoftheinternal referencestandardscodeine-d3,morphine-d3,benzoylecgonine- d3andgriseofulvinwaspreparedbydilutingstocksolutionsin20%

methanol(v/v)inwatertoaﬁnalconcentrationof200ng/mL.D3- codeine, d3-morphine and d3-benzoylecgonine werefrom Lip- omed whereas griseofulvin was from Janssen Chimica (Geel, Belgium).

2.2.Validationoforiginalscreeningmethod

2.2.1.Instrumentsensitivityandlimitofidentiﬁcation

The same UHPLC-QTOF-MS instrumental method, sample preparationand internalreferencestandard concentrationwere used for allthe samplesthroughout theperiod. The peak area resultsandRTsoftheinternalreferencestandardsinonedataﬁle

(3)

per batchwereextracted in orderto illustrate thevariation in responseovertime.Limitofidentification(LOI)wasevaluatedfora selectionof syntheticcannabinoids (MDMB-CHMICA,AB-CHMI- NACA, BB-22, JWH-018, PB-22 and THJ-018), synthetic opioids (fentanyl, remifentanil, cyclopropylfentanyl, para-fluorofentanyl, furanylfentanyl, acetylfentanyl) and designer benzodiazepines (deschloroetizolam, diclazepam, etizolam, flubromazepam, flubromazolam,pyrazolamandmeclonazepam).Bloodsampleswere spikedat0.1,0.2,0.5,1.0,2.0,5.0,10and20ng/mL,andpreparedin triplicateswiththesamemethodasdescribedforthepostmortem samples.LOIwasdefined astheminimumconcentrationwhere thecompound wasidentifiedandatleastoneMS/MSspectrum wasacquiredforlibrarysearchinallthreeparallels(seeSection2.5 fordetailsonidentification).

2.2.2.Recoveryandmatrixeffects

Recoveries(REs)andmatrixeffects(MEs)werecalculatedfor thesamecompoundsasusedintheLOIexperiment.Subsamplesof pooledwholebloodwerespikedafter(B)orbefore(C)extraction to a ﬁnal concentration of 0.1

m

^g/mL. ^The ^peak ^areas ⁱⁿ ^neat

standardsolutionofthesameconcentration(A),sampleBandC wereusedtocalculateREandME(Eqs.(1)and(2)).AnMEbelow 100% indicates ion suppression whereas a value above 100%

indicatesionenhancement.

RE ð%Þ ¼C

B100 ð1Þ

ME ð%Þ¼ B

A100 ð2Þ

2.3.Originalanalysisofthebloodsamples

Dataﬁlesincludedinthisstudywerefromtheanalysesofpost mortem blood samples from forensic autopsies sent to our laboratory in theperiod fromJanuary 2014toDecember 2018.

Inalimitednumberofcaseswherebloodwasnotavailable,spleen tissuewasused.Samplesfromatotalof1314caseswereanalysed in this period. Permission to re-process the data ﬁles (in this contextmeaningopeningthedataﬁleandrunthealgorithmwith thenewPCDL)wasgivenbytheRegionalCommitteeofMedical andHealth ResearchEthics inMidNorway (approvalNo. 2018/

2157).The datafileswere anonymized and theanalyst had no information about the original findings when doing the reprocessing.Asecondpersoncomparedthenewfindingswiththe analytical report originally attached to the relevant cases.

Accordingtothepermissiongrantedfromtheethicscommittee, re-analysis of the sample specimens as such could not be performed. The samples were originally processed with the commercially available Forensic Toxicology Personal Compound DatabaseandLibraryfromAgilent(SantaClara,CA,USA)withmore than3000compoundscontainingMS/MSspectracomplemented withbetween250and300compoundswithRTs.

2.3.1.Samplepreparation

Each blood sample was thawed at room temperature and 200mg was weighed into a micro tube and 50

m

^L ^solution ^of

internalreferencestandardand800

m

^L^ice-coldacetonitrilewere added.Thetubewasthenmixedonavortexmixerfor30sand centrifugedat7000gfor10min.before500

m

^L^of^thesupernatant was transferred to a 96-well plate, evaporated todryness and reconstituted in 50

m

^L ^of ^30% acetonitrile (v/v) in 0.03mg/mL ammoniumformate.Inthecaseswhereonlyspleenwasavailable, samplepreparationwasadjustedaccordingtotheconditionofthe tissue.Ifablood-likematerialcouldbeobtainedfromthespleen,it

washandledasabloodsample.Intheothercasesasubsampleof tissuematerialwashomogenizedwithanequalvolumeofH2O, and200mgofthismaterialwereprocessedlikeabloodsample.

The samples were prepared in weekly batches by the same procedurethroughouttheperiod.

