Positively charged molecules can be separated using CEX columns,typicallypackedwith3–10␮mparticlesandcontaining negativelychargedacidicfunctionalgroups.Thesecolumnsbind cationicspecies suchasprotonatedbasesthroughionic interac-tion.Inanion-exchange(AEX)mode,thestationaryphasecarries positivelychargedbasicfunctionalgroupsthatarecapableof bind-inganions(e.g.ionizedcarboxylicacids).Themobilephaseusually containsabuffertomaintainstablepHandvaryingthesalt con-centration(counter-ion)tocontroltheretentionofsampleions.

Thechargeofthecounter-ionhasthesamesignasthesampleions, thereforeitcanbeusedtocontroltheretentionofprotonatedbases inCEXorionizedacidsinAEX.

Thestrengthoftheinteractionisdeterminedbythenumber andlocationofthechargesontheanalyzedmoleculesandonthe functionalgroups.Byincreasingthesaltconcentration,thesamples withtheweakestionicinteractionsstarttoelutefromthecolumn first.Moleculeshavingastrongerionicinteractionrequireahigher saltconcentrationandelutelaterinthegradient.

InthepH-gradientmode,theionicstrengthofthemobilephase iskeptlowandconstant,whilethepHisvariedthankstoalinear gradient.

3.1. Theimpactofstationaryphase

Regardingthestationaryphase,therearetwomainaspects:(1) thestrengthofinteractionandassociatedretention(strongorweak ion-exchanger)andthe(2)achievablepeakwidths(efficiency).

Bothcation and anionexchangerscanbeclassifiedaseither weakorstrongexchangers.Weakcationexchangersarecomprised ofaweakacidthatgraduallylosesitschargeasthepHdecreases (e.g.carboxymethylgroups),whilestrongcationexchangersare comprisedofastrongacidthatisabletosustainitschargeovera widepHrange(e.g.sulfopropylgroups).Ontheotherhand,strong anion exchangers contain quaternary amine functional groups, whileweakanionexchangerpossessesdiethylaminoethane(DEAE) groups.Stronganionexchangersremainunderionizedforminthe pHrangebelow12,whilestrongcationexchangersareionizedat pH>2.

Asaruleofthumb,itispreferredtobeginthemethod develop-mentwithastrongexchangertoenableworkingoverabroadpH range.Strongexchangersarealsousefulifthemaximumresolution occursatanextremepH.(However,silicabasedion-exchangerscan beoperatedonlyinarestrictedpHrange.Incontrast,polymeric ion-exchangerscanbeusedinawidepHrange.)

Inthecaseofproteins,thecationexchangemodeiswellsuited, butastronganionexchangercanbeappliedtobindtheproteins iftheirpIisbelowpH7.Weakexchangerscanonlybeusefulina secondinstance,iftheselectivityofstrongionexchangersis unsat-isfactory.However,itisimportanttokeepinmindthattheion exchangecapacityofweakionexchangersvarieswithpH.

CommerciallyavailableIEXcolumnsarebasedonsilicaor poly-merparticles.Bothporousandnon-porousparticlesareavailable butforlargemoleculeswhichpossesslowdiffusivity,non-porous materialsareclearlypreferredtoavoidtheunwantedband broad-eningeffectsofthetransparticlemasstransferresistance(C-term of the van Deemter equation). Highly cross-linked non-porous poly(styrene–divinylbenzene) (PS/DVB) particles are most fre-quently used in protein separations due to their pH stability (2≤pH≤12).Thosematerialscannowwithstandpressuredrop ofuptoa500–600barandcanberoutinelyusedbeyond400bar.

Columnspackedwith10,5or3␮mnon-porousparticlesareoften used,butsub-2␮mmaterialsarealsoavailablesincerecently.On thosecolumnshighpeakcapacitycanbeattainedevenwithlarge biomolecules.Howeversomelimitationscanbeexpectedinterms ofloadingcapacityandretentionwhenapplyingthesenon-porous materials.Table1summarizesthemostpopularstate-of-the-art IEXcolumnsappliedfortheseparationofproteinchargevariants.

