Development of immunoprecipitation – two-dimensional liquid chromatography – mass spectrometry methodology as biomarker read-out to quantify phosphorylated tau in cerebrospinal

In Alzheimer’s disease (AD) brain, one of the histopathological hallmarks is the neurofibrillary tangles consisting of aggregated and hyperphosphorylated tau. Currently many tau binding antibodies are under development to target the extracellular species responsible for the spreading of the disease in the brain.

journalhomepage:www.elsevier.com/locate/chroma

Sebastiaan Bijttebiera,1,∗, Clara Theunisb,1, Farid Jahouha, Dina Rodrigues Martinsb,

Marc Verhemeldoncka, Karolien Grauwenb, Lieve Dillena, Marc Merckenb

a DMPK, Janssen Pharmaceutica, Turnhoutseweg 30, Beerse, Belgium

b R&D Neurosciences, Janssen Pharmaceutica, Turnhoutseweg 30, Beerse, Belgium

a r t i c l e i n f o

Article history:

Received 19 March 2021

Revised 17 May 2021

Accepted 24 May 2021

Available online 28 May 2021

Keywords:

Immunoprecipitation

Two-dimensional liquid chromatography

Metal oxide chromatography

Alzheimer’s disease

Phosphorylated tau

Human cerebrospinal fluid

a b s t r a c t

InAlzheimer’s disease(AD)brain, oneofthe histopathologicalhallmarks istheneurofibrillary tangles consistingofaggregatedand hyperphosphorylatedtau.Currently manytaubindingantibodiesare un-der developmenttotargettheextracellularspeciesresponsibleforthespreadingofthediseaseinthe brain.Assuch,anin-housedevelopedantibodyJNJ-63733657withpicomolaraffinitytowardstau phos-phorylatedatbothT212andT217(furthernamedp217+tau)wasrecentlytestedinphaseIclinicaltrial NCT03375697.Followingmultiple doseadministration inhealthysubjects andsubjects withAD, there weredosedependantreductionsinfreep217+taufragmentsincerebrospinalfluid(CSF)following anti-bodyadministration,as measuredwithanovelsinglemoleculeELISA assay(SimoaPT3xPT82assay), demonstratingepitope engagement ofthe therapeutic antibody[Galpern, Haeverans,Janssens, Triana-Baltzer,Kolb,Li,Nandy,Mercken,Van Kolen,Sun, VanNueten, 2020].Totalp217+taulevels alsowere reducedinCSFasmeasuredwiththeSimoaPT3xPT82assay.Inthisstudywedevelopedanorthogonal immunoprecipitation– liquidchromatography– triplequadrupolemassspectrometry(IP-LC-TQMS)assay

toverifytheobservedreductionsintotalp217+taulevels

Inthisassay,anexcessofJNJ-63733657isaddedtotheclinicalCSFtoensureallp217+tauisbound

bytheantibodyinsteadofhavingapoolofboundand unboundantigenand toimmunoprecipitateall p217+tau,whichisfollowedbyon-beaddigestionwithtrypsintoreleasesurrogatepeptides.Tryptic pep-tideswithmissedcleavagesweremonitoredwhenphosphorylationoccurredclosetothecleavagesiteas thisinducedmiscleavages.Comparedwithacidifiedmobilephasestypicallyusedforpeptideanalysis, re-versedphaseLCwithmobilephaseatbasicpHresultedinsharperpeaksand improvedselectivityand sensitivityforthetargetpeptides.Withthissetupadiphospho-tautrypticpeptideSRTPSLPTPPTREPK∗2 couldbemeasuredwithpT217accountingforatleastoneofthephospho-sites.Thisisthefirsttimethat thepresenceofadiphopsho-taupeptideisreportedtobepresentinhumanCSF.Atwo-dimensional LC-TQMS methodwasdevelopedtoremovematrixinterferences.Selectivetrappingofdiphospho-peptides viaametal oxidechromatographymechanism was achievedinafirst dimension withaconventional reversedphasestationaryphaseandacidifiedmobilephase.SubsequentelutionatbasicpHenabled de-tectionoflowpicomolarp217+taulevelsinhumanCSF(lowerlimitofquantification:2pM),resulting

inanapproximate5-foldincreaseinsensitivity.Thisenabledthequantificationoftotalp217+tauinCSF leadingtotheconfirmationthatinadditiontoreductionsinfreep217+taulevelstotalp217+taulevels werealsoreducedfollowingadministrationofthetaumAbJNJ-63733657,correlatingwiththeprevious measurementwiththePT3xPT82Simoaassay.Anorthogonalsampleclean-upusingofflineTiO2/ZrO2

∗ Corresponding author

E-mail addresses: sbijtteb@its.jnj.com (S Bijttebier), ctheuni3@its.jnj.com (C

Theunis), fjahouh@ITS.JNJ.com (F Jahouh), DRodri39@its.jnj.com (D.R Martins),

MVERHEME@its.jnj.com (M Verhemeldonck), kgrauwe@ITS.JNJ.com (K Grauwen),

LDILLEN@its.jnj.com (L Dillen), MMERCKEN@its.jnj.com (M Mercken)

1 Authors contributed equally to this work

https://doi.org/10.1016/j.chroma.2021.462299

0021-9673/© 2021 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

S Bijttebier, C Theunis, F Jahouh et al Journal of Chromatography A 1651 (2021) 462299

combinedwith1DLC-TQMSwas developedtoconfirmthepresenceofmono-ptau (pT217)tryptic pep-tidesinCSF

© 2021TheAuthors.PublishedbyElsevierB.V ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/)

1 Introduction

By 2050, it is expected that almost 19 million people will be

sufferingfromdementiainEuropeonly,representing3%ofthe

Eu-ropean population[1].WithAlzheimer’sdisease(AD)asthe

lead-ing cause,thereisan urgentneedfordiseasemodifyingtherapies

thatcanpreventorhalttheprogressionofthedisease

Besidesneurodegeneration,neuropathologicalhallmarksforAD

are the extracellular amyloidplaques andtheintracellular

aggre-gates of protein tau Due to the dominantly inherited mutations

in amyloid precursor protein or the presenilin genes PSEN1 and

PSEN2thatwerefoundtocauseAD[2,3],amyloidproteinhasbeen

the main target for disease modifying therapies for many years

Currently,thestronglinkbetweentaupathologyandcognitive

de-cline isrecognizedandtargetingtaupathologyhasbeenincluded

inresearchstrategies[4–6].Theintracellularneurofibrillarytangles

and neuropil threads consist of misfolded and abnormally

phos-phorylatedtauspeciesthatcanbereleasedintheextracellular

en-vironmentandseedpathologyintracellularlyinneighbouringcells,

contributing to the spatio-temporal progression of tau pathology

[7–11] This extracellular tau seed is targeted by active and

pas-sive immunotherapyapproachesthatarecurrentlyunder

preclini-calandclinicalinvestigation

A novel phosphorylated tauselective monoclonal Ab PT3 was

generated with picomolar affinity towards tau phosphorylated at

both T212 andT217 Reducedbinding affinity is observedfor tau

monophosphorylation at either T212 or T217, while the effect of

phosphorylation atS210andS214seemedlimited[12,13].Its

hu-manized variant, named JNJ-63733657 (JNJ’3657) is currently in

clinicaldevelopment[14].Thephosphorylatedtauspeciestargeted

byPT3andJNJ’3657willbefurtherreferredtoasp217+tau

The combinedmeasurements ofa decreaseinAβ1–42andan

increase in totaltauandtauphosphorylatedat T181inCSFwith

immuno-assays,havebeencommonlyusedasdiagnosticand

prog-nosticbiomarkersforAlzheimer’sdisease[16].However,

consider-ing currentlydeveloped therapeutic strategies andongoing

clini-caltrials,additionalbiomarkersareneededforamorepreciseand

even earlierprediction ofdisease onsetanda further increase in

theaccuracyofthediagnosis

Cerebrospinal fluid(CSF) isincontinuousexchange withbrain

interstitial fluid in direct contact with neurons [16] Tau species

foundinCSFhoweverdifferfromtauinbrain:Satoetal.[17]