2.3.2.Instrumentation

Instrumentalanalysiswas performedusinga 6540QTOF-MS (Agilent,SantaClara,CA,USA)withelectrosprayionization(ESI) coupledwitha1290InﬁnityUHPLCsystemfromAgilentequipped withanAcquityHSST3column(100mm2.1mm,1.8

m

^m)^from

Waters(Milford,MA,USA).Aninjectionvolumeof2

m

^L^was^used.

Separationwasachievedusingamobilephaseconsistingof0.05%

formic acidin 10mMammoniumformate(A) and0.05%formic acid in acetonitrile (B). A gradient witha ﬂow of 0.50ml/min startingat5%Bincreasingto50%in10min.andcontinuingto100%

overthenext6min.wasused.Aftera4-minuteholdat100%Bthe columnwasre-equilibratedfor2min.at5%B,givingatotalcycle timeof22min.Autosamplerandcolumntemperaturesweresetto 10Cand50C,respectively.

PositiveESI was usedand withfragmentorvoltageat 120V, capillaryvoltageat3500V,gastempat320C,gasﬂowat8L/min, nebulizerpressureat40psigandsheathgastemperatureat380C.

Data was acquired in data dependent Auto MS/MS mode. MS spectraandMS/MSspectrawerebothacquiredinthemassrange of50–1000m/zatarateof6Hz.Thedetectoroperatedin2GHz extendeddynamicrangegivingaresolution(m/

D

^m^at^FWHM)^of

approx.20,000atm/z322.0481.Precursorselectionwasbasedon abundanceandanintensitythresholdof1000countswasapplied.

Afteronespectrumfroma precursorwasacquired,thisspeciﬁc precursorwasexcludedfor0.03min.Precursorswerefragmented inthecollisioncellusinganelectronvoltageaccordingtoEq.(3):

CollisionenergyðeVÞ¼4þð0:06m=zof precursorÞ ð3Þ Thecomputercontrollingtheinstrument wasequippedwith theMassHunterAcquisitionsoftware(Acq)B.05.01(Agilent,Santa Clara, CA, USA). The acquireddata ﬁles consisted of MS1 (full spectrumMS-only)ofallionizedcompoundsandMS/MSspectra of theprecursorsselectedforfragmentation.Them/z massesof 121.0509and922.0098wereappliedforautomatedmasscorrec- tioninallMSspectra.Adailyperformancesampleofamphetamine (0.74ng/mL), diazepam (0.35ng/mL), 7-amino-ﬂunitrazepam (0.35ng/mL),morphine(0.35ng/mL)and

D

9-tetrahydrocannabinol(0.5ng/mL)inMeOHwasinjectedatthebeginningofevery analytical run to monitor important instrument parameters.

SampleswerenotanalysediflargedeviationsinRTs(morethan 0.2min.),massaccuracies(morethan5ppm)orpeakareasfrom thehistoricalaverageswereobservedforthecompoundsinthe dailyperformancesample.

2.4.CreatinganewPCDL

HighResNPS (highresnps.com) is a free, online, spreadsheet- format,crowdsourcedHR-MSdatabaseforNPS-screeninginitiated and managedby a group of researchers at Section of Forensic ChemistryattheUniversityofCopenhagen[14].Severalcontrib- utors worldwide submit fragmentation data when new drugs (referencestandardsorseizuresetc.)aredetectedandanalysedby aHR-MSinstrument.Also,diagnosticionsderivedfromtheoretical dissociationsofthemoleculesaresupplied.FromthisHighResNPS database(totalnumberofentriesinMay2019was1782including duplicates,and1304containedatleastonediagnosticfragment ion), 374 unique compounds with minimum one diagnostic fragment primarily belonging to the drug classes synthetic cannabinoids,syntheticopioidsordesignerbenzodiazepineswere selected. NPS already present in the screening method

(4)

implementedin2014wereﬁlteredout.Basedonthisselectiona PCDLwasdeveloped.Forthispurposeeachcompoundwasadded asanindividualdatabaseentry.Thenthesoftwaretool“Spectrum Generator”createdbyBroeckersSolutions(Berlin,Germany)was usedtoconvertthetext-basedinformationofdiagnosticionsfrom theHighResNPSdatabaseintotheAgilent“cef”ﬁleformatwhich allowsanimportoflibraryspectraforeachPCDLentry.Table1 shows the resulting HighResNPS subset PCDL content. By this approachthediagnosticfragment ions werestoredasa library spectrum. Relative abundance of the ions was not taken into account even though this would be possible by the software

“SpectrumGenerator”.Thecollisionenergyofthelibraryspectra waschosenbythesoftwareas20eVjusttohaveanyvalueinthe PCDL.Anexampleofthelibraryentryofﬂubromazepamisshown inFig.1.Acompletelistofthe374uniquecompoundsisgivenin thesupplementarymaterial(TableS1).