3.2. Theimpactofmobilephasecomposition

Inthesalt-gradientmode,themobilephasebufferpHmustbe betweenthepIofthechargedmolecule(e.g.therapeuticprotein) andthepKaofthechargedfunctionalgroupatthesurfaceofthe stationaryphase.InCEX,usingastrongcationexchangerwitha pKaof1.2,amoleculewithapI∼8(e.g.mAbs)maybeelutedwith amobilephasepHbufferof∼6.InAEX,amoleculewithapI∼6 mayberunwithamobilephasebufferatpH8whenthepKaofthe solidsupportisbeyond10.

IntheCEXmode,increasingthemobilephase bufferpHwill cause the molecule to become less protonated (less positively charged). Therefore,the proteinforms weaker ionicinteraction withthenegativelychargedstationaryphasegroups,whichresults inaretentiondecrease.Onthecontrary,decreasingthepH mani-festsinhigherretention.InAEXmode,–oppositely–decreasingthe mobilephasepHcausesthemoleculetobecomemoreprotonated (morepositivelyandlessnegativelycharged),thereforeadecrease inretentionisexpected.

ThemostoftenappliedpHrangeforproteinsIEXseparations isbetween5.5 and 7.0,however insomecaseslow pHaround 3.5isrequiredtoreachappropriateselectivityandretention.The most frequentlyused buffers for protein separations are 2-(N-morpholino)ethanesulfonicacid(MES),phosphateandcitrate.MES isusefulbetweenpH5.5and6.8(pKa∼6.15),phosphateisapplied forpHbetween6.7and7.6(pKa∼7.2),whilecitrateprovideshigh buffercapacityforpHbetween2.6and3.7(pKa∼3.1).Other addi-tivessuchasmalonicacid,aceticacidorformicacidhavealsobeen reportedforalimitednumberofapplications.Thebuffer concen-trationistypicallycomprisedbetween10and50mMandallowsa sufficientbuffercapacity.

AfterselectingthemobilephasepHandbuffer,thesalt-gradient hastobeoptimized.Typicallysodium-orpotassium-chlorideare usedforproteinscharacterization,usingasaltgradientfrom0to 0.2–0.5M.Theproteinsamplesareinjectedontothecolumnunder conditionswhereitissufficientlyretained.Then,agradientof lin-earlyincreasingsaltconcentrationisappliedtoelutethesample componentsfromthecolumn.Itisfinallyimportanttonoticethat thegradientsteepnesshasastrongimpactonretentionand selec-tivityandshouldthereforebesystematicallyoptimized.

InthepH-gradientmode,themaindifficultyistoperformlinear androbustpHgradients.Theuseofamixtureofaminebuffering

Table1

Propertiesofthemostpopularstate-of-the-artIEXcolumnsavailableforproteinseparations.

Columnname Chemistry Particle

BioMab(Agilent) Weakcationexchange (carboxylate) Antibodix(Supelco,Sepax) Weakcationexchange

(carboxylate)

MAbPacSCX-10(Thermo) Strongcationexchange (sulfonicacid) Poly(PolyLC) CATA Weakcationexchange

(polyasparticacid) 5 Ambient Information

speciesinthehigh-pHrangeandamixtureofweakacidsinthe low-pHrangeisquitecommon[38,39,42].Aspreviouslydiscussed, themostoftenusedbuffercomponentsareTrisbase,piperazine, imidazole,triethylamine, diethylamineand ammonium hydrox-ide[8,40,43].Finally,a0.25mM/minsodium-chloridegradientwas successfullyperformedconcomitantlywithapH-gradientforthe characterizationofmAbspossessingisoelectricpoints(pI)between 6.2and9.4,tohighlighttheinterestofpHgradientseparationover saltgradients[43].