dis-covered that the predominantforms oftau(99.9%) inCSFare

C-terminally truncated containing the mid-domain but lacking the

microtubulebindingregionandC-terminus,withcleavagebetween

amino acid (AA) residues 222 and 225 As a consequence, only

partofthe humanbraintauinformationispresentinCSF

More-over, theconcentration oftauinCSFisthreeordersofmagnitude

lower than in the brain [17] Barthélemy et al recently

demon-strated by usingliquidchromatography– mass spectrometry

(LC-MS) methodology that T217 phosphorylation (pT217)is

consider-ably increased in CSFof AD patients[15,18] Moreover, T217

hy-perphosphorylationoccurredsystematicallyinCSFofamyloid

pos-itive participants even at pre-symptomatic stage, thus making it

apotentialimportanttargetfor(immuno)therapeuticdevelopment

[15,18].Inanotherstudy,Barthélemyetal.showedthatCSFpT217

outperformspT181asameansofADdiagnosis[19].Thattau

phos-phorylated atT217 inCSFis superior to traditionaltau assaysin differentiatingamyloid status,wasatthesametime also demon-stratedusingasinglemoleculeELISAassaywiththePT3antibody [20,21] This singlemolecule ELISA assay (PT3 x PT82 simoa as-say) could additionallybe used asa target engagementassay for JNJ’3657whencombiningitwithadenaturingtechnique[21] ThePT3 xPT82 Simoaassaywasusedasa targetengagement assay to analyseCSFfrom thephase Iclinical trialwith JNJ’3657 (NCT03375697).Bothfreep217+tauandtotalp217+taulevelswere measured, with total p217+tau being the free p217+tau fraction plusthe JNJ’3657– p217+taucomplexeswhichcan onlybe mea-suredafterdenaturationofthecomplex.Freep217+taulevelswere foundtobereduced,whichshowsepitopeengagementofthelarge molecule [14] Additionally, total p217+tau levels were found to

bereduced,whichsuggestsclearanceofJNJ’3657andthetargeted p217+taumolecules

To verify clearance of total p217+tau levels after dosing of JNJ’3657, an orthogonal immunoprecipitation– liquid chromatog-raphy – triplequadrupole mass spectrometry (IP-LC-TQMS)assay wasdevelopedinour lab.Toachieve the selectivityand sensitiv-ity,necessaryforthedetectionoflow pMp217+taulevels inCSF,

a2DLC-TQMSapproach wasdeveloped.Thissetup wascompared withanofflineTiO2/ZrO2clean-upselectiveforphospho-peptides Thep217+taulevelsinhumanCSFclinicalsamplesweremeasured withtheoptimizedIP-LC-TQMSmethodologyandcorrelationof re-sultswiththeorthogonalPT3xPT82Simoaassaywasexamined

2 Materials and methods

2.1 Materials 2.1.1 Chemicals

Ammonium bicarbonate, formic acid (FA) 98–100%, Tween20, ammoniasolution25%(Suprapur), aceticacid(glacial),lacticacid, glycerol,3-hydroxypropanoicacid,citricacid,glutamicacid, pyrro-lidine and dithiothreitol were supplied by Merck MilliQ water witha resistivity of18.2 M.cm at25 °C wasgenerated with a MilliporeTM-purificationsystem Acetonitrile(ULC-MS)wasbought fromActu-AllChemicals.Recombinanttau-441wasacquiredfrom Promise Proteomics Synthetic peptides H-TPSLPTPPTR-OH, H-(pT)PSLPTPPTR-OH, H-TPSLP(pT)PPTR-OH, H-(pT)PSLP(pT)PPTR-OH, H- (pT)PSLP(pT)PPTREPK-OH, H-SR(pT)PSLP(pT)PPTREPK-OH and H-SR(pT)PSLP(pT)PPTREPK-OH(:Argininelabelledwith13Cand

15N) were purchased from Pepscan Phosphate buffered saline (PBS) was purchased from Roche Antibodies PT9 [12], and PT3 and its humanized variant JNJ-63733657 (JNJ’3657) [13] were produced in-house SCBiot-(dPEG4)-GTPGSR-S210-R-T212-P-S214-LP-T217-PPTREPKK-amide with different phosphorylation patterns

ataminoacidsS210,T212,S214andT217wereobtainedfromNew EnglandPeptides

2.1.2 Biological samples

All animal procedures were done strictly according to the guidelines of the Association for Assessment and Accreditation

of Laboratory Animal Care International (AAALAC), with the Eu-ropean communities Council Directive of 24th November 1986 (86/609/EEC)andwithprotocolsapprovedbythelocalInstitutional AnimalandUseEthicalcommittee

were anesthetized andforebrain wascollected, weighed and

im-mediately put on dry ice A volume of homogenization buffer

(buffer H consisting of 10 mM Tris/HCl, 0.8 M NaCl, 10%

su-crose and 1mM EGTA in MilliQ atpH 7.6 – and 1 tablet/10mL

phosphatase and protease inhibitors (Roche)) corresponding to

six times the weight of the brain tissue was added and

incu-bated on icefor 15 min.The samples were homogenized witha

FastPrep®−24 instrument (MP Biomedicals) at 6.0 m/s for 20 s

The samples were subsequently centrifuged at 20,000 × g for

5 min at 4°C (Heraeus Megafuge 8R centrifuge, Thermo Fisher

Scientific) The supernatant was centrifuged at 34,000 × g for

20 minat4°C(OPTIMATM MAX-XP, BeckmanCoulter) The

super-natantwascollectedastotalbrainhomogenate(BH).Apoolof

ho-mogenatesofforebrainwaspreparedbycombiningequalamounts

ofhomogenatefromeachanimalandvortexmixing

Human CSFsamples withlow, medium and hightau

concen-trationswerepreparedbypoolingde-identifiedCSFsamplesfrom

healthysubjectsandsubjectswithADbasedontheirtotaltau

con-centrationasmeasuredwithELISA(Innotest,Fujirebio)

Foranalysison theeffectofJNJ’3657 onp217+tauinCSF,

hu-man CSFsamples fromthe multiple ascending dosetrialin

sub-jects with prodromal or mild AD (NCT03375697) were used All

CSFsampleswere collectedwithinformed consentfromthe

sub-jects.Inthisstudy,subjectsreceivedeitherplaceboorthreedoses

ofoneofthreedoselevelsofJNJ’3657intravenously Atotalof26

baseline andpost-dose samples were analysed andwere blinded

todoselevel[14]

2.2 Sample preparation

2.2.1 Preparation of standard dilutions

All standard dilutions were prepared in low bind Eppendorf

tubes

Calibrationstandardspikesolutions:astocksolutionof200μM

waspreparedinwater:acetonitrile:aceticacid(89:10:1).Thestock

solutionwasaliquotedandstoredat−80°C.Standardworking

di-lutions were prepared in 90:10:0.1 water:acetonitrile:formic acid

(storedat−20°C)andusedfortheproductionofcalibrationpoints

ranging from 0.5pM to 10 pM asdescribed insupplemental

in-formation1.Standardworkingdilutionswerefreshlypreparedfor

eachbatchofsamples

Spike solutions forbatch acceptanceQCs (supplemental

infor-mation 1) were preparedfromthe samestocksolutions used for

productionofcalibrationstandardspikesolutions

Preparation of stable isotopically labelled (SIL) stock solution

and workingsolutions: astocksolution of200 μM wasprepared

in water:acetonitrile:aceticacid(89:10:1),aliquotedandstoredat

−80 °C SIL working dilutions were prepared in 90:10:0.1

wa-ter:acetonitrile:formicacid(storedat−20°Candfreshlyprepared

foreachbatchofsamples)