2.5.Dataprocessing

Ofthe1314dataﬁlesavailable,batchesofapprox.250werere- processed using MassHunter DA Reprocessor software B.09.00 (Agilent,SantaBarbara,CA,USA).There-processingwasrelatively fast,approximately 1min. per sample,when usinga computer equippedwitha2.67GHzprocessorand8GBofRAM.Thisprocess wasrunninginthebackgroundallowingre-processeddataﬁlesto beopenedandevaluatedinbatchesof50–80simultaneouslyin MassHunterQualitativeAnalysisSoftware(version10.0)(Agilent, SantaClara,CA,USA).

Thequalitativemethodusedinthere-processingwasbasedon thealgorithm “Find byformula” together witha librarysearch, bothusingtheHighResNPSsubsetPCDL.The“Findbyformula” searchleadtopositiveﬁndingsthatwerebasedonMS1spectral information.Thecriterionwasamasserrorlessthan5ppmanda scoreabove 80 where thescoringwas taking themass match, isotopespacingandisotopeabundanceintoaccount.Inthecase when MS/MSspectrawere acquiredfor the precursor ion of a detectedcompound, theseMS/MSspectrawerecompared with those in thePCDL. The comparison was done both by reverse search (the peaks in the PCDL are compared with theMS/MS

spectra)andbyforwardsearch(thepeaksintheMS/MSspectraare comparedwiththePCDL).Thethresholdlibrarymatchwassetto1 (ofmax.100)forbothforwardandreversescore.Asthemaximum number of fragment ions per library spectrum was three, the lowestresultingreversescoreofamatchwas33.

A ﬁlter in the software was applied in order to distinguish compoundswithMS/MSspectra(category1)andwithoutMS/MS spectra (category 2). Category 1 compounds found by the algorithm “Find by Formula” could be evaluated further by comparingtheacquiredMS/MSwiththelibraryspectrum.Ifthere wasnoagreementbasedontheMS/MScomparisonthecompound was considered a false positive. If there was a match, a visual evaluation comparing the acquired spectrum with the library spectrumwasundertakentoruleoutfalsepositivematchesdueto fragments of low abundance e.g. from contaminants. The LOIs estimatedforthecompoundsselectedinthevalidationappliesfor category1compounds.

Forcategory2compounds,noMS/MSdatahadbeenacquired andfragmentconfirmationcouldnotbedone.Thus,onlytheMS signalcouldbeusedtoevaluatethequalityofthefindings.Without the MS/MS spectra identification parameter the number of potential positives would have been large and included noisy signalsandbadpeakshapes.Apeakareathresholdof510⁴was applied to limit the number of findings to investigate.Conse- quently, higher detection limits were expectedfor these com- poundscomparedtocategory1compounds.Inordernottomiss anyimportantfindings,amassaccuracylimitof10ppmandmass matchscoreabove80wasfirstapplied(criteriona).Thiswastested with42randomdatafilesandgave74findings.Afterinvestigating the results and filtering out findings due to interferences and backgroundsignal, onlycompounds withmass accuracy better than5ppmandmassmatchscoreabove95wereleft.Thesetwo thresholdswereconsequentlyusedascriterionb.Finally,athird factorwasaddedtocriterionb,anRTrestrictionof1.5min,asthe compoundsinthegroupsunderinvestigationarehighlylikelyto eluteafterthistimeperiod(criterionc).Thenumberoffindingsin the42randomdatafilesasafunctionof criteriona,b orcare illustratedinFig.2.Criterionc(massaccuracybetterthan5ppm, massmatchscorehigherthan95 andRT1.5min.or more)was appliedforallcategory2compounds.Acompoundappearingin several data filesin the samebatch was consideredan isomer originatingfromthechemicalsusedorasendogenousmolecules withequaltheoreticalmasses.Theriskofacceptingfalsepositives ishigherforcategory2thanforcategory1findings,especiallyif thresholdsandlimitsaresettoowide.

Anynewﬁndingwasfurtherevaluatedbycomparingacquired MS/MSspectrawithothersources(e.g.mzCloud¹)oralternatively byanalysingareferencestandard, ifavailableatthelaboratory.

DuetovariationsintheRTsoverthetimeperiodthesampleswere Table1

NumberofnewcompoundsincludedintheHighResNPSsubsetPersonalCompoundDatabaseandLibrary(PCDL)groupedaccordingtodrugclassandsourceofdiagnostic fragmentions.

