Comparison
of
serum
fractionation
methods
by
data
independent
label-free
proteomics
D.
Baiwir
a,
G.
Mazzucchelli
b,1,
N.
Smargiasso
b,1,
F.
Quesada-Calvo
c,
E.
De
Pauw
b,
M.
Malaise
d,
E.
Louis
c,
M.-A.
Meuwis
c,*
aProteomicFacility,GIGA,UniversityofLiège,Belgium b
LaboratoryofMassSpectrometry,GIGA-R,ChemistryDepartment,CART,ULg,Belgium
c
GastroenterologyDepartment,GIGA-R,LiègeUniversityHospital,CHU-ULg,Belgium
d
RheumathologyDepartment,GIGA-RLiège,UniversityHospital,CHU-ULg,Belgium
ARTICLE INFO Articlehistory:
Received26January2015
Receivedinrevisedform1July2015 Accepted3July2015
Availableonline9July2015 Keywords: Serumprefractionation Repeatability Label-freeproteomics Dataindependent Depletion Equalization ABSTRACT
Off-linesampleprefractionationsappliedpriortobiomarkerdiscoveryproteomicsareoptionstoenable
moreproteinidentificationsanddetectlow-abundanceproteins.Thisworkcomparedfivecommercial
methodsefficiencytorawserumanalysisusinglabel-freeproteomics.Thevariabilityoftheprotein
quantitiesdeterminedforeachprocesswassimilartotheunprefractionatedserum.A49%increasein
proteinidentificationsand12.2%ofreliablequantificationwereobtained.A61timeslowerlimitof
proteinquantitationwasreachedcomparedto proteinconcentrationsobservedinrawserum. The
concentrationsofdetectedproteinswereconfrontedtoestimatedreferencevalues.
ã2015TheAuthors.PublishedbyElsevierGmbH.ThisisanopenaccessarticleundertheCCBY-NC-ND
license(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1.Introduction
The research of proteomic clinical biomarker was often undertakenonserumorplasmatohighlightbiomarkersavailable at thesystemic level. Indeed clinicaltrials often collect serum sampleswhich aresometimesavailable forancilarybiomarkers discoverystudies.Thelow-concentrationproteinrangeofserum containsproteinsassecretionortissueleakageproducts.Theseare expected to be more specific and sensitive potential disease biomarkersthanmoreabundantproteins[1–3].Themainissue inherenttothecomplexityofserumanalysisisthelimitationsof thetechnologiesusedforthediscoveryanalysis:limitsofdetection (LOD)andquantification(LOQ)andthelinearrangeofresponse. Indeed,serumandplasmaareverycomplexbiologicalmatricesin which the proteinconcentration covers a large dynamic range higherthan10ordersofmagnitude[4].Currentlynohighendmass spectrometercan coversucha largerange [5]. Prefractionation strategiesofferanalternativeby modifyingtheoriginalsample protein distribution inducing shifting/shrinking effects and allowingproteins initially present at low concentrations tobe
accessiblefor analysis.Several commercialsolutionsforsample prefractionationareproposedtoremoveabundantproteinswhich saturatesignalduringproteomicanalysisandalsoforreachingthe low-abundance proteins [2,6–8]. Of course, off-line or on-line fractionationstepsenhancethenumberofproteinsidentifiedby increasingtheseparationpower.Butthisisatthecostoflonger acquisitiontimes,amodifiedlimitofquantificationandahigher global variability; alldue totheadditional preprocessing steps applied[9].Suchmultistepstrategyiscertainlybetteradaptedfor celllinesoranimalmodelscompleteproteome characterization without downstream differential analysis [10]. However for differentialanalysisandthereforeforrelatedclinicalproteomics, a highernumber ofconsistentproteinidentificationsand more accurate quantification increase the probability to highlight a significant potential biomarker while keeping reasonable data aquisitiontimeandrunlength.Therefore,samplepreprocessing steps appear mandatory, but must involve at least a good repeatability, reasonable cost and manageable processing time for allowing a medium to high number of clinical sample preparations.
Amongthestrategies availablearetheimmuno-affinitydepletion of the abundant proteins as simple IgG and/or albumin or the depletionofuptothetwentymostabundantplasmaproteinsusing IgY chickenantibodies. TheseIgYdecrease therisk foraspecific protein co-depletions potentially occurring with mammalian
*Correspondingauthor.
E-mailaddress:marie-alice.meuwis@ulg.ac.be(M.-A. Meuwis).
1
Boththeauthorscontributedequallytothiswork.
http://dx.doi.org/10.1016/j.euprot.2015.07.009
1876-3820/ã2015TheAuthors.PublishedbyElsevierGmbH.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/ 4.0/).
ContentslistsavailableatScienceDirect
EuPA
Open
Proteomics
antibodiesbaseddepletion[11–15].Anotheroptionisbasedonthe principleoftheequalizationtechnique,utilizingarandomsynthetic hexapeptides librarycross-linked to micro-beads[16–21] which actuallyappearstoworkaccordingtoageneralhydrophobicbinding mechanism[4].Thelargevolumeofsamplewhichisrequiredand thefact that the“equalization mechanism” shrinks the sample proteinconcentrationdynamicrange,canbeseenasaproblemfor datainterpretationafterdifferentialproteomicdiscoveryanalysis andfurther validationoftheresults.Theenrichmentofglycoproteins on lectinesmay also be used as an alternativeapproach. Many studies wereperformedon varioussystemsdedicated todifferential analysis [12,22–24]. However, beside works done for protocols testing [14,20,23,25,26],onlyonestudycomparingplasmapreprocessing reproducibilityusingtwocommercialkits underspincolumn format and applying a downstream label-free proteomic analysis was published[27].Therefore,inthiscontext,theaimofourworkwasto compare five different commercial methods proposed under disposablespin column format, involving reasonable processing timeandcostsandtestedusingthesameserumpooloriginatedfrom healthyanddiseasedindividuals.Thequalitativeandquantitative resultsobtainedforthefiveconditionswerecomparedtogetheras wellaswiththoseobtainedwiththerawserumanalysis.Theresults arediscussedinthecontextoftheapplicationoftheprefractionation methodsonserumbeforeaclinicalbiomarkerdiscoverystudyby label-freeproteomics.Itcanbeviewedasatooltoselectthemost convenient method to apply on clinical sera before proteomic analysis.