2.2.2 Immunoprecipitation

For IP-LC-TQMS assay development, either a 1/20 dilution of

transgenic mouse brain homogenate in artificial CSF (12.4 mM

NaCl, 0.125 mM NaH2PO4.H2O, 0.13 mM MgSO4•7H2O, 0.27 mM

KCl,2.6mMNaHCO3,18mM D-glucose.H2O, 2mMascorbicacid,

2 mMCaCl2 inH2O,pH 7.25)orpoolsofhumanCSFwitheither

low (< 350pgmL−1 tau),medium(350 ><750pgmL−1 tau)or

high level(> 750pg mL−1 tau) of tauwere used andall spiked

withJNJ’3657 antibody tomimic thepresence ofthe dosed

anti-bodyintheclinicalsamples

2.2.2.1 Optimized immunoprecipitation procedure. 93.5 μL of

Dyn-abeads protein G (Thermo Fisher Scientific) corresponding to

2.8 mgbeads was washed two timesin LoBind Eppendorf tubes

(Eppendorf) with 200 μL 0.01%Tween20 in PBS per IP reaction ThetubeswereplacedinaDynaMagTM−2Magnet(ThermoFisher Scientific) andthe supernatant wasdiscarded 56 μL PBS + 0.1% Tween20,100nMJNJ’3657 and500μL(artificial)CSFwereadded

to the beads The samples were mixed and incubated overnight

at4°Cwhilerotatingon ahula mixer(Thermo FisherScientific) Thesupernatantwascollectedasimmunodepletedfraction.300μL (50 mM ammonium bicarbonate (pH8) + 10% of 0.1% Tweenin PBS)wasadded tothebeadsandsubsequentlyvortex mixedand spun (Minispin Plus centrifuge, Eppendorf).The supernatantwas collectedas washand the beads were resuspended in 150μL of

50mMammoniumbicarbonate,pH8.Thesampleswerestoredon meltingicewhilepreparingreagentsfortrypsinization: trypsiniza-tionwasstartedonthedayIPwasfinished

2.2.3 Trypsinization

15 μL of acetonitrile was added to the beads + 150 μL of

50mMammoniumbicarbonate afterIP, vortexmixedfor5 and spun down for 15 with a Minispin Plus centrifuge 15 μL of 0.1 μg trypsin mL−1 in 50 mM acetic acid was added (Trypsin Gold,Promega) andthe samples were subsequently incubatedat

37°Cfor20hwhileshakingat1200 rpm(ThermoMixer,Thermo Fisher Scientific) for on-beaddigestion Afterwards, the digestion was quenched by adding 15 μL of formic acid and briefly vor-texmixing Subsequently, 20μLof SILworkingsolution or20μL blanksolvent(for ‘blank’(supplemental information1):90:10:0.1 MilliQwater:acetonitrile:formicacid)wasadded.Thereafter,20μL

of non-labelled working solution (for calibration points and QCs forbatch acceptance (supplementalinformation 1))or blank sol-vent(90:10:0.1 MilliQwater:acetonitrile:formicacid, forCSF sam-ples)wasspiked.Thesampleswerevortexmixedandcentrifuged

at20,000× gfor10min(HeraeusMegafuge8Rcentrifuge,Thermo FisherScientific)andthesupernatantwastransferredtomicronic tubes(Micronic).Themicronicsweresealedina96wellplatewith sealingmatandstoredat4°Cuntilanalysis

2.2.4 TiO 2 /ZrO 2 clean-up

Clean-up of immunoprecipitated andtrypsinized CSF samples was performed with TiO2/ZrO2 solid phase extraction (SPE) fol-lowed by Oasis HLB SPE TopTips containing 10 mg TiO2/ZrO2 (GlySci) were conditioned with 3 × 50 μL 7.7% FA in 90:10 wa-ter:acetonitrile saturated withglutamic acid(conditioning buffer)

bycentrifugationat2000rpmfor2minwithaMinispinPlus cen-trifuge 55 μL of a MilliQ solution saturated with glutamic acid and containing 7.7% formic acid wasadded to 120 μL of digest The sample was loaded on a conditioned TopTip by centrifuga-tionat1000rpm for5min.Thesamplewasreloadedtwice The TopTip was washed with 50 μL conditioning buffer, 2 × 50 μL 50:50 water:acetonitrile + 2% formic acid and 2 × 50 μL wa-ter (at 2000 rpm for2min) Phospho-peptides were elutedwith

50 μL 10% NH4OH and 3 × 50 μL 5% pyrrolidine (each step at

2000rpmfor2min), eluantswere pooledand15μLformicacid and10μLacetonitrilewasadded.Ammoniaandpyrrolidinewere removedfromthe eluantwith OasisHLB (96-wellPlates, 30 mg,

30μm,Waters).Thestationaryphasewasconditionedwith200μL acetonitrile(centrifugedat600rpm for3min),equilibratedwith 100:2water:formicacid(600rpmfor5min),theTiO2/ZrO2eluant wasloaded (5 min at300 rpm followed by 1 min at 600 rpm), followed by washing with 2 × 800 μL 100:2 water:formic acid and 400 μL water (each step: 4 min at 1000 rpm followed by

2minat2000rpm) Peptideswereelutedwith150μL50:50 wa-ter:acetonitrile(5 minat1000rpm and2minat2000rpm) The eluantwasevaporatedto drynesswithnitrogengasat45°Cand redissolved in 150 μL of 90:10:0.1 water:acetonitrile:formic acid andstoredat4°Cuntilanalysis

S Bijttebier, C Theunis, F Jahouh et al Journal of Chromatography A 1651 (2021) 462299

Fig 1 Schematic depiction of the column connections to the 6-way valve in the column oven Dotted line: valve position 0, solid line: valve position 1

2.3 LC-TQMS analysis

Both 1DLC- and 2DLC-TQMS analyses were performed on an

ultra-highperformanceliquidchromatographfromShimadzu

con-sisting of4 NexeraLC30ADliquid chromatographssetup to

pro-videdualbinarysolventgradients,aSIL-AC30autosampler,a

CTO-20ACcolumnovenwithanintegrated6-wayvalve,a

communica-tions bus module (CBM-20A) anda sampleRack Changer II,

hy-phenatedviaaTurbo-IonsprayTM Interface(Sciex)toa6500triple

quadrupole massspectrometer(Sciex).A separateNexera LC20AD

liquidchromatograph(Shimadzu)wasusedforpost-column

addi-tion of 100 μL min mL−1 acetonitrilevia a T-piece.Analyst 1.6.3

(Sciex) wasused asinstrument control anddataprocessing

soft-ware

2.3.1 1DLC-TQMS analysis

For 1DLC-TQMS analysis, 50 μL of digest was injected on

an ACQUITY UPLC Peptide BEH C18 Column, 300 ˚A, 1.7 μm,

1 mm × 100 mm (Waters) and thermostatically (60 °C) eluted

Themobile phase(MP)solventsconsistedof100:1 water+0.05%

ammonia:acetonitrile (v:v) (A)and acetonitrile+ 0.05% ammonia

(v:v) (B), and the gradient was set as follows (min/A%): 0.0/100,

0.5/100,5.0/70,5.1/2,6.6/2,6.7/100,10/100.Theflowratewasset

at0.2mL/min.Theprobeverticalmillimetre settingwasadjusted

to 8 mm to improvesensitivity The peptides were ionized with

electrospray ionisation (ESI) in positive ion mode The ionspray

voltagewassetto4500V,temperatureto400°C,declustering

po-tentialto60Vandentrancepotentialto10V.Ionsourcegas1,gas

2andcurtaingasweresetto50,40and30,respectively.CADgas

wassetto6.TheselectedMStransitionsusedformultiplereaction

monitoringofthetargetpeptidesareprovidedinsupplemental

in-formation2

2.3.2 2DLC-TQMS analysis

For 2DLC-TQMS analysis, 50 μL of digest was injected

An ACQUITY UPLC Protein BEH C4 Column, 300 ˚A, 1.7 μm,

2.1 mm × 50 mm (Waters) was used in a first dimension with

water+ 0.1%formicacid(A)andacetonitrile(B)asmobilephase

solvents and a flow rate of 0.4 mL min mL−1 In a second

di-mension an XBridge Peptide BEH C18 Column, 300 ˚A, 3.5 μm,

1 mm× 100 mm(Waters) wasusedwithmobile phasesolvents

consisting of 100:1 water + 0.05% ammonia:acetonitrile (v:v) (C)

