approach exchange in cation chromatography, variants Part II: pHgradient Method development the separation for of monoclonal antibodycharge Journal of Pharmaceutical and Biomedical Analysis

ContentslistsavailableatScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

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

Method development for the separation of monoclonal antibody charge variants in cation exchange chromatography, Part II: pH gradient approach

Szabolcs Fekete

, Alain Beck

, Jen ˝o Fekete

, Davy Guillarme

aSchoolofPharmaceuticalSciences,UniversityofGeneva,UniversityofLausanne,Boulevardd’Yvoy20,1211Geneva4,Switzerland

bCenterofImmunologyPierreFabre,5AvenueNapoléonIII,BP60497,74160Saint-Julien-en-Genevois,France

cBudapestUniversityofTechnologyandEconomics,DepartmentofInorganicandAnalyticalChemistry,Szt.Gellérttér4.,1111Budapest,Hungary

a r t i c l e i n f o

Articlehistory:

Received10July2014

Receivedinrevisedform2September2014 Accepted10September2014

Availableonline5October2014

Keywords:

Ionexchange Monoclonalantibody pHgradient Methoddevelopment Cetuximab

a b s t r a c t

ThecationexhangepHgradientapproachwasevaluatedforthecharacterizationof10modelmonoclonal antibodiesincluding panitumumab,natalizumab,cetuximab,bevacizumab,trastuzumab, rituximab, palivizumab,adalimumab,denosumabandofatumumab.

ThisworkshowsthatretentionandresolutioncanbemodelledincationexchangepHgradientmode, basedononlyfourinitialruns(i.e.twogradienttimesandtwomobilephasetemperature).Only6hwere requiredforacompletemethodoptimizationwhenusinga100mm×4.6mmstrongcationexchange column.Theaccuracyofthepredictionswasexcellent,withanaveragedifferencebetweenpredicted andexperimentalretentiontimesofabout1%.

The10modelantibodiesweresuccessfullyelutedinbothpHandsaltgradientmodes,provingthatboth modesofelutioncanbeconsideredasmulti-productchargesensitiveseparationmethods.Formostof thecompounds,thevariantswerebetterresolvedinthesaltgradientmodeandthepeakcapacitieswere alsohigherinthesaltgradientapproach.TheseobservationsconfirmthatpHgradientapproachmaybe oflowerinterestthansaltgradientcationexchangechromatographyforantibodycharacterization.

©2014ElsevierB.V.Allrightsreserved.

1. Introduction

Monoclonalantibodies (mAbs)and related productsare the fastest growing class of therapeutic agents [1]. Suitable tech- niquesarerequestedtoqualitativelyandquantitativelyanalyze heterogeneitiesrelatedtosizeandchargevariantsofmultimers andaggregates.Commonmodificationsoftheprimarysequence include N-glycosylation,methionine oxidation, proteolyticfrag- mentation,anddeamidation[2,3].

Cationexchangechromatography(CEX)hasbeenwidelyused fortheseparationofmAbschargevariants[4–7]applyingashallow gradientofincreasingsaltconcentration(e.g.sodiumchloride)at constantpH.Besidecation-exchange,anion-exchangechromatog- raphy(AEX)hasalsoseenapplicationfortheseparationofthemore basicoxidizedvariantsofanintactmAbs[7].

InadditiontochoosingtheappropriatepHofthestartingbuffer, itsionicstrength(saltconcentration)shouldbekeptlow.Thepro- teinsarethenelutedbyincreasingthesaltconcentrationtoincrease

∗Correspondingauthor.Tel.:+41223796334;fax:+41223796808.

E-mailaddresses:szabolcs.fekete@unige.ch,szfekete@mail.bme.hu(S.Fekete).

thecompetitionbetweenthebufferionsandproteinsforcharged groupsontheionexchange(IEX)resin.Thismodeofseparationis consideredasthegoldstandardforproteinscharacterization,due toitshighrobustnessandgoodresolvingpower.

Analternativeapproachfortheseparationofchargevariants consistsinapplyingapHgradient,whilstkeepingconstanttheionic strength[8].Chromatofocusing(withinternal pHgradient)was recognizedasthechromatographicanalogytoisoelectricfocusing (IEF)[9–11]andhasbeensuccessfullyappliedforseparatingpro- teinisoforms,duetoitshighresolvingpowerandabilitytoretain theproteinnativestate[12,13].Therearehoweversomelimita- tionstothisapproachsuchasthecostofpolyampholytebuffers, columnregenerationtimeandtheinflexibilityincontrollingthe slope of pH gradient [12,14,15]. Alternatively, the pH gradient canbeconductedexternallybypre-columnmixingoftwoeluting buffersatdifferentpHvaluesconsistingofcommonbufferspecies [16].In contrasttochromatofocusing, theoutletpHgradientin apHgradientIEXisaresultofthesuperimpositionofanexter- nalpH-gradient(withrespecttotime)overan internalcolumn pH-gradient(withrespecttothecolumnlength)[17].Therefore, pHgradientIEXgeneratesawellcontrollableandrepeatablepH- gradientoverawiderangeofpH.

http://dx.doi.org/10.1016/j.jpba.2014.09.032 0731-7085/©2014ElsevierB.V.Allrightsreserved.

ForpHgradientIEXmode,theuseofamixtureofaminebuffer- ingspeciesinthehigh-pHrangeandamixtureofweakacidsin thelow-pHrangeisquitecommon[14,18,19].Insuchasystem, maintaininglinearityofthepHgradientslopemaybesomewhat difficult.FortheseparationofmAbvariants,severalbuffersystems weretestedinCEXmode.Itwasshownthatanappropriatemix- tureofTris base,piperazineandimidazoleprovidesa linearpH gradientfrompH6to9.5[20].Triethylamineanddiethylamine basedbuffer systemsalso offeredlinear pHgradientin thepH rangeof7.5–10.0[16].Formassspectrometry(MS)detection,5mM ammoniumhydroxidein20%methanolyieldedareasonablepH gradientinalimitedpHrange[16].Finally,Zhangetal.applieda salt-mediatedimprovedpHgradientthatwasusedinawidepH range(between5 and10.5)[21].In theirstudy,a0.25mM/min sodium-chloridegradientwasperformedtogetherwiththepHgra- dient.MAbspossessingisoelectricpoints(pI)between6.2and9.4 weresuccessfullyelutedinonegenericmethodbyusingthissalt- mediatedpHgradient[21].