2.Materialsandmethods
2.1.Compositionofthesampleserumpool
Alltheserumsamplesofpatientswerecollectedwithsigned informedconsentandwiththeethicapprovalofouruniversity hospital. Three healthy subjects, five patients with colorectal
cancer, five Crohn’s disease and three ulcerative colitis patient sampleswereselectedtoprepareapoolof5mL.Theserumpool aliquots ofeach patientwerestoredforseveral yearsat 80C (maximum5years)andthawedonicepriortomixing.Thisraw serum pool (RS*pool)was tested fortotal proteinquantitation usingtheRCDCkit(BioRad,Inc., Hercule,CA,USA).Aliquots of suitable volumeswere prepared for each protein depletion kit applicationsandstoredat 80Cuntilfurtheruse.
2.2.PrefractionationoftheRS*pool
Themethodsofdepletiontestedwere:ProteoPrep120Plasma kit (Sigma St. Louis, USA), under the spin column format, the ProteoMinerTM(BioRad,Inc.,Hercule,CA,USA)designedfor200
m
Lofsamplevolume,IgG&AlbuminSpinTrapcolumn(GE Health-care, USA) and WGA Glycoproteinisolation kit (Thermo-Fisher ScientificInc.,USA)withorwithoutpriorIgG&AlbuminDepletion SpinTrap(GEHealthcare).Fig.1summarizestheworkflowwith themainsampleprocessingstepsappliedtotheRS*poolbefore performingtheproteomicanalysis.
2.2.1.ProteoMinerTMkit(BioRadInc.,Hercule,CA,USA)
Thesmallcapacitykit(designedfor200
m
Lofsample)wasused for thethreetechnicalreplicates whichwereruninparallel.In brief,a700m
LvolumealiquotoftheRS*poolwasthawedonice. After centrifugation, 3200m
L of RS* pool were treated in parallel, asrecommendedby themanufacturer, andresulted in threesuccessiveelutionstepsper200m
LofRS*treated.Allthree elutedfractionswerepooledtogetherperreplicateandstoredat80C.
2.2.2.ProteoPrep120Plasmakit(Sigma,St.Louis,USA)
Thiskitwasappliedasrecommendedbythemanufacturer.We performedthreereplicatesofprocessconsecutivelyonthesame column(andonthesameday).Briefly,foreachprocessreplicate
onealiquotofRS*poolwasthawedoniceand24
m
Lwerediluted in300m
Lofbindingbufferprovidedinthekit.Thedepletionwas appliedtwiceon100m
Loffilteredanddilutedspikedserumpool. Afterwards, the two depleted fractions collected were pooled togetherandconcentratedto100m
LusingtheVivaSpin(cut-off 5kDa)includedinthekit.Theconcentratedpoolwasappliedon column for the second depletion cycle and the final depleted fractionobtained(whichwassubjectedontotaltotwocyclesof affinitydepletion)wasfurtherstoredat 80C.Reconditioningoftheaffinitycolumnkitwasperformedaccordingtothe manufac-turerrecommendation.
2.2.3.Albumin&IgGSpinTrapcolumn(GEHealthcare,UKLtd.,UK) AnaliquotofRS*poolwasthawedoniceand355
m
LofRS*were diluted with the provided kit binding buffer, to allow for six replicates(50m
Leach)applications.These6depletionswererun inparallelusingtheprotocolrecommendedbythemanufacturer. However,applicationoftheIgGandalbuminaffinitydepletionwas appliedtwiceconsecutivelyforeverysamplereplicateprocessed. Thethreedepletedfractionswerestoredat 80Cuntilfurtheruse andthethreeotheroneswerestoredat4Candprocessedusing WGAglycoproteinenrichmentonthesameday.2.2.4.WGAGlycoproteinisolationkit(Thermo-FisherScientific,Inc., USA)
2.2.4.1.OnRS*pool. Thethreetechnicalreplicatesperformedon thespikedserumpoolwereruninparallel.AnRS*poolaliquotwas thawedoniceandcentrifuged.Only20
m
Lperreplicatewereused forthekitapplicationasrecommendedbythemanufacturer,using200
m
gofWGA-lectinbeadsperreplicate.Thetwostepsofelution wereperformedusing200m
Lelutionbufferprovidedinthekitand pooledtogetherbeforestorageat 80C.2.2.4.2. Albumin andIgG depleted RS* pool. The three technical replicatesofWGA-lectinglycoproteinenrichmentperformedafter IgG &albumindepletionof thespikedserum poolwererunin parallel.Theflow-through(V=500
m
L)ofAlbumin&IgGdepletion SpinTrapwassupplementedwith150m
Lofbindingbuffer.The next steps of the process were done as recommended by the manufacturerwiththetwofinalelutionswhichwereperformed using200m
Lelutionbufferprovidedinthekit.Theelutedfractions wereallpooledtogetherbeforestorageat 80C.2.3.Generalexperimentstrategyconsiderations
Inordertoensurestandardizationofthedigestionandpeptides conditioningattheendofeachpreparationprocess,anecessarily constantquantityofmaterial(=20
m
g),wastreated.(seeFig.1).All thestepsofproteinclean-up,reduction–alkylationandthefinal trypsindigestion,wereperformedinparallelwithastandardized protocol.Nootherqualitycontroloftrypsindigestionwasapplied and run in parallel of the samples to simplify the QC of the experiment. Indeed we consideredthat the RS* pool replicates themselvesweretheQCsofdigestion,peptideextractionandMSEdataacquisitionandanalysis.However,anexternalQCcomposed of entire proteins at knownrelative quantity ratio and known concentrationscanberuninparallelofthesamplesasexternal digestion QC, but not involving the aspect of matrix of high complexityasserum.
Table1
Qualitativeandquantitativeresultssummary.