and acetonitrile + 0.05% ammonia (v:v) (D) and a flow rate of

0.2mLminmL−1.Columnoventemperaturewassetat60°Cand

the connections of the 6-way valve were as depicted in Fig 1

The LC-gradient and valve switching time program are described

inTable1.ThesameMSsettingswereusedasduring1DLC-TQMS

analysis

LC-gradient and valve switching time program of the 2DLC-TQMS method Mobile phase solvents A and B are connected, and mobile phase solvents C and D are connected

Time (min) Module Events Parameter

20 System Controller Stop

2.3.3 Analysis of clinical samples

ClinicalCSFsamples(n=26)fromthemultipleascendingdose trialinsubjectswithprodromal ormildAD (NCT03375697)were processed with the optimized IP-2DLC-TQMS protocol CSF sam-pleswerecollectedatdifferenttimepointsafterdosingwitheither placebooroneofthreedoselevelsofJNJ’3657[14].Qualitycontrol criteriaforthe optimizedIP-2DLC-TQMS protocolwere used dur-ing analysisofhuman CSFsamples asdescribedinsupplemental information1 Correlationbetweenresults obtainedfor surrogate peptide/p217+taulevelsintheclinicalCSFsampleswith IP-2DLC-TQMSandthePT3xPT82 Simoaassaywascalculatedby usinga linearregressionaftertransformingthedata(X=Log(X);Y=Log(Y))

3 Results and discussion

3.1 Digestion – formation of tryptic peptides with missed cleavages

During the current study, an IP-LC-TQMS assay was devel-oped to determine low pM quantities of p217+tau in human CSF andbrain tissue homogenates of mice Detection of

double-or triple-phosphorylated tau tryptic peptides have not been re-portedup to now astheir stoichiometry isassumed tobe lower thanmono-phosphorylatedpeptides,unlessthereisbiological co-ordination (e.g priming effect) of site phosphorylation [23] The

Ab PT3 and its humanized variant JNJ’3657 however exhibit pi-comolar affinity for tau phosphorylated at both T212 and T217, while its affinity for tau monophosphorylated at T212 or T217

is respectively 16-fold and 6-fold lower Phosphorylation at S210 and/or S214 has only a limited effect on the binding affinity

Table 2

Tryptic digests of elongated peptide SCBiot-(dPEG4)-GTPG-S210-R-T212-P-S214-LP-T217-PPTREPKK-amide with different phosphorylations show different cleavage patterns Results are expressed as peak areas of the different tryptic pep- tides formed, relative to the peak area of the tryptic peptide with the highest peak area Data are recorded on an UHPLC –HRMS system in TOF MS mode (660 0, Sciex), mass range: m/z 30 0–180 0 The 1DLC-method described in sec- tion Materials and Methods was used ∗ loss of phospho-moiety at S210 by tryptic cleavage at R211 AAs in bold represent the elongation due to missed cleavage

Phosphorylation – T217 S214/T217 T212/T217 T212/S214 T212/S214/T217 S210/S214/T217

[13] Potential capture of tau with double and triple

phospho-rylation in the Ab epitope region was therefore also considered

in this project A set of elongated peptides containing the tau

epitope region of the Ab

(SCBiot-(dPEG4)-GTPG-S210-R-T212-P-S214-LP-T217-PPTREPKK-amide,AAnumberingbasedonfulllength

2N4R tau)withdifferentphosphorylationpatterns ataminoacids

S210, T212, S214 and T217 was available from epitope mapping

experiments, all biotinylated at N-terminal and amidated at

C-terminal These elongated peptides were used to investigate the

influence of phosphorylation on formation of miscleavages

dur-ing trypsinization Separate dilutions of the elongated peptides

were trypsinized overnight.Table 2 showsper elongated peptide

the peak areas of the different tryptic peptides formed, relative

to the peak area of the tryptic peptide with the highest peak

area These data indicate that formation of miscleavages during

trypsinization isdependantonthesiteofphosphorylation,as

de-scribedpreviouslybyothers[24].Thenegativechargeof

phospho-rylated serine or threoninepositioned next to the basic arginine

or lysine forms salt bridges and competes withthe

complemen-taryasparticacidatthetrypsinactivesite[25].Forexample,when

a peptide isphosphorylated at T212, tryptic peptides miscleaved

at R211 (based on full length 2N4R tau) are abundant (Table 2),

forpeptidesphosphorylatedatT212/T217andT212/S214

predom-inantly tryptic peptides with miscleavage at R211 are detected,

while for phosphorylation at S214/T217 the tryptic peptide with

0 miscleavages is most abundant Additionally, when comparing

theresultsforpeptidesphosphorylatedatpT212/pS214/pT217and

pS210/pS214/pT217,phosphorylationattheN-terminalsideofR211

(S210)seemstohavelessinhibitoryactivityontrypticcleavageat

R211 than phosphorylation atits C-terminal side (T212) (peptide

[Btn]-GTPGSR(pS)RTP(pS)LP(pT)PPTR was not detected)

Replace-ment oftrypsin withanother enzymewith cleavagesites not

in-fluencedbythephosphorylationcouldpotentiallyresultinthe

for-mation ofpeptides withoutmissedcleavages,thereby simplifying

dataanalysis.Thiswasnotexploredinthecurrentwork

Itisknownthatwhenglutamicacidorasparticacidislocated

next toa trypsin cleavagesitea similar inhibitory mechanismis

exhibited[25].IntheAAsequenceoftheelongatedpeptidesunder

study, glutamic acidispresentnext to the R221trypsin cleavage

site(basedonfulllength2N4Rtau),enhancingtheprobabilityofa

missedcleavagebeingformed.Missedcleavagesatthislocationof

thetausequencehavebeenreportedpreviously[15–17].Formation

of miscleavages should thereforebe considered in thesearch for

p217+tautrypticpeptidesinBHandCSFsamples

3.2 1DLC-TQMS analysis

Trypticdigestsoftheelongatedpeptideswereusedasreference

during 1DLC-TQMS method optimization, next to a set of

(non-)phosphorylated synthetic peptide standards representing tryptic

taupeptideswith0miscleavages(TPSLPTPPTR).Mostoften,acidic

mobile phases(e.g 0.1% FA) are used for the separation of

pep-tidemixtureswithLC.However,broadpeakshapeswereobserved

forthe peptides of interest when using0.1% FA as mobile phase additive Moreover, this mobile phase composition rendered low chromatographic resolution of the target peptides, which consist

of multiple isomeric compounds differing only in phosphoryla-tion site In case of chromatographic overlap, diagnostic product ionsare neededfordifferentiationoftheseisomers, thereby lim-iting product ion selection options (e.g selection based on ab-senceofmatrixinterferences,sensitivity).Manystudiesreportthat phosphategroupsare easily lostduringcollision-induced dissoci-ation, asphospho-peptidesare often preferentially fragmentedat thephospho-sitestherebymakinglocalizationofthephospho-site challenging[26].Duringthe currentwork, predominantlyproline (Pro) y-ions were observed: the Pro-effect is a well-known frag-mentation inMSMSspectraofpeptides, inwhichselective cleav-age commonlyoccurs at the N-terminal side of Pro-with mobile protonstoformabundanty-ions.[27]