OneofthebenefitsofpHgradientbasedIEXis thatthesalt concentrationcanbekeptlow,yieldinglessbufferinterferences (e.g.on-lineoroff-linetwo-dimensionalLC).Ontheotherhand, pHgradientbasedseparationusingaCEXcolumnwasfoundtobe amultiproductchargesensitiveseparationmethodformonoclonal antibodies[8,21].

Inthissecondpartofthestudy,pHgradientCEXwasappliedto separatemAbchargevariants.Theimpactofgradientsteepnessand mobilephasetemperatureonretention,peakcapacityandselec- tivitywasstudiedindetailsusingsixselectedmodel mAbsand theirvariants(i.e.trastuzumab,panitumumab,natalizumab,ritux- imab,adalimumabandcetuximab).Finally,thepossibilitiesofpH andconventionalsaltgradientmodeswerecriticallyevaluatedfor 10mAbspossessingpIbetween6.7and9.1.

2. Experimental

2.1. Chemicalsandcolumns

Waterwasobtainedfroma Milli-QPurification Systemfrom Millipore (Bedford, MA, USA). CX-1 pH gradient buffer A (pH 5.6) and CX-1 pH gradient buffer B(pH 10.2) were purchased fromThermoFisherScientificAG(Reinach,Switzerland).1M2- (N-morpholino)ethanesulfonicacid(MES) solution(BioReagent), 1M sodium hydroxide (NaOH) solution and sodium chloride (NaCl)(BioChemika)werepurchasedfromSigma–Aldrich(Buchs, Switzerland).

FDAandEMAapprovedtherapeuticIgGmonoclonalantibodies includingpanitumumab, natalizumab, cetuximab, bevacizumab, trastuzumab, rituximab, palivizumab, adalimumab, denosumab andofatumumabwerekindlyprovidedbytheCenterofImmunol- ogyPierreFabre(Saint-JulienenGenevois,France).Papain(from Caricapapaya),usedforfragmentationofmAbswasobtainedfrom Sigma–Aldrich(Buchs,Switzerland).

YMCBioProSP-F 100mm×4.6mm,5␮mnon-porous strong cationexchangecolumnwaspurchasedfromStacroma(Reinach, Switzerland).

2.2. Equipmentandsoftware

Alltheexperimentswereperformedusinga WatersAcquity UPLC^TMsystemequippedwithabinarysolventdeliverypump,an autosamplerandfluorescencedetector(FL).TheWatersAcquity systemincludeda5␮lsampleloopanda2␮lFLflow-cell.Theloop isdirectlyconnectedtotheinjectionswitchingvalve(noneedle seatcapillary).Theconnectiontubebetweentheinjectorandcol- umninletwas0.13mmI.D.and250mmlong(passivepreheating

included),andthecapillarylocatedbetweenthecolumnanddetec- torwas0.10mmI.D.and150mmlong.Theoverallextra-column volume(Vext)isabout14␮lasmeasuredfromtheinjectionseatof theauto-samplertothedetectorcell.Themeasureddwellvolume isaround100␮l.Dataacquisitionandinstrumentcontrolwasper- formedbyEmpowerPro2Software(Waters).Calculationanddata transferringwasachievedbyusingExceltemplates.

ThemobilephasepHwascheckedandadjustedusingaSeven- MultiS40pHmeter(MettlerToledo,Greifensee,Switzerland).

MethodoptimizationwasperformedusingDryLab^®2000Plus chromatographic modelling software (Molnar-Institute, Berlin, Germany).

2.3. Apparatusandmethodology

2.3.1. Mobilephasecompositionandsamplepreparation

ForthepHgradientCEXseparationofmAbsandtheirfragments, themobilephase“A”wasa10timesdilutedCX-1pHgradientbuffer A(pH5.6)whilethemobilephase“B”wasa10timesdilutedCX- 1pHgradientbufferB(pH 10.2)–asdescribedin theprotocol providedbythevendor(ThermoFisherScientific).

Forthesaltgradientseparations,themobilephase“A”consisted of10mMMESinwater,whilethemobilephase“B”was10mMMES inwatercontaining1MNaCl.ThepHofbothmobilephaseswas adjustedbyadding1MNaOHsolutiontoreachtherequiredpH.

Thedigestionofcetuximabwasinitiatedbyadditionofpapain (dilutedto100␮g/mlwithwater)toreachafinalprotein:enzyme ratioof100:1(m/m%).Thedigestionwascarriedoutat37^◦Cfor3h.

Thefinaldigestionvolumewas200␮landdirectlyinjectedusing lowvolumeinsertvials.

2.3.2. Investigationofretentionpropertiesofantibodies

IntactantibodieswereelutedinpHgradientmode.Forstudying theretentionpropertiesofintactmAbs,sixofthe10availableanti- bodieswereselectedbasedontheirtype(IgGclassandisotype) andcalculatedpI,namelypanitumumab(huIgG2,pI=6.7),natal- izumab(hzIgG4,pI=8.6),cetuximab(chIgG1,pI=8.7),adalimumab (huIgG1, pI=8.8), trastuzumab (hzIgG1, pI=8.8) and rituximab (chIgG1,pI=9.1).BecauseIgGfromdifferentsubclassesmayhave differentproperties(Fcregions arevery closebutthestructure of thehinge regions may berelatively different), therefore the strengthofinteractionsbetweentheion-exchangeresinandIgG canvaryconsiderablybetweenspeciesandsubtypes.Ourpurpose wasto cover thewhole pIrange and toinclude chimeric (ch), humanized (hz)andhuman(hu)referenceIgG1, IgG2and IgG4 isotypes,todrawoverallandreliableconclusions.

First,theeffectofpHgradientsteepnessontheretentionwas evaluated.Differentgradienttimesweretestedatagivenmobile phasetemperature.Agenericlineargradient,startingfrom0%to 100%B(equivalenttoalinearpHgradientfrom5.6to10.2)was appliedataflowrateof0.6ml/minforallsamples.Thegradient time(tg)wasvariedas10,15,20,30and40min(atT=30^◦C).The observedapparentretentionfactors(kapp)andpeakcapacity(Pc) valueswereplottedagainstthegradienttime(steepness).