Characteristics ProteoMinerTM
ProteoPrep120 IgG&albumin IgG& albumin+WGA WGA RS* Sample data before MS analysis
Totalstartingproteinquantity(RCDConserum pool)(mg)
13220 1057 3305 3305 1322 20 VolumeofRS*pooltreated(mL) 200 16 50 50 20 0.3031a
Recoveryoftotalproteinsafterprefractionation (meanof3replicatesSD,expressedinpercentof totalstartingquantitytreated)(%)
2.30.34 1.910.17 24.474.93 7.580.52 32.913.21 NR
Theoreticalexpectedproteinquantityrecovery (percentageofstartingproteinquantity)(%)
0.4–0.7b
1 16 <40 NC 100
Manipulationtimeforkitapplication(h) 4 3 3 5.5 2.5 0.1 PLGS 2.5ion count tables derived data
Quantityofproteinsexperimentalyidentifiedand quantifiedby2D-nanoUPLC–MSErun(meanof
3processreplicatesSD)(ng)
805.29113.73 1450.43241.89 1652.35315.88 653.1224.78 686.775.15 721.1141.1
Numberofproteinsidentifiedinthe3replicatesof process(3R3)(numberofproteinsidentifiedand quantifiedin3R3)
130(124) 191(178) 134(122) 127(117) 130(128) 128(126)
Numberofproteinsidentifiedandquantifiedin 3R3withindividualQtySD50%;QtySD30%
97;70 127;83 100;88 98;89 31;21 93;74 PercentageincreaseobtainedcomparedtoRS*pool
forallproteins;forproteinswithQtySD50%; withQtySD30%(numberofproteinsgained comparedtoRS*pool)(%)
1.5;4.3; 5.4 49.2;36.6;12.2 4.7;7.5;18.9 1;5.4;20.3 1.5; 66.6; 71.6
NR
Numberofproteinsidentifiedandquantifiedin 3R3whateverQtySD,reportedtothemeanprotein quantitydetected(numberofprot/ng)
0.154 0.123 0.074 0.179 0.186 0.175
Numberofuniqueproteinsquantifiedinall 3replicatesofprocess(withfilteronQtySD50%; 30%)
6(8;7) 38(31;16) 1(2;1) 1(7;4) 3(0;0) 7(4;4)
Theconcentrationdynamicrangeofproteins identifiedandquantifiedin3R3c
(withlogarithmic transformation)
3.10 3.86 3.52 3.60 3.30 3.78
NC,notcommunicatedbymanufacturer,NR,notrelevant.
aTheexactvolumtreatedwas10mLofdilutedRS*poolinPBSsuchas20mgofproteinweretreatedandcorrespondtoonly0.303mLofundilutedRS*pool. b
Datadeducedfromprevioustestsperformedinthelaboratory(datanotshown).
c
Concentrationdynamicrangeistheratiobetweenthehighestandthelowestquantitiesoftheproteinsidentifiedandquantifiedinallthreereplicates(ameanis calculatedusingthethreereplicatesofprocessperformedforboththehighestandthelowestabundantproteinsobtained);thisratiovaluewassubjectedtobase 10logarithmictransformation.
Weperformedsomeofthedepletionstwiceconsecutivelyon the same starting material (for IgG & albumin depletion and ProteoPrep120kit)togetoptimalresults.Despitetheapplication ofsomedepletioncycles twice,thetime necessaryforallsteps completionstayedinferiorto6hpersamplereplicate(seeTable1). Moreover, these methods at the exception of the ProteoPrep1 20couldberun inparallelreducing againtherequiredsample preprocessing time and in theory theirrespective repeatability variation.
2.4.Proteindigestionandpreparationforlabel-freeproteomics AfterquantitationusingtheRCDCkit(BioRad,Inc.,Hercule,CA, USA), an aliquot of each fractionated sample replicateand the diluted RS* pool (20
m
gof total proteins) were precipitated in parallelusing2D-Cleanupkit(GEHealthcare,UK).Thesamples were all reduced and alkylated as previously described and digested using porcin recombinant trypsin (Roche, Ltd., Switzerland)[28].Eachproteindigestwasplacedinaqueous0.1%trifluoroacetic acid final solution and further purified using C18 ZipTip high capacity(MerckMillipore,Billerica,MA,USA).Proteindigestswere speed-vacuumdriedandreconstitutedusing100mMammonium formatesolutionadjustedtopH10ataconcentrationof278ng/
m
L. Eachinjectedsamplewasspikedwitha commercialmixtureof proteindigest standardsoriginated fromnon-human biological material: the MassPREPTM digestion standards (Waters, Corp.,Milford,USA)at150fmolofADHpersampleprocessreplicate.This commercialstandard consistsof two standard mixtures (MPDS Mix1andMPDSMix2)containingproteindigestsofyeastalcohol dehydrogenase(ADH), rabbit glycogenphosphorylase b, bovine serum albumin, and yeast enolase present at known protein quantities;allowingthereforetocheckforrelativequantitationof the samples spiked. Sample injection replicates order on 2D-nanoUPLC-ESI–MSE(Waters,Corp.,Milford,USA)wasrandomized
and 2.5
m
g of total protein digest (in 9m
L) were injected per samplereplicate.2.5.2D-nanoUPLC-ESI–MSEsystemconfigurationandsettings
All eighteen samples (threefive prefractionated digested samplesandthreeRS*pooldigested),wereinjectedontothe 2D-nanoAquity UPLC (Waters, Corp., Milford, USA) coupled online withtheQ-TOFSynaptHDMSTMG2system(Waters,Corp.,Milford, USA) using ion mobility as supplementary separation. The configuration ofthe2D-nanoUPLCsystemwas areversedphase pH10–reversedphasepH3basedtwodimensionseparation.The firstdimensionseparationwasmadeonanX-BridgeBEHC185
m
m column(300m
m50mm).ThetrapcolumnSymmetryC185m
m (180m
m20mm)andanalyticalcolumnBEHC181.7m
m(75m
m 250mm)(Waters,Corp.,Milford,USA)wereusedafteranonline dilutiontolowerpHvalues(minimumbeenpH3).Thesamples wereloaded at 2m
L/min (20mM ammonium formate solution adjustedtopH10)onthefirstcolumnandsubsequentlyelutedin fivesteps(10,14,16,20and65%acetonitrile).Eachelutedfraction wasdesaltedonthetrapcolumnafteratentimesonlinedilutionto pH3andsubsequentlyseparatedontheanalyticalcolumn;flow rate250nL/min,solventA(0.1%formicacidinwater)andsolventB (0.1%formic acidin acetonitrile), linear gradient0min,97% A; 90min,65%A.(Thetotalruntimeofanalysiswas5120=600 min).Themassspectrometerwasoperatedinpositiveionmodeand thedataacquisitionwereperformedin the50–1500m/zrange, withascantimeof0.6sandwithcollisionenergyvoltagessetin independent alternative scanning (MSE) mode. The IMS cell
pressure was set at 2.5mbar, the variable IMS wave velocity
rangedfrom850m/sto1200m/sandthewaveheightwas40V.A lock mass correction was done using [Glu1]-fibrinopeptide B ([M+2H]2+:785.84206m/z)(WatersCorp.,Milford,USA).
2.6.ProcessingofrawdatausingProteinLynxGlobalServer(PLGS2.5, WatersCorp.,Milford,USA)andanalysisofionaccounttables
Rawdataprocessing(deconvolutionanddeisotoping),protein identificationandrelativequantificationwereallperformedusing ProteinLynx GlobalSERVER (PLGS) v2.5 (Waters Corp., Milford, USA).Theprocessingparametersweresetasfollows:theMSTOF resolutionandthechromatographicpeakwidthwerebothsetto automatic,thelow-/elevated-energydetectionthresholdwasset to150/15counts,respectively,theidentificationintensity thresh-oldwas500countsandthelockmasswindowat785.84206m/z wassetto0.30Da.