Nexttobroadpeakshapesandlowchromatographicresolution obtainedwithacidified MP, highcarry-over levels were observed for the target phospho-peptides, most probably due to interac-tionsofthephospho-moietieswithstainlesssteelandfreesilicaat low pH[28].It hasbeenstatedthatthe useofhigherpHmobile phasescan limittheseinteractions [28,29].Othershavesuggested thattheseinteractions havenorelationtopH asindicated bythe ionic condition ofphosphate compounds, and because the inter-actionscan besuppressed bymakinguseofmetal chelatorssuch

asEDTAandmedronicacidorion-pairingreagents[29,30].Asthe peaks observed for the non-phosphorylated target peptides (e.g TPSLPTPPTR)werealsobroadwhenusingacidicmobilephaseand

asthiswasnotthecaseforothertryptictaupeptides(investigated withrecombinanttaudigests onHRMS,datanotshown),this in-dicatesthat peakbroadeningofthetargetpeptides isalsorelated

toits aminoacidsequence.It hasbeenreportedthat thiscan be caused by slow cis trans peptide bond isomerisation of Pro-Pro-moieties[31].Mostpeptidebondsoverwhelminglyadoptthetrans

isomericformunderunstrained conditionsmainly becauseofthe weakerstericrepulsioneffects,butpeptidebondstoN-substituted aminoacidssuchasProcanpopulatebothisomers[32].Becauseof thepartial doublebondcharacterofthe amidebond and reason-able high barrier of conformational transformation, the cis –trans

isomerization of peptide bonds is a relatively slow process [32] Increasein column temperature howeverincreases isomerization rates,resultinginsymmetricpeaksforpeptides containinga Pro-Pro-moiety[31].Duringthiscurrentstudytheinfluenceofcolumn temperatureandpHonchromatographyofthetargetpeptideswas investigated

TheresultsinTable3showthatpeakwidthsatpH3are simi-larforTPSLPTPPTRandTPSLP(pT)PPTRat25°C.Whenthecolumn temperatureisincreased,peakwidthsofthenon-phospho-peptide decrease to 6 s, which is in agreement with the increasing iso-merization rates describedby Griffits andCooney [31] The peak widthsofTPSLP(pT)PPTRhoweverremainconstantwithincreasing columntemperature:thisis mostprobablyduetointeractions of thephospho-moietywithfreesilanolsandstainlesssteel

Increas-S Bijttebier, C Theunis, F Jahouh et al Journal of Chromatography A 1651 (2021) 462299

˚ A, 1.

column te

ingtheretentionbyloweringthegradientspeed(from1%to30%B

in15mininsteadof5min)resultsindoublingofthepeakwidths

at25°C.Thiscanbeexplainedbythelongerresidencetimeofthe peptidesinthecolumn(retentiontimesapproximatelydouble)and thereby moretimeforpeak broadeningbecauseofisomerization Peakfrontingalsoincreaseswithlongerresidencetimeinthe col-umn.Similartowhenafastgradientisapplied,increasingcolumn temperatureresultsinsharperpeak widthsforthe non-phospho-peptide(from21 to9.6s)whilethepeakwidthofTPSLP(pT)PPTR remainsconstant.IncontrastwithchromatographyatpH3,peaks

of the TPSLP(pT)PPTR peptide eluted at pH 11 become narrower whenthecolumntemperatureisincreased,indicatingless interac-tionsofthephospho-moiety withthe stationaryphase and stain-lesssteel

Nextto that, changeof themobile phase pHalso resultsin a shift in selectivity The peptides TPSLPTPPTR and TPSLP(pT)PPTR co-elute at pH 3 while at pH 11 they are easily separated: TP-SLP(pT)PPTR elutes earlier, hence its narrower peak widths ob-tained at pH 11 (Table 3) Chromatography was compared with

a largerset ofreferencetau peptidesto investigatethe influence

of the mobile phase pH on selectivity (Fig 2) Best results were obtainedwith0.05% ammoniumhydroxideasmobilephase addi-tive(pH11),renderingsharpchromatographicpeaksand separat-ingtargetpeptidesbasedondegreeofphosphorylation(following the trendforcapacity factor:non-phospho-> mono-phospho- >

di-phospho-peptides) Changes in the charge-states of phosphate groupsand (e.g basic) amino acidsby changes inpH contribute

tothe overall netcharge andthereforehydrophilicityof peptides thereby altering chromatographic selectivity and retention [33] Nextto increasedselectivityandsmaller peakwidths, less carry-overwasobservedatpH11.Moreover,an average5-foldincrease

in peak area andheight wasobserved forboth TPSLPTPPTRand TPSLP(pT)PPTR Therefore,ammonium hydroxide was retained as mobilephase additiveforfurtherexperiments.Tooptimize sensi-tivity,anACQUITY UPLCPeptide BEHC18 Column,300 ˚A,1.7μm, 1.0mm× 100mmwasusedwithmobilephaseflowrateof0.2mL min−1 andpost-column addition of 0.1mL min−1 acetonitrile to enhancemassspecionizationefficiency.Notwithstandingthesmall innerdiameterofthecolumn,injectionsof50μLofstandardsand samplesrenderedsharppeaksforthepeptidesofinterest

3.3 Optimization of IP and digestion of diluted mouse BH and human CSF

TobeabletoquantifylowpM totalp217+taulevelsinhuman CSFderivedfromthemultipleascendingdosetrialinsubjectswith prodromal or mild AD (NCT03375697) with LC-TQMS,an IP pro-tocolwithJNJ’3657wasdeveloped.TheseclinicalCSFsamples al-readycontainJNJ’3657derivedfromthetreatment.Asdescribedin [14],thedosedantibodyisboundtop217+tauinCSFresultingin

afractionofp217+tauincomplexwithJNJ’3657anda fractionof freep217+tau.Thefractionoffreep217+tauinCSFdecreasesafter antibodydosingin adosedependant manner, asmoreandmore antibodywillformacomplexwithp217+tau.Totalp217+tau(i.e thesumofp217+tauincomplexwithJNJ’3657andfreep217+tau) also was observed to be decreased In order to measure the to-talp217+taulevelswithimmunoassays, JNJ’3657cannot be cova-lentlybound to thebeads prior tosampleaddition, asthen only freep217+tauwouldbecaptured.Instead,aprotocolwasdesigned

inwhichJNJ’3657isfirstaddedinexcesstothesampletobindall p217+taupresentafterwhichmagneticbeadswithproteinGwere used to captureall JNJ’3657 either bound or not to p217+tau.A scheme ofthe samplepreparation procedure is provided in sup-plementalinformation3.AshumanCSFsamplesarescarce,apool

oftotalhomogenatesofforebrainof3-month-oldP301Ltau trans-genicmice[22] wasprepared.Thesemiceoverexpresshumanfull