Fortheinvestigationofmobilephasetemperature,15mingra- dientruns(0–100%B)werecarriedoutusingvarioustemperatures between 30^◦C and theupper temperature limit of thecolumn (60^◦C).TheretentionpropertiesofintactmAbsandtheircharge variantswereevaluatedbyplottingthelogarithmkappagainst1/T (Van’tHofftyperepresentation).Peakcapacityandresolutionwere alsostudiedasafunctionofmobilephasetemperature.

2.3.3. Systematicmethodoptimization

A commonapproach in methoddevelopment is tosimulta- neously model the selectivity and/or resolution as a function of temperature and gradient steepness on a selected column

fromalimitednumberofinitialruns[22,23].Inthiscase,initial experimentsarethoseonwhichthecomputer-modelsarebased (calculated)tomodelafewthousandexperiments.Then,withthe helpofresolutionmaps–whichshowthecriticalresolutionofthe peakstobeseparated–thegradientprogramandcolumntemper- aturecanberapidlyandaccuratelyoptimized.Thisapproachwas currentlyappliedforthesaltgradientCEXseparationofantibody variants[24].Inthissecondpart,thisprocedurewasimplemented forpHgradientCEXbasedseparations,toevaluateifthesamerules formethoddevelopmentcanbeapplied.

Basedontheobservedeffectsofthefactorsonretentionand resolution of mAbs peaks, a 4 runs based initial experimental setupisrecommendedformethodoptimizationinthepHgradi- entmode.Performinggradientrunswithtwogradienttimes(as tg1=10min,tg2=30min)attwotemperature(T1=25^◦C,T2=55^◦C) ona100mm×4.6mmcolumnallowedareliableoptimizationof theseparation.

Theoptimizationwasperformedbycomputersimulationusing aDryLabtwodimensionalmodel.Cetuximabpapaindigestedsam- pleswereinjectedtobuilduptheDryLabmodelandstudythe predictionaccuracyerror.CetuximabisaheterogeneousmAbpos- sessingtwoN-glycosylationsitesintheheavychainandseveral chargevariantsincludingC-terminallysinesandsialicacids[25].It isthereforeacomplexexampleformethoddevelopment[24].

2.3.4. GenericpHgradientformultiproductanalysis

Afterstudyingtheretentionbehaviorofa limitednumberof modelantibodies,a genericpHgradientwasproposedallowing theelutionandseparationofallthe10mAbswithinreasonable analysistime(20min).ApHgradient,startingfrompH5.6to10.2 wasappliedataflowrateof0.6ml/minforallsamples.Themobile phasetemperaturewassetatT=30^◦C.Fluorescencedetectionwas carriedoutat_ex=280and_em=360nm.

2.3.5. ComparisonofpH-gradientandsalt-gradientCEX

Thegenericmethodsforthe10mAbsandtheoptimizedsepa- rationofcetuximabfragmentsinbothsaltandpHgradientmodes werecompared.Thesamestationaryphaseandmobilephaseflow ratewereemployedallowinga faircomparison ofthetwoCEX modes.Peakcapacity,selectivityandelutionorderwerestudied.

Themethodoptimizationandfinalconditionsemployedforthe saltgradientCEXmethodshavebeendetailedinthefirstpartof thisarticleseries[24].

3. Resultsanddiscussion

InpHgradientmode,theproteinsnetchargecanbemodified duringthepHgradient,duetoprotonation–deprotonationoffunc- tionalgroups.InCEX,theproteinisexpectedtoeluteat,orcloseto itspI.Accordingtotheory,whenusingpHgradientelutionmode andlowionicstrengthmobilephase,theproteinsarefocusedin narrowerbandsenablinghigherresolutioncomparedtoapHgra- dientperformedathighionicstrength.Thewidthofaproteinpeak alongalinearpHgradientexpressedinpHunitscanbewrittenas follows[9,10,16]:

(pH)²≈ D

dpH/dV

ϕ

dZ/dpH

whereDisthediffusioncoefficientoftheanalyte,dpH/dVisthegra- dientslope,ϕistheDonnanpotentialanddZ/dpHisthechangein proteinnetchargealongthepHgradient.SincetheDonnanpoten- tialdependsontheionicstrength,apeakfocusingeffectisexpected atlowerionicstrength.In agreementwiththis expectation,pH gradientsatlowionicstrengthshowedbetterresolutionformAb variantscomparedtopHgradientsperformedinhighionicstrength

medium[16].However,themechanismof saltgradientandpH gradientmodeisdifferentandthereforehardlycomparable.

To achieve an optimalexperimentalsetup for separation or purificationofmAbs,theinfluenceofvariousparametersonsep- aration,suchasgradientsteepness,temperatureorflowratehas tobetakenintoaccount.TheoptimizationofmAbseparationsin CEXmodewasperformedbythesystematicvariationofgradient steepnessandflowrateandtheirimpactonthepeakwidthand resolution.Thisapproachisknownasiso-resolutioncurveconcept andallowsdevelopingastepwisegradient[26].

TheworkofSnyderandco-workersshowedthatsaltgradient basedIEX systemsfollow non-linearsolventstrength(non-LSS) typemechanism[27–29].Inthefirstpartofourstudy,itwasshown thatwhenthenetchargeofaproteinislarge(e.g.suchaswith mAbs), thesocalled stoichiometricdisplacement model (SDM) givesvirtuallylinearretentiondependenceonthegradientspan [24].

Tothebestofourknowledge,theoptimizationprocedureinpH gradientmodeCEXhasnotbeenyetstudied.Here,theimpactof gradientsteepnessandtemperatureonmAbsretentiontimeand peakcapacitywasevaluatedinasystematicway.

3.1. TheeffectofpHgradienttime(gradientsteepness)onthe retention

ForCEXseparationofmAbvariants,thesolutesareelutedin orderofincreasingbindingcharge(correlatesmoreorlesswith thepI)andequilibriumconstant.ItisgenerallyassumedthatmAbs eluteclosetotheirpIinthepHgradientmode[17].Therefore,the appliedpHrangeclearlydeterminestheproteinsthatcanpossi- blybeeluted.Ontheotherhand,retentiontimesandpeakwidths dependonthegradientsteepnessasbotharefunctionofdZ/dpH.