For the identification of the proteins, the UniProt human database(UniProtrelease 2011_12-December14th,2011) imple-mentedwiththesequencesofthenon-humanproteinstandards spikedasproteinsdigestsbeforeinjectioninrawserum(accession #: P00489,P00924, P02769, P00330), was used. The searched parametersusedwereasfollows:carbamidomethylation(C)and oxidation (M) of peptides as fixed and variable modifications respectively,twopossibletrypsinmissedcleavageswereallowed, minimum fragment ion matches per peptide of three and minimumfragmentionmatchesperproteinofseven,minimum two peptides matches per protein and finaly a protein false discoveryrate(FDR)forproteinwassetatmaximum4%(whichis equivalenttoa1%FDRatthepeptidelevel).
Proteinandpeptidesionaccounttablesweregeneratedduring proteinidentification.Theproteinquantitationwasthenbasedon calibration using the response factor of the ADH spiked. The normalization on ADH was also performed using the PLGS vs 2.5 Expression module to control technical variability. The technical variability was evaluated using comparison between expected and obtained protein ratios for the different sample standardMixMassPREPTMdigestionstandardsMix1or2(Waters
Corp.,Milford,USA).Thesewerecalculatedfromtheexpression analysis tables generated using P00330-ADH normalization to verify that the process was technically sound and that the variabilityofthis experimentsatisfiedtheoneof thislabelfree technology(maximum30%)[29].Ionaccounttableswereusedto extractdataconcerningtheproteinsidentifiedandquantifiedin eachreplicatedoneforeachsamples.Themethodofquantitation usedforionaccounttablegenerationbyPLGSvs2.5,isthe“Top3” [30].Toenablecomparisonofdifferentprefractionationmethods appliedtothespikedrawserum,weexpressedthequantityofeach proteinasthepercentageof totalproteinquantityobtainedper replicate run. We calculated each meanprotein quantity value basedonthethreereplicatevaluesforthetotalproteinquantity andalsoforeachindividualproteinquantity.Wealsocalculated theproteinconcentrationsequivalenttotheinitialquantity(and volume)ofRS*pooltreated.Thisoneisreferedasthe“calculated rawserumequivalentconcentrationofprotein”andisexpressedin
m
g/mL.Itwasobtainedbydividingby100each proteinrelative quantity(obtainedusingthe3replicatesofprocessandexpressed inpercents),andbymultiplyingthequotientobtainedbyafactor. Thisfactorwasdefinedastheratiobetweenthequantityoftotal protein recovered after the prefractionation and the quantity analyzed oncolumn. For example, for ProteoPrep1 20 sample: 20.1887m
g/2.5m
g=8.0755.Thislast productwas further multi-plied by the mean total protein quantity (obtained with the 3 process replicates and expressed inm
g) and divided by the numberofmLofrawserumtreatedinitialy(forProteoPrep120: 0.016mL).Theratioobtainedwasthereforedifferentforeachkit. For example, for the ProteoPrep1 20, the ratio is 8.075480 as20.1887
m
g was recovered and that only 2.5m
g was loaded on column.ForRS*pool,thisratiois8(20m
g/2.5m
g).Thesedatawere usedtocomparethe5prefractionationmethodstogetherandwith theRS*poolresults.TheRSDwasobtainedforthetotalproteinquantityrecovery calculatedusingeachprocess replicates(=relativeQtySD/mean relativeQty100,expressedin%).
AlltheresultsdataweretreatedusingtheExceltabler. TheExpressionAnalysismodulewasusedtoperformdifferential analysisusingthethreetechnicalreplicatesdoneperdepletion/ enrichment kit, performing exclusively pairwise comparisons. Expressionanalysiswasusedtoverifythespikeproteindigestof thestandard MixMassPREPTMdigestionstandardsMix1and2ratio.
NormalizationwasperformedontheADH(P00330).Thisanalysis wasdoneonlyconsideringtheproteinsidentifiedandquantifiedin threeoutofthreereplicatesofprocess(filterapplication:3R3). 3.Resultsanddiscussion
3.1.Totalproteinquantityobtainedfortheprefractionatedsamples The measured protein concentration of the RS* pool was 66.1061.890
m
g/m
L.Thetotalproteinquantitiesandthe corre-spondingsamplevolumestreatedwitheachmethod,theexpected theoreticaldata andtheempiricalresultsobtainedare summa-rizedin Table 1.The repeatability of each kit applications was evaluatedusingthetotalproteinquantitymeasuredjustpriorto sampleclean-upanditprovidedaSD<5%.3.2.Proteomicresultsanalysis 3.2.1.Datatechnicalqualitycontrols
TheMPDSmixproteinratioswerecalculatedusingthequantity of ADH (P00330) spiked amount,while performing expression
analysistocompareprocessreplicatesofeachmethodofserum preprocessing. All the Mix MassPREPTM digestion standards proteins showed process replicates SD equal or lower to the expected30%[31].
3.2.2.Qualitativeandquantitativeproteomicresults
Table1summarizesthemainresultsobtainedforeachmethod ofprefractionationandtheRS*pool.
3.2.2.1.Totalproteinrecovery. Thetotalproteinrecoveryobtained afterprefractionationandthequantityforeachproteinidentified andquantifiedaredetailedinTable1.Thequantityofeachprotein isthemeanofthethreeprocessreplicatesstandard deviation (SD),obtainedforeachprefractionationmethod.
3.2.2.2. Individual protein variability. The variability in protein quantity identified and quantified per process replicates is expressed in ng (see Table 1).The relative standard variations (RSD)ofeachsamplewerelowerthan20%.TheRS*poolshowedin thisexperimentthehighestRSDat19.57%countingonlyfor clean-up,digestionandpeptideresolubilisation/extractionyieldsandthe applied downstream nanoUPLC–MSE step repeatability.The
smallest RSD was obtained withthe WGA lectine glycoprotein enrichmentappliedafterIgG&albumindepletion(RSD=3.79%). TheRSDofProteoMinerTMandProteoPrep120were14.12%and
16.65%,respectively.Therelatedproduct“prot20column”(Sigma) prefractionationonanoff-lineFPLC-systemcombinedwith label-free LC–MS/MS analysisprovidedinter-runsCVassays of30.9% [12].Moreover,thedistributionofeachproteinRSDobtainedfor theRS*poolwassimilartotheoneofotherpretreatedsamples (datanot shown). Hakimiet al. using plasmasample and spin columnformatofProteoMinerTMandProteoPrep120,reporteda
betterreproducibilityofprocessusingProteoMinerTMthanusing
theProteoPrep1 20depletionstrategy [27].On theserumpool
Fig.2.(A)VenndiagramswiththepoteinsidentifiedperprocessreplicateofeachprefractionationmethodandtheRS*pool,allanalyzedbylabel-freeMSE
proteomics.(B) ComparisonillustratedbyVenndiagramsoftheproteinsidentifiedin3R3fortheProteoPrep120,fortheProteoMinerTM
,IgG&albumindepletedsamplesandtheRS*pool. FromtoptobottomtheVennanalysisresultsareshownapplying,respectively,no,50%and30%thresholdquantitySDfilteronproteinsincluded.
testedinourwork,ProteoMinerTMappearedslightlybetterthan
ProteoPrep120.