Synthetic cannabinoids

Synthetic opioids

Designer benzodiazepines

Total

Libraryspectrabasedondiagnosticionsfromstandards 126 47 22 195

Libraryspectrabasedondiagnosticionsfromtheoreticalevaluation 116 40 0 156

Libraryspectrabasedondiagnosticionsfromseizures 4 2 4 10

LibraryspectrabasedondiagnosticionsfromRESPONSEproject^a(seizuresortestpurchase on-line)

13 – – 13

Totalnumberofuniquecompounds(databaseentries) 259 89 26 374

aAEuropeanprojectnamedResponsetochallengesinforensicdruganalysis.https://www.policija.si/apps/nﬂ_response_web/seznam.php.

Fig.1.Libraryspectrumofﬂubromazepam. ¹ https://www.mzcloud.org.

(5)

originallyanalyzed, RTdeviationsupto0.5min.weretolerated when comparing these samples to reference standards. If a consistencyin fragmentsor RTswas observed, theﬁndingwas reportedtoapersonwithaccesstotheoriginalcasereport.Ifa presumablynovelmoietywasidentiﬁedandareferencestandard was available, this standard was analysed and RTsand MS/MS spectrawerecompared.

3.Resultsanddiscussion

3.1.Validationoforiginalanalyticalmethod

3.1.1.Instrumentsensitivityandlimitofidentiﬁcation

The instrument response and RT variation over time was expressedbyplottingthepeakareaandRToftheinternalreference standardsextractedfromonecalibratorfromeachanalyticalrun (Fig.S1insupplementarymaterial).Morphine-d3showedanRT difference(maximumminimum)of0.28min.andameanpeak areaof 2.710⁵(standard deviation(SD)1.310⁵).Codeine-d3 showed anRTdifferenceof 0.35min.and a mean peakareaof 4.410⁵ (SD 1.610⁵). Benzoylecgonine-d3 showed an RT difference of 0.32min. and a mean peak areaof 7.810⁵ (SD 3.910⁵).Finally,griseofulvinshowedaRTdifferenceof0.44min.

andameanpeakareaof2.810⁵(SD1.410⁵).Thepeakareasof internalreferencestandardsinthedatafilesarenotonlyreflecting the variation in instrument response but also variation in extraction efficiencyand matrix effects over time. This gives a morerelevantexpressioncomparedtoadirectinjectionofaneat performancetestsample.

LOIswereestimated for a representativegroup ofsynthetic cannabinoids, synthetic opioids and designer benzodiazepines (Table 2). LOIs are unknown for new compounds but the experiment indicated that synthetic cannabinoids could be detectedifpresentaboveapproximately10–20ng/mL,synthetic opioidsabove1ng/mLanddesignerbenzodiazepinesabove10ng/

mL. Electrospray ionization is best suited for analysis of compoundswithmedium-to-highpolaritybutisnotoptimalfor allcompounds[28].TheLOIsinTable2areonlyestimatesofthe instrumentsensitivitythroughtheacquisitionperiod.Asseenby theresultsfromtheinternalreferencestandards,thepeakareas varied during theperiod due to e.g. instrument condition and periodicmaintenance.HowthisinturnaffectedtheLOIsisdifficult todetermine,asthevalueisnotonlyaresultofsignalintensity,but alsotheautomaticselectionofprecursorionsbasedontheDDA settings.Ifthecompoundstillisamongtheprecursorsselectedfor fragmentationitwillprobablybeidentified.Giventhepeakarea threshold applied to detect category 2 substances, a higher concentration must be present in order to detect them as compared to category 1 substances. A review of the data files from the LOI experiments shows that a peak area of 510⁴

generallycorrespondstotwo-orthreefoldtheconcentrationofthe LOI of category 1 substances (see Table S2 in supplementary material).Thesedataalsoindicatethatmassmatchscoreof95is achievedformostcompoundswhenapeakareaaround510⁴is measured.

3.1.2.Recoveryandmatrixeffects

MajordifferenceswereobservedintheestimatedRE(%)ofthe syntheticcannabinoids,withvaluesrangingfrom32%(THJ-018)to 91%(AB-CHMINACA)(Table2).TheremainingcompoundshadREs above82%.AllcompoundsshowedanMEbetween69%and127%

demonstrating that both ion-suppressionand ion-enhancement occur.MEvalueswithrelativelylittledeviationfrom100%forthe studiedcompoundsindicatethatsevereionsuppressionisunlikely forothercompoundsinthesegroups.

3.2.Retrospectivedataﬁleanalysis

Atotalnumberof1314datafiles(242,252,273,242and305, respectively,fromtheyears2014to2018)wereprocessedwiththe newPCDL.Theretrospectiveanalysisrevealedsixnewfindingsof category 1 in addition to two compounds (fluorofentanyl and cyclopropylfentanyl)thathadbeenreportedwhenthedatafiles wereprocessedwiththeoriginalmethod,but onlyafterseized materialhadbecomeavailable(Tables3and4).Inadditionthere were35possiblefindingsofcategory2(Table5)notreportedwhen thedatafileswereprocessedwiththeoriginalmethod.