Fig 2 Chromatography of reference peptides representing the target tryptic peptides of p217 + tau and tau, analysed with the 1DLC-MSMS method described in Materials

and methods, using an XBridge peptide BEH 300 ˚A, 3.5 μm, 1.0 mm × 100 mm Lower chromatogram: 0.1% FA in mobile phase Upper chromatogram: 0.05% NH 4 OH in mobile phase

length2N4Rtauunderthethy1promotoranddonotdevelop

ag-gregated forms of tauat thisage As a humanCSF surrogate for

method optimization, the BH pool was diluted 20-fold in

artifi-cial CSF.Experiments were conductedtomake sureall p217+tau

wascapturedandtoincreasemethodsensitivity,byoptimisingthe

amount ofJNJ’3657,magneticproteinG beads,sampleintake and

volume ofammonium bicarbonate buffer added after IP A

com-parison wasmadebetweenon-beaddigestionafter IP and

diges-tion following tauelution after IP.The elution protocol consisted

of5minofboilingat98°Cinthepresenceofdithiothreitol(DTT),

immediatelyfollowedby10mincooling ofthesampleinice

wa-ter anda centrifugation stepof 30min at20,000× g As tau is

a heatstableprotein[34] itwillremainin thesupernatantwhile

mostother proteins,including thecaptureantibodyJNJ’3657,will

denature andbe centrifuged down tothe pellet When

perform-ingon-beaddigestion,moresubstrates(bothproteinGandAb

to-gether withptau)fortrypsinareavailable,potentially resultingin

less efficientdigestionof p217+tauandmore matrixeffects

dur-ing LC-TQMSanalysis.Ontheother hand,taucouldbe partlylost

duringelutionafterIP,e.g.bynon-completeelutionorbyinclusion

betweenthedenaturedantibodyaggregates Moreover,itwas

ob-servedthatDTTpartlyinhibitedtrypsin,potentiallyleadingto

vari-ability, andthatmatrix effectswere still presentduringLC-TQMS

analysis ofsampleextracts obtained withdigestionfollowing tau

elution Itwasdecidedtouseon-beaddigestioninfuture

experi-mentsandminimizematrixeffectsviachromatography

An experiment was conducted to compare p217+tau tryptic

peptidesdetected inthe20-folddilutedBHpool andpoolsof

hu-manCSFsampleswithlow,medium orhightotaltaulevels(2, 6

and 11 pM ptau, respectively, asmeasured with the PT3 x PT82

simoaassay [21]).Thesampleswere IP-edwithJNJ’3657 andtwo

humanIgG1antibodies(isotype control1and2) asnegative

con-trolstoassessnon-specificbinding(Table4).Whenp217+taufrom

diluted BH samples were IP-ed with JNJ’3657, both

monophos-phorylated (pT217)anddoublephosphorylatedtrypticpeptidesof

tauweredetected.ForpT217,bothTPSLP(pT)PPTR(0miscleavages)

andTPSLPTPPTREPK p(one miscleavageat C-terminalside,

phos-phorylation site undetermined) were detected The double

phos-phorylated tryptic peptide was only detected with two missed cleavages(SRTPSLPTPPTREPK 2p).Thisformation ofmissed cleav-ages agrees with what was observed with the set of elongated p217+taupeptides:inhibition oftrypsinizationdueto phosphory-lation.In the negative control samples andwhen using JNJ’3657, non-phosphorylatedtautrypticpeptidesarealso detected.Morris

etal.alsodetectedatrypticpeptidediphosphorylatedmostlikely

atT212andT217inmouseBH[35].Whenanadditionalwashstep wasconductedafterIP, thepeak areasofnon-modified tau tryp-tic peptides diminished, suggestingmainly antibody independent binding ofnon-phosphorylatedtauto thebeads.Some low affin-itybindingofnon-modifiedtauwithJNJ’3657duringIPcannot be completely excluded asnon-phosphorylatedtau ispresent in ex-cessoverp217+tauinCSFandbrainhomogenates

Alsoin human CSFpools, double phosphorylated tryptic pep-tideswere detected(SRTPSLPTPPTREPK2p).Moreover,anincrease

in peak areas was observed for SRTPSLPTPPTREPK2p in the or-der low<medium<high tau CSF samples Detection of tau dou-ble phosphorylated around T217 has never been reported before

in CSF It was not expected to be present in high enough stoi-chiometry to be able tobe detectable The presence ofthis pep-tide inhumanCSFwasconfirmedwithmultipleMRM-transitions andon differentLC-TQMS instruments.Nonetheless, atthisstage

MS fragmentation did not allow to unambiguously confirm the phosphorylation sites of the SRTPSLPTPPTREPK2p peptide Nei-therwere theSRTPSLPTPPTREPK2ptrypticpeptidesofthe above-mentioned elongated peptides with double phosphorylation at T212/S214,T212/T217orS214/T217chromatographicallyseparated

In contrast with mouse BH samples, no non-phosphorylated tryptic tau peptides were detected in human CSF samples after IP:thesedifferencesinnon-specificbindingcanbetentatively ex-plained by on one hand the larger amountof tau proteinin the

BH samples andon the other hand by the different tau species presentinthe samplematrices Tau speciesin mouseBHconsist

offull-lengthtauwhiletauinhumanCSFistruncated,potentially affectingnon-specificbinding.Anotherdifferencebetweenthetwo sample matrices was that in human CSF samples an interfering peakinthetraceofTPSLP(pT)PPTRwasdetected, alsointhe

neg-S Bijttebier, C Theunis, F Jahouh et al Journal of Chromatography A 1651 (2021) 462299

IP-1DLC-TQMS analysis of 20-fold diluted P301L tau transgenic mouse brain homogenate samples and human CSF pool samples with low, medium and high tau levels,

as determined with Simoa Immunoprecipitation with JNJ’3657 and non-tau Abs as negative controls (Isotype control 1 and 2) Values in peak areas ND: not detected SRTPSLPTPPTREPK, (pT)PSLPTPPTR, SRTPSLPTPPTR ∗ p, SRTPSLPTPPTREPK ∗ p, SRTPSLPTPPTR ∗ 2p and TPSLPTPPTREPK ∗ 2p were not detected in the samples Additional wash step was performed with 300 μL 0.1% Tween in PBS

TPSLPTPPTR TPSLPTPPTREPK TPSLP(pT)PPTR TPSLPTPPTREPK ∗ p SRTPSLPTPPTREPK ∗ 2p P301L mouse brain

homogenate

High Tau - Isotype control 1

High Tau - Isotype control 2 additional wash

ative controls samples Due to this interference, the presence of

taumonophosphorylatedatT217inhumanCSFcouldnot be

con-firmed.Notwithstandingthatthedeveloped1DLC-TQMSmethodis

highlysensitiveforthedetectionofthetarget peptidesinsolvent

(LLOQsatlowpM level);ionisationsuppression,interferingpeaks

and elevated noise levels during sample analysis reduced

sensi-tivity These matrix effects were predominantly caused by

tryp-tic peptides ofprotein G and Ab,and detergents used during IP

(Tween20).OtherdetergentsmoresuitableforLC-MSanalysissuch

asoctylβ-D-glucopyranosidewillbetestedinfutureexperiments

At this stage, it was decided to further optimize the LC-TQMS

method withfocus on thequantification ofthe double

phospho-rylated peptide in digests of IP-ed CSF Since JNJ’3657 has

high-est affinity fortauphosphorylationatT212/T217, itwas

hypothe-sized that SRTPSLPTPPTREPK2p ismostprobably phosphorylated

atT212/T217.Areferencestandardandstableisotopicallylabelled

standardofSR(pT)PSLP(pT)PPTREPKwerepurchased

3.4 2DLC-TQMS analysis

A 2DLC-TQMS method wasdeveloped to increase method

se-lectivity andminimize matrixeffects(ionizationsuppression,

ele-vated noiselevelsandinterferingpeaks).RPLCwithmobilephase

solvents atpH 11waschosen assecond dimensionbecauseofits

superior selectivity and sensitivity Hydrophilic interaction

chro-matography (HILIC), electrostaticrepulsion hydrophilicinteraction

chromatography (ERLIC), weak anion exchange chromatography

(WAX) and online metal oxide chromatography were evaluated

asfirstdimension.However,preliminaryoptimisationexperiments

didnotshow obviousaddedvalue ofthesetechniquesina

2DLC-TQMSsetup for the currentapplication RPLC withacidified

mo-bile phase solvents was evaluated as first dimension because of

its alternativeselectivitytowardstarget peptidesandmatrix

com-pounds(evaluatedwithhighresolutionMS,datanotshown)