3.2.2.3.Numberofdifferentproteinsidentifiedandquantified 3.2.2.3.1.DataforeachmethodandtheRS*pool. Table1detailsthe numberofproteinsidentifiedinthreereplicatesoutofthree(3R3). The range and the percentage of conserved proteins after the 3R3filterapplicationshowsthequalitativerepeatabilityofeach method(=prefractionation process and nanoUPLC–MSE). Thisis
illustratedFig.2AwithVenndiagramsbuiltwiththenumberof proteinsidentifiedin eachreplicate ofprocess (spikedproteins excluded). The numbers of proteins quantified and showing a quantity SD higher than 30% (technical variation of the instrumental system) and lower than 50% (arbitrary cut-off value that might be suitable for a discovery phase) are annotated in the Table 1. As a discoverystudy should include somebiologicalreplicatestobeconsistentandasthevariability associated to biological replicates is certainly higher than the technicalone,weincludedananalysisofthedataafterapplying this50%cut-offforquantitySD(%).Thisindicatedthenumberof proteinsthatmightbeoftruevalueforconsistentresultsanalysis aftersamplepreprocessing.The30%thresholdfiltercorresponding tothetechnicalvariabilityoftheinstrumentalsystemusedinthis experimentcouldbeconsideredastoostringent.
3.2.2.3.2. Comparison of the results obtained after the prefarctionationstotheRS* Pool. Fig.2BshowsVenndiagrams comparingthreeselectedmethodstogetherandwiththeRS*pool. Thecomparisonofthesixproteinlistsidentifiedin3R3,without anyquantitySDfilter,canbedownloadedasSupplementarydata Table1.Intheanalysisincludingalltheproteinswhateverquantity SD obtained, only 71 proteins were found common to the 6 conditions compared. The maximum number of “unique proteins”,identifiedsolelyinoneprefractionationmethod,with 3R3filterapplication and witha quantity SD lowertothetwo thresholdsarealsoreportedinTable1.Thenumbersof“unique proteins”indicatetheadvantageofonemethodovertheothersand mostlyovernoprefractionation(RS*poolsample).ProteoPrep1 20providedthehighestnumberof“uniqueproteins”whatever
quantitySDfilterisapplied.Inaddition,ProteoPrep120process enableda49%increaseinnumberofproteinsidentifiedcompared totheRS*pool(in3R3;noQtySDfilter).Butwhenanalyzingthe numberofproteinswithasuitablequantitySD,thegainof36.6% obtained for 50% Qty SD dropped to 12.2%, when using the technical30%threshold.Atthislevelofstringency,theAlbumin& IgGDepletionSpinTrapkitorthecombinationofIgG&albumin depletion prior to WGA enrichment appeared to be slightly superiorwithagainof18.9%and20.3%,respectivelyovertheRS* pool. The apparent increases observed for some of the prefractionation kits used must be nuanced depending on the stringency of analysis. Altogether, ProteoPrep1 20 enrichment provided thehighest advantage,regarding thequantity/volume ratioofrawbiologicalmaterialneededforthekitapplicationand thenumberofproteinssignificantlyidentifiedandquantified.All theotherprefractionationmethodstestedregardingthesecriteria didnotprovidedbetterresultsthantheRS*pool.
3.2.3.Themostabundantproteinsandtheirdepletionefficiencies 3.2.3.1. The 20 most abundant protein families analysis. Supplementary data Table 2 details all the proteins and proteinfamilies targetedbytheProteoPrep120plasma kit (according to information provided by the manufacturer data sheet).Thedataprovidedtofacilitatecomparisonofkitefficiencies are the relative quantities (% of the total quantity of protein detectedandquantifiedforeachcondition).Whensummingthe relativeproteinquantitypercentagesoftheIgGfamilymembers identified(Prot.entryandProt.Acc#with**inSupplementarydata Table 2),the ProteoPrep1 20 methodshowed the lowest total relative quantity,whichindicatestheirmoreefficientdepletion withthiskit.Thecalculatedproteinconcentrations,equivalentto raw serum (
m
g/mL) and the corresponding SD are also communicatedforeach sample.But,notallthemostabundant proteinspointedshowedanindividualquantitySD<30%. 3.2.3.2.Theefficiencyofalbumin(HSA)depletion. Themethodthat provided the best results concerning HSA depletion was the combinationofIgG&albumindepletionwithdownstreamWGAFig.3. Thesixrangesofcalculatedrawserumequivalentproteinconcentrationobtainedfortheproteinsidentifiedandquantifiedin3R3(mg/mL)forthemethodsof prefractionationandfortheRS*poolsample.
ProteinwithquantitySD30%(filledlabel:*);QuantitySD>30%(emptylabel: )
Thelowestabundantproteinsreliablyquantifiedineachsample(QtySD30%)areK1C25(0.0180.031mg/mL)forProteoMinerTM
;ACTH(5.8410 3
10.13103m
g/mL) forProteoPrep120;K1C28(0.2970.514mg/mL)forIgG&albumin;K1C28(0.0520.089mg/mL)forIgG&albumin+WGA;KRT15(0.2270.394mg/mL)forWGA;HBD (0.2880.500mg/mL)forRS*pool.
enrichment which showed the lowest relative quantity percentage. The second best was WGA enrichment and the worst was ProteoMinerTM with results closer to these of RS*
pool.KnowingthatHSAispresentinserumatapproximately75% oftotalproteinquantity[32],thelowpercentageinRS*pool(9.15% of HSA quantified) illustrated the compression of the protein concentration dynamic range obtained. This was however expected using this proteomic technology and these settings, potentiallyinvolving alsosuppression effects wellknown with electrosprayionization(ESI)andthefactthatthequantitationof analytes withthis technology is concentrationdependent [33]. However,onemustkeep inmindthatveryhighlyconcentrated proteins(asHSA)andthereforethemostabundantones,couldbe saturating and not be correctly quantified. Importantly, HSA showed,onlyforRS*pool,ProteominerTMandWGA enrichment, quantitySDlowerto30%.