3.2.1.Category1ﬁndings

Flubromazepamwasdetectedintwodatafilesfrom2015and 2016respectively.Therewasamassmatchscoreinbothdatafiles higherthan95,amassaccuracybetterthan3.46ppmandanRT deviationoflessthan0.07min.Themassmatchcanbevisualized bytheresemblance ofthespectrumofflubromazepamandthe theoretical pattern indicated by the boxes in Fig. 3. The three diagnosticfragmentsinthelibraryspectrumwerealsofoundinthe MS/MSdataacquiredfromtheprecursorinthetwodatafiles(see Fig.4A).AnadditionalcomparisonoftheMS/MSspectrafromthe data file and theanalysisofa referencestandard showed good Fig.2.Numberofcategory2findingsin42randomdatafilesasafunctionof

criteriona(peakareathresholdof510⁴,massaccuracylimitof10ppmandmass matchscoreabove80),criterionb(massaccuracylimitreducedto5ppmand massmatchscoreabove95)orcriterionc(massaccuracybetterthan5ppm,mass matchscorehigherthan95andRT1.5min.ormore).

Table2

Retentiontime(RT),limitofidentiﬁcation(LOI),recovery(RE)andmatrixeffect (ME) fora selectionof compoundsinthe threegroups of newpsychoactive substancesincludedinthepresentstudy.

Substance RT[min] LOI[ng/mL] RE[%] ME[%]

Syntheticcannabinoids

MDMB-CHMICA 14.0 10 68 97

AB-CHMINACA 11.9 20 91 107

BB-22 14.3 10 57 86

JWH-018 14.4 2 51 85

PB-22 13.8 10 68 89

THJ-018 14.8 10 32 69

Syntheticopioids

Fentanyl 7.1 1 87 132

Remifentanil 5.4 1 94 123

Cyclopropylfentanyl 7.5 1 82 128

Para-ﬂuorofentanyl 7.2 0.5 88 124

Furanylfentanyl 7.3 0.5 100 124

Acetylfentanyl 6.0 1 100 127

Designerbenzodiazepines

Deschloroetizolam 8.8 2 107 119

Diclazepam 10.7 5 87 110

Etizolam 9.3 2 110 121

Flubromazepam 9.1 10 110 72

Flubromazolam 8.5 5 113 121

Pyrazolam 6.4 10 114 122

Meclonazepam 9.3 10 110 105

(6)

agreement also for additional fragment masses (see Fig. 4B).

Flubromazepamwasﬁrstdescribedin1962andisahighlypotent andincompletelyevaluatedbenzodiazepinestructurallyrelatedto phenazepam[29,30].Flubromazepamstartedtoemergeinonline shops in Europe in 2012. In Norway it was detected in seized material by the Norwegian National Criminal Investigation (KRIPOS)fortheﬁrsttimein2013.

Phenibutwasdetectedinadataﬁlefrom2015andshoweda massmatchscorehigherthan85,amassaccuracyof1.63ppm andanRTdeviationof0.12min.comparedtoareferencestandard analysedin2018.EvaluationoftheRTovertimeshowedthat a deviation up to 0.5min. could be expected due to change of analyticalcolumnlotandtubing.Phenibutisaneuropsychotropic drugwithpossiblecognitionenhancingeffectsthatwasdiscovered andintroducedintoclinicalpracticeinthe1960sSovietUnion[31].

ThedrugiswidelyusedinRussiaandisclaimedtohavevarious clinicaleffects,e.g.torelievetensionandanxietyandtoimprove sleep. Phenibut can cause dependency. It is not scheduled or classiﬁedasamedicinaldruginNorwayandisnotforlegalsale.

Privateimportisprohibitedbylaw.KRIPOSdidnotdetectphenibut in any cases before2019.Our laboratory reporteddetection of phenibutinseized materialand biologicalsamplesfor theﬁrst time in 2016, and it has since then been part of the routine analyticalrepertoireatourlaboratory.

Fluorofentanylwasdetectedinonedataﬁlefrom2016witha massmatchscorehigherthan97,amassaccuracyof0.21ppm andgoodagreementinthediagnosticions.Analysisofreference

materialshowedanRTdeviationoflessthan0.05min.Moreover,a compoundwithmolecularformulaC23H28N2Owasdetectedina dataﬁlefrom2018withmassmatchscorehigherthan96andmass accuracyof2.87ppm.Thediagnosticfragmentsofm/z105.0699 and 188.1434showedthatthecompound mostprobablywas a fentanyl analogue and the software suggested either cyclopropylfentanyl,methacrylfentanylorcrotonylfentanyl.Thesethree compounds share the same formula and diagnostic fragments.