Dur-ing optimization of chromatographyof target peptides on

differ-ent stationary phaseswith acidified (0.1% FA) mobile phase

sol-vents, it was noticed that diphospho-peptides exhibited a

pecu-liar elutionbehaviour.The diphospho-peptideselutedasexpected

when using an ACQUITY UPLC Peptide BEH C18 column, as

de-scribed above.However, diphospho-peptidesdidonly partlyelute

whenusinganACQUITYUPLCCSH phenyl-hexylstationaryphase

Moreover,they didnoteluteatallwithanACQUITYUPLCProtein

BEHC4orACQUITYUPLCShieldRP18stationaryphase,evenwhen

a gradient upto 98% acetonitrilewasapplied.Non-phospho- and

monophospho-peptideswerenotretained.Thedivergentretention

behaviour of diphospho-peptides can be explained by chemical differences between stationary phases The ACQUITY UPLC Pep-tide BEH C18 stationaryphase is endcapped, thereby minimizing potential interactions with free silica, while the ACQUITY UPLC Protein BEH C4 phase is non-endcapped and the ACQUITY UPLC CSH phenyl-hexyl and ACQUITY UPLC Shield RP18 phases con-tain a positive surfacecharge andan embedded hydrophilic car-bamate group, respectively We propose a retention mechanism herewherebythetargetdiphospho-peptidesareretainedeitherby complexationwithsurface-immobilizedmetal ions(e.g.FE(III))or directlywithembeddedpositivecharges(carbamategroupsare ex-pectedto be chargedatlow pH).Thisis inlinewithother stud-ies that reported strong retention and difficult elution of multi-plephosphorylatedpeptidesduetointeractionwithmetalionson the surfaceof the stationary phase, through an interaction simi-lartoFe(III)-IMACandCr(III)-IMAC[30,36].Liuetal.describethat solvents,stainlesssteel, glasswareandC18materialare sourcesof ironandaluminium[36].DePraetal.reportedthatevenwith bio-compatibleHPLCspeaktailingcanoccurduetotitaniumleaching from the system (investigated with fluoroquinolones) [37] Since themajorportionofsilanolgroupsisexpectedtobeneutral(Si— OH)under theexperimental conditions,hydrogen bonding rather thanion-exchangeinteractionisbelievedtobethemajor contribu-tionfromtheresidualsilanolgroups[38].However,evenatpHas lowas2.5thefractionofnegativelychargedsilanolscancause im-mobilisationofmultiple-chargedtitaniumcationsbyion-exchange, resultinginthe formationofa positivelycharged environmentat thestationaryphasesurface[37].Themoreresidualsilanolgroups are available, the more predominant these interactions become

Asdescribedabove,phospho-peptidesdidelutefroman ACQUITY UPLCPeptide BEH C18 stationaryphase withan acidified mobile phase,althoughbroadpeakshapeswereobtained Thiseffectcan

be tentativelyexplainedby thelow availability offree silanolsin thisstationaryphase,thusminimizingmetalioninteractions

As focuswasset on analysisofdiphospho-peptidesin human CSF, ion-exchange interaction of diphospho-peptides with immo-bilized metal ionswasused toour advantage: an ACQUITYUPLC ProteinBEHC4 stationaryphase withacidified (0.1%FA)MPwas usedtoselectivelytrapdiphosphorylatedpeptides.Tests with dif-ferent ACQUITYUPLC ProteinBEHC4 columns confirmedthe re-tentionbehaviour.Elutionfromthe first dimensioncolumn, refo-cusing on a second analytical column (XBridge peptide BEH C18

1 × 100 mm, 300 ˚A, 3.5 μm) and subsequent elution was per-formedunderbasicmobilephaseconditions(0.05%ammonia).The

Fig 3 Analysis of SRTPSLPTPPTREPK ∗ 2p in a human CSF sample after IP and trypsinization with 1DLC-TQMS (left) and 2DLC-TQMS (right)

optimized 2DLC-TQMS setup enables to remove matrix

interfer-ences,resultinginnarrowpeaks(4s)andanappr.5-foldincrease

in sensitivity(LLOQ of2 pM in CSF)for SRTPSLPTPPTREPK2p in

comparisonwith1DLC-TQMS(Fig.3)

3.5 TiO 2 /ZrO 2 clean-up to enable 1DLC-TQMS of target

phospho-peptides

1DLC-TQMS sensitivity of monophospho-peptides was

lim-ited by matrix compounds originating from sample

prepara-tion The optimized 2DLC methodology did not allow to confirm

the presence of tau phosphorylated at T217 in human CSF as

the monophospho-peptides are not retained on the C4-column

Notwithstanding these mono ptau peptides were previously

re-ported by Barthélemy et al [15,23] A sample clean-up protocol

was developed in our lab with TiO2/ZrO2 SPE in order to

con-firm the presence oftau phosphorylation atT217 in human CSF

In contrast to online metal oxide chromatography, its offline

ap-plicationisnotconstrainedbythechemicalsused,thatcould

oth-erwise decrease sensitivityof theLC-MS systemor limit lifetime

of instrument consumables Metaloxides such as TiO2 andZrO2

show HILIC properties as well as anion exchange properties

un-der acidic conditions: below pH 2.7, the carboxy group at

Asp-andGlu-residuesandtheC-terminusoftrypticdigestsarelargely

undissociatedandtheaminogroupatLys-,His-,andArg-residues

and the N-terminus are positively charged [39] Therefore,

non-phosphorylatedpeptidesincludingordinaryacidiconesarelargely

unretained by a WAX column because of electrostatic repulsion

between the solute and thesolid phase, while phospho-peptides

are slightly retained since phosphate groups are dissociated

un-der these conditions[39] Nonetheless,Asp- orGlu-rich peptides

were proven to exhibit non-specific binding to TiO2 beads [26]

Variousacidicadditivessuchaslacticacidandglutamicacidhave

been reportedtodecrease non-specific bindingof acidicpeptides

because of competition for interaction with metal oxide binding

sites[40].Itwasshownthat theseadditivesdonot competewith

phospho-peptide binding, probablybecauseof adifferent

geome-try of phospho-peptide binding comparedwith non-specific

pep-tidebinding[40,41].Duringthisstudyseveraladditives(lacticacid,

glycerol,hydroxypropanoicacid,citricacidandglutamicacid)were

tested in the loading solvent to minimize non-specific binding:

based on the recovery of target phospho-peptides and removal

ofmatrixinterferences(matrixpeptidesanddetergents),glutamic

acid provided an overall bestperformance The releaseof bound

peptidesduringphospho-peptideenrichmentisusually donewith

alkaline solutionscontainingforexampleammoniumbicarbonate,

ammoniumhydroxideorpyrrolidine:differenteluentsprovide

dif-ferentphospho-peptidespectraandsuccessiveelutionwithvarious

elution buffers can significantly improve phospho-peptide

recov-ery[42].Fukudaetal.observedthat thereislittlecorrelation

be-tweenthemaximalphospho-peptiderecoveryandthesolventpH: the authors suggested that effective elution of phospho-peptides

is also caused by nucleophilicity of amine groups against metal ions [43] We developed a consecutive elution with 10% ammo-niumhydroxideand5%pyrrolidine,followedbyanadditionalSPE step to remove ammonia and pyrrolidine from the purified ex-tract.Thisclean-upenableddetectionofTPSLP(pT)PPTRinhuman CSF samples IP-ed with Abs PT9 (total tau Ab with epitopes in theprolinerichregion)[12],andJNJ’3657.Asmatrixinterferences were removed,a moresensitive transitioncould be used,namely 573.8–>650.0 with product ion corresponding to ‘y6 phospho moiety’, instead of573.8–>748.2 withproduct ioncorresponding

to‘y6’fragmentation TPSLP(pT)PPTRischromatographically sepa-ratedfromitsisomersphosphorylatedatothersitesallowingusage

oftheproduction‘y6 phosphomoiety’forunambiguous identi-fication This increasedsensitivity atleast 5-fold (Fig 4) No im-provement in sensitivity was observed for SRTPSLPTPPTREPK2p during 1DLC-TQMS analysis, indicating that not all matrix inter-ferenceswereremovedwiththeclean-up.Forthispeptide, 2DLC-TQMSprovedtobethemostsensitiveoption