Duetotherelativelyhighnetcharge(zvalue)ofmAbs asmall changeinpHcouldleadtosignificantshiftinretention.

Theeffectofgradientsteepness(gradienttime)onthereten- tionofintactmAbsandtheirvariantswasstudiedinasystematic way.Thegradienttime(steepness)wasvariedas10,15,20,30and 40min(atT=30^◦C).TheretentionofthesixselectedmAbsand thevariantsoftrastuzumab,adalimumabandcetuximabshowed thesamebehaviour.Fig.1illustratestheeffectofgradienttimeon theapparentretention(kapp)ofintactmAbsandchargevariants.

Therelationbetweenkappandtgcanbeaccuratelydescribedby fittinglinearfunction(R²>0.999forallsolutes).Onthecontrary, theretentionbehaviourofmAbfragmentsinRPLCshowedamod- eratedeviationfromlinearrelationship[22].Thisbehaviourwas explainedbysomepossiblechangesinconformationandcontri- butionofdifferentretention mechanisms(hydrophobicand ion exchange interactions and/or irreversibleadsorption)[30]. Sur- prisingly,inpHgradientCEXmode,anLSStypemodelperfectly describestheretentionbehaviourofmAbs.Thereisnoneedfor logarithmicorpolynomialfitting(asitisoftenappliedinRPLCor NPLCmodes).

Assuggestedbyourobservations,theretentiontimeofintact mAbs canbepredictedfor anygradientsteepness onthebasis of only two initial gradient runs (e.g. with t_g1=10min and t_g2=30min)inthepHgradientmode.

3.2. Theeffectofmobilephasetemperatureonretention

The temperature dependence of analyte retention factor in liquid chromatographyis generally expressedby thevan’t Hoff equation.With regular compounds,the van’t Hoffplots (log(k) vs 1/T) follow a linear relationship. However, with ionisable compoundsandlargebiomolecules,deviationsfromlinearitywere described[31].Dependingonthestabilityofproteinssecondary structure, the molecules unfold to various extents and hence

Fig. 1.Effect of gradient time (steepness) on the apparent retention factor ofnativeantibodies(A)andantibodyvariants(B).Column:YMC BioProSP-F (100mm×4.6mm).Mobilephase“A”CX-1BufferApH5.6,“B”CX-1BufferBpH 10.2.Flowrate:0.6ml/min,gradient:0–100%Bin10,15,20,30and40min,temper- ature:30^◦C,detection:FL(280–360nm),injectedvolume:2␮l.

interactwiththestationaryphasewithvariousstrengths[32–34].

TheeffectofmobilephasetemperatureonproteinretentioninpH gradientIEXmodehasnotyetbeenreported.

Fig.2illustratestheobtainedvan’tHofftypeplots.Thelog(k) vs1/Tplotsshowlinearbehaviourintheinvestigatedtemperature range.SimilarlytotheconclusionsdrawninsaltgradientbasedCEX separations,theimpactoftemperatureseemstobelessimportant comparedtoRPmode[24].Theslopeofthecurveswasindeed significantlylowerinIEXvs.RPLCmode.TheslopeinpHgradient IEXmodewascomprisedbetween−0.47and0.11K*10³,whilefor intactmAbsinRPLCmodetheslopeistypicallyaround0.5and 1.0K*10³[22].

Panitumumab (possessing the lowest pI) showed the most importantretention–temperaturedependence.Itisprobablydueto derelativelylownumberofnetchargeswhilstrunningthepHgra- dientbetweentheinitialmobilephasepH(5.6)andpanitumumab pI(6.7).Theselow netchargesformrelativelyweakinteraction withthestationaryphasecomparedtotheothermAbs.Moreover, theretentionofpanitumumabincreaseswithtemperatureproba- blyduetoamodificationofpIwithtemperature.

Finally,theslopesoflog(k)–1/Tcurves forrelatedmAbs(e.g.

chargevariantsofa givenmAb)arequitesimilar.Fig.2Bshows theseplotsforadalimumab,cetuximabandtrastuzumab,asrep- resentativeexamples.Thissuggeststhatselectivitycanhardlybe tunedwith temperature. However, temperature has significant impactonthepeakwidth(peakcapacity)–seeSection3.3–and hasaneffectontheoverallseparationquality(resolution).

Fig.2.Effectoftemperatureontheapparentretentionfactorofnativeantibodies (A)andantibodyvariants(B)(van’tHofftyperepresentation).Column:YMCBioPro SP-F(100mm×4.6mm).Mobilephase“A”CX-1BufferApH5.6,“B”CX-1BufferB pH10.2.Flowrate:0.6ml/min,gradient:0–100%Bin15min,temperature:30,40, 50and60^◦C,detection:FL.(280–360nm),injectedvolume:2␮l.

3.3. Peakcapacity

Thevariationofpeakcapacityasafunctionofgradientsteepness andtemperaturewasalsoestimated.Thefollowingequationwas usedtoestimatepeakcapacitybasedonpeakwidthathalfheight [35]:

Pc=1+ tg

1.7·w_50% (2)

A logarithmic relationship perfectly described the evolution ofpeak capacity withincreasinggradient time.Similarly tothe observationsmadeinsalt-gradientCEXmodewithtrastuzumab, adalimumab and natalizumab, the peak capacities achieved in pHgradientCEXforthesamemAbsandusingthesamecolumn and gradient steepnesswerecomparable [24]. Peakcapacity of Pc∼50–60wasobservedwitha 10minlonggradient,whilethe longest 40min gradient provided Pc∼100–120 (Fig. 3A). Since retentionalsoincreasedwithgradienttime(Fig.1),relativelylong gradientsarerequiredforhighresolutionseparations,albeitthe costofanalysistime.

Surprisingly,panitumumabandrituximabelutedwithsignifi- cantlywiderpeaksthantheothermAbs.Thepeakcapacitywas comprisedbetween40and75forpanitumumabandonlybetween 20and60forrituximab.Thisefficiencywasclearlylowerthanthe oneachievableinsaltgradientCEXmodeonthesamecolumnand usingthesamegradientsteepness[24].