Consequently,theyarenotpossibletodistinguishfromeachother basedoncategory1criteriaonly,butanalysisofreferencematerial showed good RT agreement (deviation 0.01min.) with cyclopropylfentanyl.Infact,fluorofentanylandcyclopropylfentanylhad alreadybeenconfirmedbytargetedanalysisofthedatafilesbased upon information from analysis of seizures from the scene requestedby thepolice [32,33]. However, asthese compounds would not have beendetected originally ifwe had not known which substances to suspect, theyare included in thepresent material.

Identificationofflubromazepam,phenibut,fluorofentanyland cyclopropylfentanyl (of category 1) was based on the mass accuracyof themonoisotopicMS signal,presenceof diagnostic fragment ions and, finally, RT agreement. Fulfilment of these criteriagavethehighestlevelofconfidencethatcanbeachievedin aretrospectivereviewwhenre-analysisoftheactualspecimenis notpossible.DetectionandconfirmationofcompoundswithHR- MS can be divided in different levels of confidence based on informationavailablefromthedataacquisition,assuggestedby Fig.3.MS1-spectrumofflubromazepamextractedfromadatafile(redlines)withtheoreticalisotopicpatternillustratedbytheblackboxes.(Forinterpretationofthe referencestocolourinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle).

Fig.4.(A)AcquiredMS/MS-spectrumofﬂubromazepamwithdiagnosticfragmentsmarkedwithasterisk(atthetop),libraryspectrumfromPCDL(atthebottom)anda comparison(inthemiddle).(B)AcquiredMS/MS-spectrum(atthetop),fullMS/MS-spectrumfromaﬂubromazepamreferencestandard(atthebottom)andacomparison(in themiddle).

(7)

Schymanskietal.[34].Inthatapproach,level5throughlevel1 requiresincreasinginformationfromtheMSsignaltodiagnostic fragmentsandRTs[34].Findingsofcategory1inourretrospective methodcanbecomparedtoa situationclosetolevel1.Level1 requiresconfirmationwithareferencestandard,whichwas the casewithournewfindings,butaslongasthesampleandstandard are notanalysed simultaneously, a definite confirmation is not achieved.

Inaretrospectiveapproachco-identificationofmetabolitescan furtherstrengthen theconfidenceof afinding.Searchesfor the majormetabolitesofthedetectedcompoundsweredoneinthe relevantdatafiles.Metabolitesfrompublishedinvivoandinvitro studieswereselected[29,35–37]. Neitherof themetabolites of fluorofentanyl were detected in the data file containing this compound. In the data file containing cyclopropylfentanyl the N-dealkylatedmetaboliteandtwohydroxylatedmetaboliteswere detected.Themetabolitesofflubromazepamfoundin literature to be the most abundant (hydroxylated flubromazepam and debrominatedflubromazepam)werenotdetected in anyof the two positive samples. The metabolism of phenibut has to our knowledgenotbeenstudied,andnoputativetargetmetabolites havebeendescribedintheliterature.

Threeotherpositivecategory1ﬁndingscouldberefutedafter furtherinvestigation(Table4).FormethoxyacetylfentanyltheRT deviationcomparedwiththereferencestandardwassigniﬁcant,

indicating thatthecompound ratherwasanisomer ofmethox- yacetylfentanyl withsimilarfragmentationpatterns.Therewere nootherdescribed fentanylanalogues withidenticalmolecular formula.Thepresenceoffragmentsofm/z105.0699and188.1434 washoweverastrongindicatorthatthecompoundconsistedof thepiperidineandphenylmoietycharacteristictofentanylitself as well as many fentanyl analogues. Metabolites of fentanyl hydroxylated at the alkyl or phenetyl moeity have the same monoisotopicmassasmethoxyacetylfentanylandthediagnostic fragments 105.0699 and 188.1434 will be the same (Fig. 5).

Fig.5.Fragmentationofhydroxyfentanyl(left)andmethoxyacetylfentanyl(right).

Thesuperimposedareaindicatespositionofhydroxyl-group.

Table3

Newcompoundsfoundafterapplyingcategory1criteria,includingidentiﬁcationdataandcaseinformation.

Compound(year) Molecular formula

Retentiontime sample/reference standard(D^min)

Mass match score

Diagnostic fragment

Mass (calculated)

Mass accuracy [ppm]

First reported inNorway

Caseinformation

Flubromazepam (2015)

C15H10BrFN2O 9.08/9.15(0.07) 95.55 314.0049 3.46 2013^a Male,approx.30yrs.old.Historyofdrug abuse,founddeadafterdruguse.