3.3. TheimpactofsamplepI

Thechargeofproteinsdependsonthenumberandtypeof ion-izableaminoacidgroups.Lysine,arginineandhistidineresidues haveapositivelychargedsidechaingroupwhenionized,whereas glutamicacidand asparticacidresiduesare negativelycharged whenionized.EachionizablesidechaingroupshasitsownpKa. Therefore,theoverallnumberofchargesonaparticularproteinat agivenpHdependsonthenumberandtypeofionizableamino acidgroups.ProteinstendtohavedifferentchargesatagivenpH and socan befractionatedonthebasisof theirnetand acces-siblecharges.EachproteinhasapIvalue,which correspondsto

thepHvaluewhereithasnonetcharge.Then,whenpHisequal topI,theproteinwillnotbindtotheion-exchangeresin.Below this pH value,the protein hasa net positive charge and binds toacationexchanger,whileabovethispH,ithasanetnegative chargeandbindstoananionexchanger.Inpractice,proteinsare stable and functionallyactive withina fairlynarrow pHrange, sothechoiceofion exchanger isoftendictatedby thepH sta-bilityofthedesiredprotein.IftheproteinisstableatpHvalues belowitspI,acationexchangershouldbeusedifitisstableat pHvaluesaboveitspI,thenananionexchangerphasehastobe chosen.

ThepIoftheproteinalsodeterminesthemobilephasepH.ThepI oftherapeuticproteinsdistributebetween3.6and11.0,andamong them,mAbspossesspIvaluesbetween6and10.Forsalt-gradient basedCEXmode,themobilephasepHshouldpreferablybeatleast 1–2unitsbelowthepIofthesample,tomaintainappropriate reten-tion.InAEXmode–oppositely–thepHhastobesetatleast1–2 unitsabovethepIoftheprotein.

InthepHgradientmode–performedonCEXcolumns–the startingpHshouldbebelow thepIof theless retainedprotein, whilethefinalpHhastobesomewhathigherthanthepIofthe mostretainedprotein.

3.4. Theimpactoftemperature

Theeffectoftemperatureonretentionfactor(k)isgenerally expressedinliquidchromatographywiththeGibbsfreeenergyor van’tHoffequation:

log k=−H RT +S

R +logˇ (5)

where H representstheenthalpy changeassociated withthe transfer of the solute between phases, S the corresponding entropychange,Rthemolargasconstant,Ttheabsolute temper-atureandˇthephaseratioofthecolumn.Whenlog(k)isplotted against1/T,theenthalpyisgivenbytheslopeofthecurve.With reg-ularcompounds,theseplotsgenerallyfollowalinearrelationship.

However,non-lineardependenceoflog(k)versus1/Toverawide rangeoftemperaturewasnoticedbydifferentauthorsusing silica-basedaswellasnonsilica-basedstationaryphases[45].Theeffect oftemperatureontheretentionofpartiallyionizedcompounds whichmayexistintwoforms(i.e.molecularandionizedforms)can alsobedescribedwithEq.(5).However,bothenthalpyandentropy areexpectedtobedifferentforthetwoformsandasaresult,both HandScanvarywithtemperaturewhenbothformsarepresent toa significantextent[45].Withlargebiomolecules, theeffect oftemperatureonretentionbecomesmorecomplex.Depending onthestabilityofthesecondarystructure,themoleculesunfold tovariousextents andhenceinteractwiththestationaryphase withvariousstrengths[46]. Duetothedifferent conformation-dependentresponses of proteinsat elevated temperatures, the changeinretentioncanbedifficulttoassess[47,48].InRPLC sepa-rationofproteins,temperatureisausefulparameterforadjusting selectivity.InIEXseparationsofproteins,theimpactof tempera-turewasfoundtobeespeciallyimportantforpeakcapacity(and thereforeforresolution),buthasa limitedimpactonselectivity [32,41].Itseemedthatinbothsalt-andpH-gradientbased sepa-rations,thetemperaturedoesnotmodifyseverelyselectivity,but impacttheachievablepeakcapacity.Therefore,insomecases, tem-peratureoptimizationcouldalsobeofimportanceduringtheIEX methoddevelopmentprocedure.

3.5. Optimizationprocedure

In contrastwithRPLC, the methoddevelopment inIEX was mostly basedon trialand erroror “one factor at time”(OFAT) approaches.However,therearesomeguidelinesavailablefrom col-umnproviders,whichexplainthebasicrulesformethodscreening (e.g.columnselection,bufferselection...).

Baietal.showedthedependenceofretentionandselectivityof IgGantibodiesonmobilephasepH,stationaryphasetypeand salt-gradientsteepnessinCEXmode[49].Theystudiedtheeffectof thethreevariablesindependently,andfoundthatmobilephasepH wasthemostimportantparameterinCEXseparationsofproteins.