Fig.3Bshowsthechangeinpeakcapacityasafunctionofmobile phasetemperatureatagivengradientsteepness.Adecreaseinpeak capacitywasobservedforhalfofthemAbswhiletheotherhalf

Fig.3.Peakcapacityasafunctionofgradienttime(steepness)(A)andtemperature (B).Column:YMCBioProSP-F(100mm×4.6mm).Mobilephase“A”CX-1BufferA pH5.6,“B”CX-1BufferBpH10.2.Flowrate:0.6ml/min,detection:FL(280–360nm), injectedvolume:2␮l.

showedtheoppositetendency.Themostdifferentbehaviourwas observedforadalimumabandnatalizumab.Foradalimumab,the peakcapacitywasreducedfrom80to60whenincreasingthetem- peraturefrom30to60^◦C,whilefornatalizumabitwasenhanced from70 to80inthesametemperaturerange. Thisobservation suggeststhattemperaturemaybeanimportantfactorinmethod optimization.Indeed,selectivitydoesnotchangesignificantlywith temperaturebut itcan affectresolutionthrough anincrease or decreaseofpeakcapacityinpHgradientCEXmode.

ForlargeproteinsinCEXmode,itisalwayshardtopredictand explaintheirbandbroadeningprocessoccurringatdifferenttem- peratures.Ontheonehand,duetothechangesinmobilephase viscosityandsolutediffusion,anincrease inefficiency (sharper peak)isexpected.Ontheotherhand,temperaturecanhaveastrong impactonproteinconformationaswell.Undercertainconditions, thenativeconformationand/orotherintermediateconformations maybepresentduringtheanalysis.Eachofthesewillinteractdif- ferentlywiththestationaryphase,resultinginslightvariationsin retentiontimesofthedifferentspecies.Theexistenceofthediffer- entspeciesatelevatedtemperaturemayresultin“apparent”peak broadening.

3.4. Methoddevelopment,creatingatwodimensionalDryLab modelforpHgradientCEX

Linearmodels are generally employed for the simultaneous optimizationoftwoorthreevariablesinliquidchromatography.

Polynomialrelationshipoftwovariablescanbewrittenas:

y=b₀+b₁x₁+b₂x₂ (3)

whereyistheresponse(retentiontimeoritstransformation),x₁ andx2arethemodelvariables(e.g.tgandT),whileb0,b1,b2are themodelcoefficients.

AsobservedwithmAbs,thedependenceofretentiontime(or itstransformation)onpH,gradientsteepnessandmobilephase temperaturecanbedescribedbylinearmodels.Thisobservation suggeststhatmethodoptimizationwithgradientsteepnessand mobilephasetemperatureasmodelvariablesrequiresthemea- surementofvariableeffectsattwolevelsonly.

Gradient runs with two gradient times (as t_g1=10min, tg2=30min) at two temperatures (T1=25^◦C, T2=55^◦C) on a 100mm×4.6mmcolumnwereperformedtobuildupthemodel.

The modelling software implements an interpretive approach, wheretheretentionbehaviourismodelledonthebasisofexperi- mentalruns,andthentheretentiontimes,peakwidths,selectivity andresolutionatotherconditionsarepredictedinaselectedexper- imentaldomain.Thisallowscalculatingthecriticalresolution,and accordingly,theoptimalseparationcanbefound.Forthispurpose, retentiontimesweretransformedintoretentionfactors,andlinear modelswerechosenforbothgradienttime(steepness)andtem- perature.Thismodellingwasperformedonarectangularregionin thet_g–Tplane,determinedby2gradienttimes(steepness)and2 temperatures.Hence,thisapproachrequires4initialexperimen- talrunsfor creatingthemodel. Followingtheexecution ofthe inputexperimentalruns,data(retentiontimes,peakwidthsand peaktailingvalues)wereimportedintoDryLabandpeaktrack- ing performed. Then, the optimization was carried out onthe basisofthecreatedresolutionmap,inwhichthesmallestvalue ofresolution(Rs)ofanytwocriticalpeaksinthechromatogram was plotted as a function of gradient time and mobile phase temperature.

To establish the accuracy of this 2 dimensional linear model, the predicted and experimentally derived chro- matograms(retentiontimes)undertheoptimalconditionswere compared.

3.5. OptimizationoftheseparationofFabandFcfragmentsof cetuximabinpHgradientmode

Afastandefficientmethodoptimizationprocesswasappliedfor thedeterminationofvariantsanddegradationproductsofrecom- binantcetuximab,usingthepHgradientapproachinCEXmode.

ThenativemAbwasinitiallydigestedwithpapainandourpurpose wastoseparateasmanyvariantsoftheFabandFcfragmentsas possible,withintheshortestachievableanalysistime.Fig.4shows theobtainedchromatogramsofthefourinitialruns.Thepredicted resultsaredemonstratedinFig.5asaresolutionmap.Basedon theresolutionmap, a16mingradientwasfoundtoprovidethe highestresolution atthemobilephasetemperature ofT=25^◦C.

Then,thepredictedoptimumconditionwassetandexperimen- talchromatogramsrecordedtoevaluatethepredictionaccuracy.

Fig. 6 shows thepredicted and experimentally observed chro- matograms,whileTable1providesthecorrespondingretention times.

AsshowninTable1,thepredictedretentiontimeswereingood agreementwiththeexperimentalones.Theaverageretentiontime relativeerrorswassystematicallyunder1.0%(seeTable1),which canbeconsideredasexcellent.Thehighest individualdeviation was1.5%.

Inconclusion,thismethodoptimizationapproachcanbecon- sideredasreliableandthesuggestedinitialexperiments(i.e.10 and30mingradientona100mmlongstandardborecolumnat 25and55^◦C)aresuitablefordailyroutinework.Thetimespent formethoddevelopmentinthisexamplewasaround6h(2gradi- ents×2temperatures×3samples+equilibrationtime).

Fig.4. Cetuximabpapaindigestedsample.Column:YMCBioProSP-F(100mm×4.6mm).Mobilephase“A”CX-1BufferApH5.6,“B”CX-1BufferBpH10.2.Flowrate:

0.6ml/min,gradient:0–80%B,detection:FL(280–360nm),injectedvolume:2␮l.Gradienttimes:tg1=10min,tg2=30min,temperaturesT1=25^◦C,T2=55^◦C.