3.2.4.Theproteinconcentrationdynamicrangesandtheabilitiesto measurelow-abundanceproteinsinserum
3.2.4.1.Proteinconcentrationdynamicrange. Table1providesthe dynamic ranges of protein concentrations observed (with base 10logarithmic transformation) for thesix conditions analyzed. Fig. 3 shows for every method tested, the distribution of the proteinsidentifiedandquantifiedandrankedindecreasingorder ofcalculatedequivalenttorawserumconcentrations(usingabase 10logarithmicscale).Thelargestdistributionintermofprotein concentrationswasobtainedusingtheProteoPrep120andsecond best was the RS* pool itself. The smallest range observed was obtainedafterequalizationwithProteoMinerTMwhichmightbe
obviousconsideringthetheoreticalprincipleofthiskit.Thisisin opposition with what was reported previously using similar workflows: DIA and label-free strategy [27]. In this work, ProteoMinerTM application on plasma enabled detection of
proteins within a 5 orders logarithmic range of protein concentrations.This demonstrates againthe shrinkingeffect of thisstrategywhenappliedonserummatrix,asreportedpreviously [21,34].Butalsohighlighthedifferenceofresultsobtainedwhen usingplasmaorserum.Fig.3showsthatthemaximumgainin proteinidentificationsnumber(49%)obtainedwithProteoPrep1 20, were proteins lying within medium values of the protein relativequantification range.TheProteoMinerTMcurveobtained
wasnotas flatascouldbeexpectedafterequalization. Butthe results obtained for the 6 conditions when considering only individualproteinswithquantitySD30%,(illustratedusingthe filedlabelonthegraph),allowedthesameconclusions,butwitha dynamic range 0.5–1 order shorter. Supplementary data Table 3 illustrates also the deformation in the proteins concentration dynamic range observed for the six conditions compared.Someproteinswereselectedatdifferentrepresentative relativequantitiesanddistributedwithintherangeobservedinthe RS*pool.Theircorrespondingcalculatedequivalenttorawserum protein concentrations are provided, as well as known concentrations measured in serum [35], plasma concentration observed[1,36,37]orestimated[38].Therankingofabundances were affected when using prefractionation and some proteins showedquantitybelowthelowerlimitofquantitation(LLOQ)of the method. The deformation observed might be driven by compression effects and also by some truncations in the concentration range due to the depletion of some specific proteins, also visible in Fig. 3. For example the vitamin D bindingprotein(VTDP-P02774),notrelatedtocellstructurebut totheactiveanalyteofvitaminDtransport,wasnotexpectedtobe withinthelowconcentrationrange in serum.Its concentration rangesfrom193to4350mg/Linserumofhealthyindividuals(all phenotypesconsidered)[35].But,somevaluesobtainedforsome
of the kits tested (see Supplementary data Table 3) are very differentand cannotall bereliably considered(because of Qty SD>30%).ForexampleinProteoMinerTM,itisdetectedatonly0.6%
ofthevaluefoundfortheRS*poolsample.Butbindingofproteins withProteoMinerTMhasbeenshowntobedrivenbynonspecific hydrophobicityrather than by the hexapeptideaffinity. In this mechanism the hydrophobicity of each proteinpresent in the complex mixture drives their own respective binding and depletion[4].However,abundanthydrophilicproteinsasIgG or apolipoproteinA1(ApoA1)weremerelydepletedefficientlywith thismethod,whichwasalsoobservedinthisserumdataset(see Supplementary data Table 2).Importantly, some proteins were quantifiedfor somesamplesonlywitha quantity SDhigher to 30 or 50%and might or shouldthereforenot beconsidered as reliablyquantified.
3.2.4.2.Theabilitytoreachlow-abundanceproteins. Toevaluateif the prefractionations tested enabled the analysis of proteins presentatlowconcentrationsinserum(thedeepproteome),one shouldcompare theproteinsquantified aslowestonesin each sampletotheirlevelinserumdeterminedbyothertechniques.But mostoftheseserumproteinlevelsarenotavailableinliterature, despitethatvalidreferencemeasurementsorestimationsmightbe availableforplasma[36].Therefore,noreferencecouldbeusedto comparethegainin low-abundanceserumproteinsreached by thesekits.However,weanalyzedforeach processreplicate,the lowestquantitydetected.Allthekitsprovidedsimilarresultswith nosignificant difference (Mann–Withney test,Pvalue>0.05)to theseobtainedwiththeRS*poolsample(lowestsinglereplicate value=0.010ng, with a mean value over the triplicates of 0.0220.027ng). This was due to the main limitations of the experimentaldesign, wherea constantquantityof totalprotein digestwas loadedoncolumn,in addition tothe lowerlimitof quantification (LLOQ) of the technology itself. However, when consideringtheresultsexpressedintermsofcalculatedrawserum equivalent protein concentrations, and with quantity SD<30% filter, the lowest values werefound for Dermicin (P81605). Its mean concentration is 0.067
m
g/mL21% in the ProteoPrep1 20 sample. The ratio between the lowest protein quantities obtainedintheRS*poolandtheProteoPrep120is49(61when considering the proteins with quantity SD30%). The three methods enabling to measure proteins present at lower equivalent raw serum concentrations compared to RS* pool were IgG &albumin depletioncoupled withWGA enrichment, ProteoMinerTM and ProteoPrep1 20, showing itself the lowestvalue. Hence, despite that the comparison of our results with literature serum references was not always possible, we could conclude that in our experiment and data set, ProteoPrep1 20 enabled the lowest calculated equivalent raw serum concentrations of protein,with 49–61 times lowervalues than theseobtainedwiththeRS*pool.