Ethanol,amphetamine,metamphetamine, methylenedioxymetamphetamine, metylenedioxyamphetamine,diazepam, desmetyldiazepam,7-aminoclonazepam, alprazolam,pregabalin,mephedrone, buprenorphine,norbuprenorphineand gamma-hydroxybutaratefoundinblood.

C14H11FN2 226.0901 3.29 C7H7BrN 183.9756 5.25 C14H11N2FBr 305.0084 7.10

Phenibut^c (2015)

C10H13NO2 1.53/1.65(0.12) 85.36 180.1019 1.63 2016^b Samesubjectasabove.

C9H9 117.0699 1.81 C10H9O 145.0648 16.67 Flubromazepam

(2016)

C15H10BrFN2O 9.21/9.15(0.06) 97.45 333.0033 0.24 2013^a Female,approx.50yrs.old.Historyofdrug abuse,founddeadathome.

Ethanol,paracetamol,gabapentin,pregabalin, tramadol,O-desmethyltramadol,

amitriptyline,nortriptyline,sertralineand chlorprothixenefoundinblood.

C14H11FN2 226.0901 3.48 C7H7BrN 183.9756 0.95 C14H11N2FBr 305.0084 16.4

Fluorofentanyl^d (2016)

C22H27FN2O 7.18/7.17(0.01) 97.73 355.2180 0.21 2016^b Male,approx.20yrs.old.Founddeadathome withdrugparaphernalia.

7-aminoclonazepam,diazepam, desmethyldiazepam,alprazolam, tetrahydrocannabinolandgamma- hydroxybutyratefoundinblood.

C13H18N 188.1434 5.42 C8H9 105.0699 3.56 C14H17FNO 234.1289 9.92

Cyclopropylfentanyl^d (2018)

C23H28N2O 7.46/7.47(0.01) 96.53 349.2274 2.87 2017^a Male,approx.30yrs.old.Founddeadathome withpillsonsite.

Morphine,morphine-3-glucuronide, morphine-6-glucuronide,buprenorphine, norbuprenorphine,pregabalin,amphetamine, methylenedioxymetamphetamine,

metyhlenedioxyamphetamine,

benzoylecgonine,7-aminoclonazepamand tetrahydrocannabinolfoundinblood.

C13H18N 188.1434 2.52 C8H9 105.0699 1.05 C15H18NO 228.1383 4.46

aDetectedinseizedmaterialbytheNorwegianNationalCriminalInvestigation.

b Detectedinseizedmaterialbyourdepartment.

c4-amino-3-phenylbutyricacid.

d Reportedoriginallybutincludedheretoillustratemethodsuitability.

(8)

Fentanylwasreportedintheoriginalanalysisofthesample,which explainsthepresenceofametabolite.Thus,itcouldbeconcluded thattheﬁndingwascausedbyfentanylintake.

Category1findingsofJWH-167andtilidineweredetectedin onedatafileeach,from2014and2015,respectively.Thefragments intheMS/MSspectrawereinrelativelygoodagreementwiththe diagnosticfragmentsfromthelibraryspectrum,andinaddition them/zClouddatabasewasconsultedandshowedagreementwith oneadditional fragment. Reference standards wereacquired to compareRTsandsignificantRTdifferences clearlyshowedthat neitherJWH-167nortilidinewerepresent.Theseexamplesoffalse positiveresultsillustratetheimportanceofhavingaccesstothe referencesubstanceinordertocheckRTconformity.

3.2.2.Category2ﬁndings

Atotalof35possiblecategory2findingswastheresultwhen applyingcriterionc(betterthan5ppmmassaccuracy,massmatch scorehigherthan95andRT1.5min.orlater).Theinitialfindings arepresentedinTable5.AfurtherevaluationofRT,signal-to-noise ratioandchromatographicpeakshapeforeveryfindingwasdone.

ThemetaboliteAB-FUBINACAM3(#1315),carfentanil(#17and 18),tilidine(#35)andthreeoffourﬁndingsofphenibut(#31,32 and 34) could be disproved due to large RT deviations from reference standards. Based on the RTs of other synthetic cannabinoidsanalysedwiththesamechromatographicconditions (seeTable2)ﬁndingsofsyntheticcannabinoidswithRTslessthan 5min.wereregardedashighlyunlikelyandremovedfromthelist.