Fig.5.Cetuximabpapaindigestionresolutionmap(tg–Tmodel).Column:YMC BioProSP-F (100mm×4.6mm). Mobilephase “A”CX-1 Buffer ApH 5.6, “B”

CX-1BufferBpH10.2.Flowrate:0.6ml/min,gradient:0–80%B,detection:FL (280–360nm),injectedvolume:2␮l.Gradienttimes:tg1=10min,tg2=30min,tem- peraturesT1=25^◦C,T2=55^◦C.

Table1

Predictionaccuracy.ConditionsarethesameasspecifiedinFig.7.

Peak Retentiontime

Experimental Predicted Difference(min) Error(%)

1 3.47 3.42 0.05 1.37

2 3.79 3.74 0.05 1.34

3 4.18 4.14 0.04 1.01

4 4.72 4.68 0.04 0.88

5 5.20 5.12 0.08 1.50

6 5.92 5.97 −0.05 −0.89

7 6.57 6.48 0.09 1.33

8 6.92 7.00 −0.08 −1.09

9 7.39 7.48 −0.10 −1.27

10 7.76 7.82 −0.06 −0.73

11 8.31 8.39 −0.08 −0.99

12 9.03 9.13 −0.10 −1.10

13 9.85 9.98 −0.13 −1.30

14 11.09 11.14 −0.05 −0.47

Average −0.02 −0.03

3.6. GenericpHgradientCEXmethodforvariousmAbs

ThemainadvantageofpHgradientbasedseparationsusinga CEXcolumnisdescribedasamulti-productchargesensitivesepa- rationmethodforvariousmAbs[16,20,21].Inthisstudy,wetriedto checkthishypothesisbyanalyzingamixtureof10differentmAbs (possessingpIbetween6.7and9.1),withagenericpHgradient frompH5.6to10.2.

Fig.6. Comparisonofpredictedandexperimentalchromatograms.Column:YMC BioProSP-F(100mm×4.6mm).Mobilephase“A”CX-1BufferApH5.6,“B”CX-1 BufferBpH10.2.Flowrate:0.6ml/min,gradient:0–55%Bin16min,temperature:

25^◦C,detection:FL(280–360nm),injectedvolume:2␮l.

Fig.7. GenericpHgradient.Column:YMCBioProSP-F(100mm×4.6mm).Mobile phase“A”CX-1BufferApH5.6,“B”CX-1BufferBpH10.2.Flowrate:0.6ml/min,gra- dient:0–100%Bin20min,temperature:30^◦C,detection:FL(280–360nm),injected volume:2␮l.

BasedonSection 3.5,the pHgradientsteepnessand mobile phasetemperaturewerevariedtofindappropriateconditionsfor these10mAbsandtheirvariants.

Fig.7showstheobtainedchromatogramsof10intactmAbs, andsuggeststhatpHgradientCEXseparationisindeedadequate formulti-productmAbseparations.Theoptimalconditionsona strongcationexchanger resinwerefoundas20minlonggradi- ent(0–100%B)at30^◦C.OnFig.7,itcanbeclearlyseenthatmAbs donoteluteexactlyintheorderoftheirpI.Theoretically,proteins shouldeluteaccordingtotheirpIinthepHgradientmode.Ahamed etal.observedthatbasicproteins(pI>8)elutedatpHverycloseto theirpI,whileacidicproteins(pI<6)elutedatslightlyhigherpH (0–0.5pHunit)thantheirpI[17].Onthecontrary,neutralproteins (6<pI<8)elutedatmuchhigherpHthantheirpIandhadaten- dencytoeluteatpH∼9regardlesstheirpI[17].Thedistribution ofchargesonthesurfaceofproteinsisgenerallyconsideredasthe reasonfortheminordeviationsbetweentheelutionpHandpI.

Forneutralproteins,thesourceofdeviationisprobablyoriginated fromthepolypeptidicchain.Indeedthehugedeviationbetweenthe elutionpHandpIofneutralproteinsoriginatedfromtheinherent natureoftheproteinstitrationcurve[17].

Inourexample,natalizumabclearlyelutesearlier,whiledeno- sumabelutesathigherpHthanexpected.Onepossibleexplanation maybethedifferences in glycosylationprofiles ofthese mAbs.

Moreover,somesupplementaryinteractions withthestationary phase canalso occurthat superposes tothe charge-interaction basedelutionmechanism.Basedontheseobservationsandthefact thatretentiontimesandpIarenotperfectlycorrelated,careshould betakenwhenevaluatingtheproteinspI,basedonapHgradient CEXexperiment.

3.7. ComparisonofgenericpHandsaltgradientCEXseparations ThegenericpHgradientmethodwascomparedtotherecently developedgenericsaltgradientmode.Theconditionsofthesalt gradientmethodaredetailedinthefirstpartofthisarticleseries [24].

Forthesake ofcomparison, thesamecolumn(strong cation exchange),instrument,flowrate,gradientsteepness,temperature,

Fig.8.Comparisonofretentionfactors(A)andpeakcapacity(B)insaltandpHgra- dientmodecationexchangechromatography.Conditionsasdefinedinthelegend ofFig.7forthepHgradient,andasdefinedinRef.[24]forthegenericsaltgradient.

injectedamountanddetectionmodewereapplied.Theonlydif- ferencewasthemobilephaseconditionsandthemodeofelution.

Thegradientprogramsweredevelopedtokeepasimilarretention rangeforallthemAbs.ThehighestretentionfactorinpHgradient modewaskapp=13.7whileinsaltgradientmodeitwaskapp=14.2.

Therefore,theseconditionsallowedafaircomparisonofthetwo modes.

First, the retention factors were compared. The kapp values observedinthepHgradientmodewereplottedagainstthek_app values observed in the salt gradient mode. Fig. 8A shows the observedplotforthe10intactmAbsandtheirmainvariants.As shown,denosumab,palivizumab,bevacizumabandadalimumab showdeviationfromthelinearfittedfunction.Thissuggeststhat thetwoCEXelutionmodesoffersomealternativeselectivityfor theseparticularmAbs.ForthesixothermAbs,nosignificantdiffer- encesinelutionorder(selectivity)wasobserved.