3.6 Analysis of clinical samples

Theproject goalwasto confirmthe declineoftotal p217+tau levels overtime in CSFfrompatientsdosed withJNJ’3657 inthe multipleascending dosetrial insubjectswith prodromalormild

AD (NCT03375697), asmeasured withthe PT3 x PT82 Simoa as-say[21] Weused the optimized2DLC-TQMSmethodology as or-thogonalassaytomonitorSRTPSLPTPPTREPK2passurrogate pep-tide for p217+tauin human CSF clinical studysamples The im-munoprecipitationprocedures weredone asdescribedinthe Ma-terialsandMethodssection.Inshort,optimizedlevelsofJNJ’3657 and protein G beads to have an excess of both to connect all p217+tautothebeadsbutaslittleaspossibletominimizematrix interference wereaddedtotheclinicalstudyCSFsamplesand in-cubatedovernight at4°C.After incubation,theunboundfraction was removed, beads were washed andtrypsinization buffer was added Method validation could not be executed in a traditional waybecause of thelimited availability of humanCSF andas (C-terminallytruncated)ptauisnaturallypresentinCSF.Nonetheless, multipletest batches were analysed to optimizesample prepara-tionandLC-TQMSmethodsandtoconfirmmethodreliability Dur-ing analysisofclinical samples,all quality controlparameters for batchacceptance,agreedpriortosamplepreparationandanalysis, weremet(e.g.acceptancecriteriaforcalibrationpointsand qual-itycontrolsamples,etc.,asdescribedinsupplementalinformation 1) Carryover (appr.5%) wasobservedduring2DLC-TQMS analy-sis: a duplicated gradient test as describedby Vu et al.[44] re-vealed that carry over is not caused by theinjection systembut originates from retention of diphospho-peptides in the LC-TQMS

S Bijttebier, C Theunis, F Jahouh et al Journal of Chromatography A 1651 (2021) 462299

Fig 4 1DLC-TQMS analysis of TPSLP(pT)PPTR in tryptic digest of IP-ed human CSF before (left) and after (right) TiO 2 /ZrO 2 clean-up Results were obtained with the 1DLC- MSMS setup described in Materials and Methods with MP gradient as follows (min/A%): 0.0/100, 5.0/70, 5.1/2, 6.6/2, 6.7/100, 10/100

Fig 5 IP-2DLC-TQMS analysis of SRTPSLPTPPTREPK ∗ 2p in human CSF clinical samples Graph A:% Reduction of total p217 + tau (LC-TQMS) after normalization for total tau levels (Simoa) compared to baseline samples Graph B: correlation of total p217 + tau measured with IP LC-TQMS and p217 + tau simoa measurement (PT3-PT82: + boiling) Only values within calibration range are included Two samples were below quantification limit and one sample was above calibration range

systemandcolumn.Asp217+tautrypticpeptidesareverylow in

abundance (pMquantities)inhumanCSF,endogenous

concentra-tionsareclosetothemethodLLOQ.Limitingthedynamicrangeof

calibrationstandardsreducedcarryover

Fig 5 shows the quantitative IP-2DLC-TQMS measurement of

the SRTPSLPTPPTREPK2p tryptic peptide inthe human CSF

clin-icalsamples,depictedasthenormalizedp217+tauleveltothe

to-tal tau level (left), and the correlation between the IP-LC-TQMS

data and the PT3 x PT82 Simoa data (right) An average

maxi-mum reductionof50%oftotalp217+tau(LC-TQMS)after

normal-ization fortotaltau (HT-7x PT82simoa assay [21]) comparedto

baseline samples wasobserved,thereby confirmingthereduction

of total p217+tau afterdosing as observed withthe PT3 xPT82

Simoa Absolute levels measured with Simoa are slightly lower

than with IP-2DLC-TQMS, which isreflected in the slope (<1) of

the linear regressionin the correlation graphof the two

orthog-onal assays in Fig 5 (right) The small differences observed

be-tweenthetwoanalyticalmethodscanbeexplainedbydifferences

insamplepreparation,standardmaterialforcalibrationcurvesand

methods for analysis[45].Forexample, forSimoa analysisof

to-talp217+tau,samplesareboiledtodisruptantibody-antigen

com-plexes which is not needed for IP-2DLC-TQMS Additionally the

captureantibodyPT3(themouseIgG2aparentversionofJNJ’3657

– [13]) used in theSimoa assay binds all the different

phospho-rylated p217+tau speciesaroundpT217 epitope(e.g.pT212/pT217

and pT217) that are consequently contributing to the detected

signal, while during the IP-2DLC-TQMS assay only the surrogate

peptide SRTPSLPTPPTREPK2p is quantified Notwithstanding the

observed difference, there is a strong correlation between total

p217+tau measured withIP LC-TQMS versus Simoa, asshown in

Fig.5

Two extracts of clinical study samples containing a relatively high concentration of SRTPSLPTPPTREPK2p were used to try to identifythelocation ofthetwophosphorylations.The peptide se-quence contains 4 potential phosphorylation sites, namely, S210, T212,S214,T217andT220.TauphosphorylationatT220 has how-ever never been reported before [46] Moreover, it was shown before that Ab PT3, of which a humanized variant was used in thisstudy,showsvirtuallynoaffinityforphosphorylationatT220 [12,13] T220wasthereforeexcluded aspotentialphosphorylation site.As nochromatographicseparationofthediphospho-peptides wasobtained, MS fragmentation was the only wayto differenti-atebetweenphosphorylationsites.ItwasnoticedduringMS tun-ing of a SR(pT)PSLP(pT)PPTREPK standard solution that the 4+ charge-state(atm/z 456.71) rendersmuchmorediagnostic prod-uct ionsthanthe3+ charge-state(atm/z 608.62).Theabove de-scribed2DLC-MSmethodwasmodifiedtobeabletoperform anal-ysis on a newly installed 6500+ TQMS from Sciex (method de-tails describedin supplemental information 4) This methodwas used to analyse a standard solution of SR(pT)PSLP(pT)PPTREPK,

a tryptic digest of the elongated peptide containing phosphory-lations at S214 and T217 and two CFS clinical study sample di-gests, while monitoring diagnostic product ions of the precursor ion with a 4+ charge-state In both the standards and samples,

a peak wasdetected at m/z 551.77 corresponding to a y9 prod-uction(2+charge-state)containingonephospho-moiety,thereby confirmingthe presence ofphosphorylation at T217 (supplemen-tal information4).Next tothat, apeak wasdetected inboth the SR(pT)PSLP(pT)PPTREPK standard andthesamplesat m/z591 23 (1+ charge-state) corresponding to a b5 – H2Oproduct ion con-tainingone phospho-moiety, indicating phosphorylationat either S210, T212 or S214 (supplemental information 4) However, this