Thiswasthecasefor5-fluoro-PY-PINACA(#3and4),5-fluoro-3,5- AB-PFUPPYCA (#5), AB-BICA (#9 and 10) and MA-CHMINACA (#20).5-fluoro-AB-PINACAN-(4-hydroxypentyl)(#2),a metabo- lite and presumably more polar compound than its parent substance,is likely to have a shorter RT. Still, it will probably noteluteasearlyas3.6min.Asimilarlimitof4min.wasappliedon the synthetic opioids which lead to the rejection of 3-fluoro methoxyacetylfentanyl(orocfentanyl)(#1)andtwofindingsofN- methylnorcarfentanil(#25and26).Thesignal-to-noiseratiowas3 or less for AB-CHMINACA 3-carboxylindazol (#11 and 12), N- methylnorcarfentanil(#27)andPB-223-carboxyindole(#30).The initial finding determined as benzyl carfentanil (#16) was disprovedduetopoorpeakshape.Acategory2compoundfound inoneormoredatafiles,andalsofoundwiththesameRTinother datafileshavingMS/MSspectraacquiredbutnolibrarymatch,was likewiserejected.Onesuchexamplewasohmefentanyl,whichwas foundintwodatafileswithRTsof7.8min.Theioncouldalsobe

found in other data files withthe same RT but with acquired MS/MS spectranot in agreement with thePCDL. This strongly indicatedthatthetwofindingsofohmefentanyl(#28and29)were falsepositives.Thesamewasthecasewithpresumablefindingsof AB-FUBINACA(#6–8),JWH-200analog1(orA-796260)(#19)and methoxyacetylfentanyl(#21–24).

Thus,afterreviewingthe35suggestedcategory2findings,only onefindingofphenibut(#33)remained.AsnoMS/MSspectrawere availableforlibrarycomparison,thisfindingcould,however,not be confirmed with the same degree of confidence as those of category1.

3.3.Strengthsandweaknesses

ThePCDLconstructedinthisstudyisbasedondataacquiredon instrumentsfromdifferentmanufacturersandbasedondifferent principles.Apreviousstudyhasshownthatlibrariesconstructed from data acquired on either Orbitrap or QTOF can be used interchangeably by both instruments providing that suitable collisionenergiesareapplied[38,39].Anessentialfeatureofthe PCDL is the mass accuracy of the diagnostic fragments. In HighResNPS the masses of the fragments are added by either typing theformula, selecting thecorrect formula from a drop- downlistofcommonfragmentsortypingthetheoreticalmassof the acquiredfragment. This ensures that masserrors from the acquisitionarenottransferredtothedatabase.Asecondimportant settingisthechoiceofcollisionenergyappliedwhenacquiringthe diagnosticfragmentsthatareaddedtothedatabase.Thecollision energyappliedcan eitherbediscrete (e.g.10,20 and40eV) or ramped,providingacombinedresult.Informationonthechoiceof strategy used in the individual entry was not present in the database. In the Auto MS/MS method used in this study, the collision energy was a voltage correlated to the mass of the precursor. Potentially this can result in differences in relative abundancewhencomparingalibraryspectrumandanacquired MS/MS spectrum. However, the settings in the retrospective reprocessing algorithm ensure a hit even if only one of the diagnosticfragmentionscouldbefoundintheacquiredspectrum.

Theriskoffalsenegativesampleswillalwaysbepresentwhen searchingforcompoundsthat havenotbeensubjecttospeciﬁc evaluation of LOI, which is the case for the majority of the compoundsinthePCDL.Inaddition,theinstrumentresponsehas beenshowntoﬂuctuatetosomeextentduringtheperiodofdata acquisition. Due totherelatively highLOIs and low recoveries Table4

Newcompoundsfoundafterapplyingcategory1criteria,butrefutedas“falsepositive”ﬁndings.

Compound(year) Molecular formula

Retentiontime(RT) sample/reference standard(D^min)

Mass match score

Diagnostic fragment

Mass (calculated)

Mass accuracy [ppm]

Comment

Methoxyacetylfentanyl (2016)

C22H28N2O2 5.13–5.75(0.62) 91.36 353.2224 0.20 RTnotinagreementwithreferencestandard.

Monoisotopicmassanddiagnosticfragmentssuggest fentanylhydroxylatedatthealkylorphenetylmoiety C13H18N 188.1434 3.8

C8H9 105.0699 17.23 C9H12N 134.0964 Notfound JWH-167

(2014)

C21H23NO 11.82–13.86(2.06) 95.32 306.1852 0.90 RTnotinagreementwithreferencestandard C14H16NO 214.1226 7.94

C7H7 91.0542 9.63 C13H18N^a 188.1434 11.05 Tilidine

(2015)

C17H23NO2 9.24–5.56(3.28) 90.87 274.1802 1.65 RTnotinagreementwithreferencestandard C15H17O 229.1223 Notfound

C12H11 155.0855 5.88 C7H7 91.0542 45.1

aDiagnosticfragmentfrommzCloud.