Exceptselectivityandretention,thepeakcapacitiesachieved withthesetwomodeswerealsocompared.Fig.8Bshowsthepeak capacityobservedforsixrepresentativemAbs,whenperforming thegenericgradientseparations.Inagreementwithourexpecta- tions,thesaltgradientmodeoffershigher peakcapacity(when applyingthesamegradientsteepness,flowrateandcolumndimen- sion).Inaverage,a25%higherefficiencywasobservedinthesalt gradientmodecomparedtothepHgradientmode.Inthecaseof rituximab,a 2.5fold higherpeak capacity wasobservedin salt gradientmode.RituximabelutedinatailedbroadpeakinpHgra- dientmodewhileitwassymmetricalandsharpinthesaltgradient mode.Foradalimumab,thetwomodesprovidesimilarefficiency.

FormostofthemAbs,thevariantswerebetterresolvedinthesalt gradientmode.

ToconcludeonthesetwoCEXmodes,thetheoreticallyexpected focusingeffectofthepHgradientmodedoesnotbringanygainin resolvingpowerorpeakcapacityforintactmAbscomparedtothe saltgradientmode.Inaddition,thesaltgradientmodecanalsobe successfullyappliedasamultiproductgenericmethodforalarge varietyofintactmAbs[24].

4. Conclusion

Inthissecondpart,thepossibilitiesofferedbyCEXpHgradient modewereevaluatedfor10differentmodelmAbs,andsystemati- callycomparedtoCEXsaltgradientapproach.

Firstof all,wehaveevaluatedwhethertheretentioncanbe modelledasafunctionofgradientsteepnessandmobilephasetem- peratureinCEXpHgradientmode.Becausetheretentionmodels werealwayslinear,onlyfourinitialexperiments(2gradientstimes attwotemperatures)wererequiredtomodelthebehaviourinCEX pHgradient.Then,only6hwererequiredtofindouttheoptimal conditionsforthecharacterizationofseveralmAbsvariants.

Next,wealsodemonstratedthattheretentiontimesobservedin CEXpHgradientwerenotsystematicallyproportionaltothemAbs pI.Thus,theapplicationofthisapproachformeasuringaccuratepI valuesshouldbeconsideredwithcaution.

Finally,our10 modelmAbsweresuccessfullyeluted inboth CEXpHgradientand saltgradient,showingthat bothmodesof elutioncanbeconsideredasmulti-productchargesensitivesep- arationmethods.These twoapproaches werealsocompared in termsofselectivityandpeakcapacity.Selectivitywerequitecom- parable (exceptfor denosumab, palivizumab,bevacizumab and adalimumab) andfor mostofthemAbs,thevariants werebet- terresolvedinthesaltgradientmode.Thepeakcapacitieswere generallyimprovedwiththeregularsaltgradientapproach.These observationsconfirmthatpHgradientapproachmaybeoflower interestthansaltgradientCEXmodeformAbscharacterization.

Acknowledgments

WeacknowledgeElsaWagner-Rousset,LauraMorel-Chevillet and Olivier Colas (Physico-Chemistry Department, Centre d’Immunologie Pierre Fabre, Saint-Julien en Genevois, France) forpIcalculationandexperimentaldetermination(IEFandcIEF) and LC-MS analysis and Stephane Tonelotto (QC Department, Centre d’Immunologie Pierre Fabre, Saint-Julien en Genevois, France)forhelpfuldiscussionsonCEX.

References

[1]A.Beck,S.S.Cianferani,A.VanDorsselaer,Biosimilar,biobetter,andnextgener- ationantibodycharacterizationbymassspectrometry,Anal.Chem.84(2012) 4637–4646.

[2]Z.Zhang,H.Pan,X.Chen,Massspectrometryforstructuralcharacterizationof therapeuticantibodies,MassSpectrom.Rev.28(2009)147–176.

[3]M.Perkins,R.Theiler,S.Lunte,M.Jeschke,Determinationoftheoriginof chargeheterogeneityinamurinemonoclonalantibody,Pharm.Res.17(2000) 1110–1117.

[4]S.Fekete,A.L.Gassner,S.Rudaz,J.Schappler,D.Guillarme,Analyticalstrategies forthecharacterizationoftherapeuticmonoclonalantibodies,TrendsAnal.

Chem.42(2013)74–83.

[5]K.G. Moorhouse, W.Nashabeh,J. Deveney, N.S.Bjork,M.G.Mulkerrin, T.

Ryskamp,ValidationofanHPLCmethodfortheanalysisofthechargehet- erogeneityoftherecombinantmonoclonalantibodyIDEC-C2B8afterpapain digestion,J.Pharm.Biomed.Anal.16(1997)593.

[6]H.Lau,D.Pace,B.Yan,T.McGrath,S.Smallwood,K.Patel,J.Park,S.S.Park,R.F.

Latypov,InvestigationofdegradationprocessesinIgG1monoclonalantibodies bylimitedproteolysiscoupledwithweakcation-exchangeHPLC,J.Chromatogr.

B878(2010)868–876.

[7]G.Teshima,M.X.Li,R.Danishmand,C.Obi,R.To,C.Huang,J.Kung,V.Lahidji,J.

Freeberg,L.Thorner,M.Tomic,Separationofoxidizedvariantsofamonoclonal antibodybyanion-exchange,J.Chromatogr.A1218(2011)2091–2097.

[8]D.Farnan,G.T.Moreno,Multiproducthigh-resolutionmonoclonalantibody chargevariantseparationsbypHgradiention-exchangechromatography,Anal.

Chem.81(2009)8846–8857.

[9]L.A.Æ.Sluyterman,O.Elgersma,Chromatofocusing:isoelectricfocusingonion exchangecolumns.I.Generalprinciples,J.Chromatogr.150(1978)17–30.

[10]L.A.Æ.Sluyterman,J.Wijdenes,Chromatofocusing:isoelectricfocusingonion exchangecolumnsII. Experimentalverification,J.Chromatogr.150(1978) 31–44.

[11]L.A.Æ.Sluyterman,J.Wijdenes,Chromatofocusing:IV.Propertiesofanagarose polyethyleneimineionexchangeranditssuitabilityforproteinseparation,J.