3.2.5.ComparisonoftheresultsofProteoPrep120treatedsampleto referenceconcentrationsofproteins
ThelistofproteinsidentifiedandquantifiedintheProteoPrep1 20sampleisavailableasSupplementarydataTable4.Thistable providesthequantityofeachproteinandSD(bothexpressedin percentage oftotal proteinquantitydetected) andeach protein QuantitySD(in%).ThecalculatedProteoPrep120depletedserum concentration(
m
g/mL) of proteins equivalenttothe initialraw serum andalso,someknownreferenceconcentrationsof these proteins taken from the “high confidence plasma proteome referenceset” arereported[38].Theseconcentrationreferences inplasmaorserumareprovidedfor rankingcomparison ofthe proteinsfoundinanondepletedserumorplasmaandidentified after ProteoPrep1 20 serum depletion. Regarding the proteinslisted (191 protein identified in 3R3 and 178 quantified) only 38proteins(inbold)wereidentifiedonlyafterusingthisdepletion method(=“uniqueproteins”).Comparedtothe1929noncanonical sequencesreportedbyFarrahetal.,45proteinswereidentified onlyinouranalysis.Thesearenotallrelatedtothecomplement activationand thecoagulationprocessaswould beexpectedin serumcomparedtoplasma.Theseproteinsmightalsoberelatedto thediseasestateofthepatientseraincludedinourRS*pool. 3.3.Generalconclusions
The five methods tested together with label-free proteomic analysis on a 2D-nanoUPLC–MSE
Synapt G2 system showed acceptable repeatabilities, similar to the one obtained for the RS* pool sample analysis. All five methods showed equivalent performancesregardingtotalproteinquantityrecovery. Surpris-ingly,theRS*poolalreadyprovidedveryinterestingresultswith this system settings, which illustrates its power in terms of peptidesandionsseparationsandproteinsidentifications,aswell asrelativequantification.But,someoftheprefractionationstested didnotaddmuchinformationovertherawserumanalysis.Thisis worthknowing before spending time,energy and resourceson precious clinical samples prefractionation. In summary, in our hands,theProteoPrep120 whichis theonlyfullyre-usablekit present the following advantages. First, it requires the lowest volumeorquantityofpreciousclinicalsera.Seconditenablesthe maximum number of protein identifications increase (49%) compared to theRS* pool with the highest number of unique proteins.Howeverotherdepletionprotocolsmightevenbemore valuableif verystringent QtySD threshold havetobe applied. TheseinvolvesIgG&albumindepletionstrategy.Nevertheless,the third main advantage of the ProteoPrep1 20 relies on the possibilitytoreachupto61times lowerproteinconcentration valuescomparedtonondepletedserumwhicharelowabudance proteinnotalwaysdetectableforotherkits.
Butthisdepletionkitisnotperfectasitisexpensiveandtime consuming(twosuccessivecycleofdepletionswerenecessaryto reachthe purity obtainedin this work). This antibodyaffinity capturespincolumnisareusableformatanddespitethewarranty that one hundred applications might be done without losing depletionefficiency,biasinsampleprocessingforaclinicalsample setduetotheorderofsamplesprefractionationcouldstilloccur. Hence, the order of samples should always be randomized, shufflingdiseases and controls(andlikely QC)tominimizethe potential effect as such a bias risk. Moreover, potential cross-contaminationsremainpossible,evenifcarefulreconditioningof thespincolumnisperformedbetweenthedepletioncycles.
Altogether,albeitnotperfect,ProteoPrep120appearedinthis experimentaldesignanddatasettobeasuitablecompromisefor application asoff-line serum samplesprefractionation strategy. The results of a differential label-free proteomic analysis for clinicalbiomarkersdiscoveryphasewouldprofitfromthistypeof prefractionation as long as results are analyzed knowing the technicalrepeatabilityoftheglogalstrategy.
Conflictofinterest
Theauthorsdeclarethattherearenoconflictsofinterest. Acknowledgements
WethanktheGIGA ProteomicFacilityfromULg,involvedin shotgun proteomic analyses and Mrs. Trzpiot for technical assistance.
Thisresearchwassupportedbythe“CentreAnti-Cancereux”, CHU de Liège and the FEDER action: “Fédération Wallonie Bruxelles”.
ThisworkwasalsomadepossiblethankstothesupportofMSD. NSwasaFNRSlogisticcollaborator.
AppendixA.Supplementarydata
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j. euprot.2015.07.009.
References
[1]M.Polanski,N.L.Anderson,Alistofcandidatecancerbiomarkersfortargeted proteomics,Biomark.Insights1(2007)1–48.
[2]M.Pernemalm,R.Lewensohn,J.Lehtio,AffinityprefractionationforMS-based plasmaproteomics,Proteomics9(2009)1420–1427.
[3]N.L.Anderson,N.G.Anderson,Thehumanplasmaproteome:history, character,anddiagnosticprospects,Mol.Cell.Proteomics1(2002)845–867. [4]E.M.Keidel,D.Ribitsch,F.Lottspeich,Equalizertechnology–equalrightsfor
disparatebeads,Proteomics10(2010)2089–2098.
[5]B.Domon,R.Aebersold,Optionsandconsiderationswhenselectinga quantitativeproteomicsstrategy,Nat.Biotechnol.28(2010)710–721. [6]J.R.Whiteaker,H.Zhang,J.K.Eng,R.Fang,B.D.Piening,L.C.Feng,etal.,
Head-to-headcomparisonofserumfractionationtechniques,J.ProteomeRes.6(2007) 828–836.
[7]S.J.Geromanos,J.P.Vissers,J.C.Silva,C.A.Dorschel,G.Z.Li,M.V.Gorenstein, etal.,Thedetection,correlation,andcomparisonofpeptideprecursorand productionsfromdataindependentLC–MSwithdatadependantLC–MS/MS, Proteomics9(2009)1683–1695.
[8]G.Z.Li,J.P.Vissers,J.C.Silva,D.Golick,M.V.Gorenstein,S.J.Geromanos, Databasesearchingandaccountingofmultiplexedprecursorandproduction spectrafromthedataindependentanalysisofsimpleandcomplexpeptide mixtures,Proteomics9(2009)1696–1719.
[9]Y.S.Choi,Reachingforthedeepproteome:recentnanoliquidchromatography coupledwithtandemmassspectrometry-basedstudiesonthedeep proteome,Arch.Pharm.Res.35(2012)1861–1870.
[10]Q.Zhang,V.Faca,S.Hanash,Miningtheplasmaproteomefordisease applicationsacrosssevenlogsofproteinabundance,J.ProteomeRes.10(2011) 46–50.
[11]L.Huang,X.Fang,Immunoaffinityfractionationofplasmaproteinsbychicken IgYantibodies,MethodsMol.Biol.425(2008)41–51.
[12]M.P.Smith,S.L.Wood,A.Zougman,J.T.Ho,J.Peng,D.Jackson,etal.,A systematicanalysisoftheeffectsofincreasingdegreesofserum
immunodepletionintermsofdepthofcoverageandotherkeyaspectsin top-downandbottom-upproteomicanalyses,Proteomics11(2011)2222–2235. [13]C.Tu,P.A.Rudnick,M.Y.Martinez,K.L.Cheek,S.E.Stein,R.J.Slebos,etal.,
Depletionofabundantplasmaproteinsandlimitationsofplasmaproteomics, J.ProteomeRes.9(2010)4982–4991.
[14]W.J.Qian,D.T.Kaleta,B.O.Petritis,H.Jiang,T.Liu,X.Zhang,etal.,Enhanced detectionoflowabundancehumanplasmaproteinsusingatandem IgY12-SuperMiximmunoaffinityseparationstrategy,Mol.Cell.Proteomics7(2008) 1963–1973.
[15]T.Shi,J.Y.Zhou,M.A.Gritsenko,M.Hossain,D.G.Camp2nd,R.D.Smith,etal., IgY14andSuperMiximmunoaffinityseparationscoupledwithliquid chromatography-massspectrometryforhumanplasmaproteomicsbiomarker discovery,Methods56(2012)246–253.
[16]P.G.Righetti,E.Boschetti,L.Lomas,A.Citterio,Proteinequalizertechnology: thequestforademocraticproteome,Proteomics6(2006)3980–3992. [17]P.G.Righetti,E.Boschetti,TheProteoMinerandtheFortyNiners:searchingfor
goldnuggetsintheproteomicarena,MassSpectrom.Rev.27(2008)596–608. [18]C.Sihlbom,I.Kanmert,H.Bahr,P.Davidsson,Evaluationofthecombinationof beadtechnologywithSELDI-TOF-MSand2-DDIGEfordetectionofplasma proteins,J.ProteomeRes.7(2008)4191–4198.
[19]L.Guerrier,S.Claverol,F.Fortis,S.Rinalducci,A.M.Timperio,P.Antonioli,etal., Exploringtheplateletproteomeviacombinatorial,hexapeptideligand libraries,J.ProteomeRes.6(2007)4290–4303.
[20]R.C.Dwivedi,O.V.Krokhin,J.P.Cortens,J.A.Wilkins,Assessmentofthe reproducibilityofrandomhexapeptidepeptidelibrary-basedprotein normalization,J.ProteomeRes.(2010)1144–1149.
[21]J.Rivers,C.Hughes,T.McKenna,Y.Woolerton,J.P.Vissers,J.I.Langridge,etal., Asymmetricproteomeequalizationoftheskeletalmuscleproteomeusinga combinatorialhexapeptidelibrary,PLoSOne6(2011)e28902.
[22]J.E.Bandow,Comparisonofproteinenrichmentstrategiesforproteome analysisofplasma,Proteomics10(2010)1416–1425.
[23]M.DeBock,D.deSeny,M.A.Meuwis,A.C.Servais,T.Q.Minh,J.Closset,etal., Comparisonofthreemethodsforfractionationandenrichmentoflow molecularweightproteinsforSELDI-TOF-MSdifferentialanalysis,Talanta82 (2010)245–254.
[24]L.A.Echan,H.Y.Tang,N.Ali-Khan,K.Lee,D.W.Speicher,Depletionofmultiple high-abundanceproteinsimprovesproteinprofilingcapacitiesofhuman serumandplasma,Proteomics5(2005)3292–3303.
[25]Z.Cao,H.Y.Tang,H.Wang,Q.Liu,D.W.Speicher,Systematiccomparisonof fractionationmethodsforin-depthanalysisofplasmaproteomes,J.Proteome Res.(2012).
[26]D.C.Liebler,A.J.Ham,Spinfilter-basedsamplepreparationforshotgun proteomics,Nat.Methods6(2009)785.
[27]A.Hakimi,J.Auluck,G.D.Jones,L.L.Ng,D.J.Jones,Assessmentof
reproducibilityindepletionandenrichmentworkflowsforplasmaproteomics using label-free quantitative data-independent LC–MS, Proteomics(2013).
[28]M.Fleron,Y.Greffe,D.Musmeci,A.C.Massart,V.Hennequiere,G.Mazzucchelli, etal.,Novelpost-digestisotopecodedproteinlabelingmethodfor phospho-andglycoproteomeanalysis,J.Proteomics73(2010)1986–2005.
[29]J.C.Silva,R.Denny,C.Dorschel,M.V.Gorenstein,G.Z.Li,K.Richardson,etal., SimultaneousqualitativeandquantitativeanalysisoftheE.coliproteome:a sweettale,Mol.Cell.Proteomics5(2006)589–607.
[30]J.C.Silva,M.V.Gorenstein,G.Z.Li,J.P.Vissers,S.J.Geromanos,Absolute quantificationofproteinsbyLCMSE:avirtueofparallelMSacquisition,Mol. Cell.Proteomics5(2006)144–156.
[31]J.P.Vissers,J.I.Langridge,J.M.Aerts,Analysisandquantificationofdiagnostic serummarkersandproteinsignaturesforGaucherdisease,Mol.Cell. Proteomics6(2007)755–766.
[32]L.Urbas,P.Brne,B.Gabor,M.Barut,M.Strlic,T.C.Petric,etal.,Depletionof high-abundanceproteinsfromhumanplasmausingacombinationofan affinityandpseudo-affinitycolumn,J.Chromatogr.A1216(2009)2689–2694. [33]K.Tang,etal.,Chargecompetitionandthelineardynamicrangeofdetectionin electrosprayionizationmassspectrometry,J.Am.Soc.MassSpectrom.15(10) (2004)1416–1423.
[34]L.Guerrier,P.G.Righetti,E.Boschetti,Reductionofdynamicprotein concentrationrangeofbiologicalextractsforthediscoveryoflow-abundance proteinsbymeansofhexapeptideligandlibrary,Nat.Protoc.3(2008)883– 890.
[35]D.Blanton,Z.Han,L.Bierschenk,M.V.Linga-Reddy,H.Wang,M.Clare-Salzler, etal.,ReducedserumvitaminD-bindingproteinlevelsareassociatedwith type1diabetes,Diabetes60(2011)2566–2570.
[36]G.L.Hortin,D.Sviridov,N.L.Anderson,High-abundancepolypeptidesofthe humanplasmaproteomecomprisingthetop4logsofpolypeptideabundance, Clin.Chem.54(2008)1608–1616.
[37]B.B.Haab,B.H.Geierstanger,G.Michailidis,F.Vitzthum,S.Forrester,R.Okon, etal.,ImmunoassayandantibodymicroarrayanalysisoftheHUPOplasma proteomeprojectreferencespecimens:systematicvariationbetweensample typesandcalibrationofmassspectrometrydata,Proteomics5(2005)3278– 3291.
[38]T.Farrah,E.W.Deutsch,G.S.Omenn,D.S.Campbell,Z.Sun,J.A.Bletz,etal.,A high-confidencehumanplasmaproteomereferencesetwithestimated concentrationsinPeptideAtlas,Mol.Cell.Proteomics10(2011)M110.6353.