Chromatogr.206(1981)441–447.

[12]A. Rozhkova, Quantitative analysis of monoclonal antibodies by cation- exchangechromatofocusing,J.Chromatogr.A1216(2009)5989–5994.

[13]X.Kang,D.Frey,High-performancecation-exchangechromatofocusingofpro- teins,J.Chromatogr.A991(2003)117–128.

[14]L.Shan,D.J.Anderson,Effectofbufferconcentrationongradientchromatofo- cusingperformanceseparatingproteinsonahigh-performanceDEAEcolumn, J.Chromatogr.A909(2001)191–205.

[15]L.Shan,D.J.Anderson,GradientchromatofocusingversatilepHgradientsepa- rationofproteinsinion-exchangeHPLC:characterizationstudies,Anal.Chem.

74(2002)5641–5649.

[16]M.Talebi,A.Nordbog,A.Gaspar,N.A.Lacher,Q.Wang,X.Z.He,P.R.Haddad, E.F.Hilder,Chargeheterogeneityprofilingofmonoclonalantibodiesusinglow ionicstrengthion-exchangechromatographyandwell-controlledpHgradients onmonolithiccolumns,J.Chromatogr.A1317(2013)148–154.

[17]T.Ahamed,B.K.Nfor,P.D.E.M.Verhaert,G.W.K.vanDedem,L.A.M.vanderWie- len,M.H.M.Eppink,E.J.A.X.vandeSandt,M.Ottens,pH-gradiention-exchange chromatography:ananalyticaltoolfordesignandoptimizationofproteinsep- arations,J.Chromatogr.A1164(2007)181–188.

[18]L.Shan,D.J.Anderson,Gradientchromatofocusing.versatilepHgradientsepa- rationofproteinsinion-exchangeHPLC:characterizationstudies,Anal.Chem.

74(2002)5641–5649.

[19]Y.Liu,D.J.Anderson,Gradientchromatofocusinghigh-performanceliquidchro- matography:I.Practicalaspects,J.Chromatogr.A762(1997)207–217.

[20]J.C.Rea,G.T.Moreno,Y.Lou,D.Farnan,ValidationofapHgradientbasedion- exchangechromatographymethodforhigh-resolutionmonoclonalantibody chargevariantseparations,J.Pharm.Biomed.Anal.54(2011)317–323.

[21]L.Zhang,T.Patapoff,D. Farnan,B.Zhang,ImprovingpH gradientcation- exchange chromatography of monoclonalantibodies by controlling ionic strength,J.Chromatogr.A1272(2013)56–64.

[22]S.Fekete,S.Rudaz,J.Fekete,D.Guillarme,Analysisofrecombinantmonoclonal antibodiesbyRPLC:towardagenericmethoddevelopmentapproach,J.Pharm.

Biomed.Anal.70(2012)158–168.

[23]S.Fekete,J.Fekete,I.Molnár,K.Ganzler,Rapidhighperformanceliquidchro- matographymethoddevelopmentwithhighpredictionaccuracy,using5cm longnarrowborecolumnspackedwithsub-2␮mparticlesandDesignSpace computermodeling,J.Chromatogr.A1216(2009)7816–7823.

[24]S.Fekete, A. Beck,J. Fekete, D. Guillarme, Methoddevelopment forthe separationofmonoclonalantibodychargevariantsincationexchangechro- matography,partI:saltgradientapproach,J.Pharm.Biomed.Anal.102C(2014) 33–44.

[25]D.Ayoub,W.Jabs,A.Resemann,W.Evers,C.Evans,L.Main,C.Baessmann,E.

Wagner-Rousset,D.Suckau,A.Beck,Correctprimarystructureassessmentand extensiveglyco-profilingofcetuximabbyacombinationofintact,middle-up, middle-downandbottom-upESIandMALDImassspectrometrytechniques, mAbs5(2013)699–710.

[26]T.Ishihara,S.Yamamoto,Optimizationofmonoclonalantibodypurification byion-exchangechromatography,applicationofsimplemethodswithlinear gradientelutionexperimentaldata,J.Chromatogr.A1069(2005)99–106.

[27]R.W.Stout,S.I.Sivakoff,R.D.Ricker,L.R.Snyder,Separationofproteinsbygradi- entelutionfromion-exchangecolumns:optimizingexperimentalconditions, J.Chromatogr.353(1986)439–463.

[28]M.A.Quarry,R.L.Grob,L.R.Snyder,Predictionofpreciseisocraticretentiondata fromtwoormoregradientelutionruns.Analysisofsomeassociatederrors, Anal.Chem.58(1986)907–917.

[29]C.M.Roth,K.K.Unger,A.M.Lenhoff,Mechanisticmodelofretentioninprotein ion-exchangechromatography,J.Chromatogr.A726(1996)45–56.

[30]S.Fekete,S.Rudaz,J.L.Veuthey,D.Guillarme,Impactofmobilephasetemper- atureonrecoveryandstabilityofmonoclonalantibodiesusingrecentreversed phasestationaryphases,J.Sep.Sci.35(2012)3113–3123.

[31]S.Heinisch,J.-L.Rocca,Senseandnon-senseofhigh-temperatureliquidchro- matography,J.Chromatogr.A1216(2009)642–658.

[32]Y.Chen,C.T.Mant,R.S.Hodges,Temperatureselectivityeffectsinreversed- phaseliquidchromatographyduetoconformationdifferencesbetweenhelical andnon-helicalpeptides,J.Chromatogr.A1010(2003)45–61.

[33]C.T.Mant,Y. Chen,R.S.Hodges, Temperatureprofilingofpolypeptidesin reversed-phaseliquidchromatography:I.Monitoringofdimerizationand unfoldingofamphipathic␣-helicalpeptides,J.Chromatogr.A1009(2003) 29–43.

[34]C.T.Mant,B.Tripet,R.S.Hodges,Temperatureprofilingofpolypeptidesin reversed-phaseliquidchromatography:II.Monitoringoffoldingandstability oftwo-stranded␣-helicalcoiled-coils,J.Chromatogr.A1009(2003)45–59.

[35]U.D.Neue,Theoryofpeakcapacityingradientelution,J.Chromatogr.A1079 (2005)153–161.