Ithadthebiggestimpactontheseparationandthereforeshould bedeterminedfirst[49].Itwasalsofoundthat(i)peakwidthof IgG-smostlydependsonthetypeofthestationaryphaseand(ii) resolutioncanbetunedbychangingthegradientsteepness.Fig.3 showstheimpactofsalt-gradientsteepnessontheseparationof IgGproteins.

Themobilephaselinearvelocityalsohasastronginfluenceon theseparationqualityoflargeproteins[50,51].Indeed,the lon-gitudinaldiffusionisnegligiblewithlargemolecules,whileband broadeningismostlydeterminedbythemasstransferresistance.

Therefore,lowflowrateisalwayspreferredforhighresolution sep-arations,butacompromisehastobefoundbetweenresolutionand analysistime.

Fig.3. Theeffectofsalt-gradientslopeontheretention,selectivityandpeakwidth inCEXseparation.Sample:IgG1,mobilephase:40mMphosphatepH6.5,applying a0–0.4MNaClgradient.

AdaptedfromRef[49],withpermission.

Theinfluenceofsalttypecanalsobeimportant.Itseffecton theretentionofbovineserumalbuminwasreportedbyAl-Jibbouri [52].

Computerassistedmethoddevelopmentandoptimizationin RPLCproteinseparationsisquitecommon[53,54]andwasalso recently appliedin ion-exchange mode. Because of the system non-linearity,findingtheoptimumforprocessoptimizationis chal-lenging[55].Thiemoetal.developedasoftwarecalledChromX fortheestimationofparameters,chromatogramsimulation,and processoptimization[55].ChromXprovidesnumericaltoolsfor solving varioustypes of chromatography models,includingthe modelcombinationofTransportDispersiveModel(TDM)andSMA.

SimilarlytoRPLCmethoddevelopment,anon-LSSand LSStype computerassistedmethoddevelopmentprocedurewasrecently reportedforbothsalt-andpH-gradientmodesinagreementwith QualitybyDesign(QbD)concept[32,41].

Forthesalt-gradientbasedproteinseparation,itseemedthat temperaturewasnotarelevantparameterfortuningselectivityand shouldbekeptat30C,toachievehighresolvingpower(elevated peakcapacity) [32].Becausetherelationship betweenapparent retentionfactorsandgradienttime(slope)canbedescribedwitha linearfunction,onlytwoinitialgradientrunsofdifferentslopes are required for optimizing the salt gradient program. For pH dependence,asecondorderpolynomialmodel(i.e.basedonthree initialruns)ispreferredtodescribekversuspHdependence.When combiningtheexperimentsina designofexperiments(DoE),it appearedthatmethodoptimizationcanbeperformedrapidly,in anautomatedwaythankstoaHPLCmodelingsoftware,usingtwo gradienttimesandthreemobilephasepH(e.g.10and30min gra-dientona100mmlongstandardborecolumnatpH=5.6,6.0and 6.4).Suchaprocedurecanbeappliedroutinelyandthetimespent formethoddevelopmentwouldbeonlyaround9h.Therelative errorinretentiontimepredictionwaslowerthan1%,makingthis approach highlyaccurate[32]. Fig.4Ashows a generic DoEfor

Fig.4. Designofexperiments(A)andresolutionmap(B)fortheoptimizationofsalt-gradientbasedCEXseparationofmAbs(tg–pHmodel).Column:YMCBioProSP-F (100mm×4.6mm).Mobilephase“A”10mMMES,“B”10mMMES+1MNaCl.Flowrate:0.6mL/min,gradient:0–20%B,temperature:30C.Gradienttimes:tg1=10min, tg2=30min,pH1=5.6,pH2=6.0,pH3=6.4.Ontheresolutionmap,red-orangecolorsshowthehighestresolutionwhilethedark-blueareasindicatetheco-elutionofpeaks.

(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.) AdaptedfromRef[31],withpermission.

themethoddevelopmentofsalt-gradientbasedCEXseparationof mAbs(possessingawiderangeofpIbetween6.7and9.1),while Fig.4Brepresentstheobtainedresolutionmapshowingthe criti-calresolutionofthepeakstobeseparatedasafunctionofmethod parameters.

InthepH-gradientmode,itwasfoundthattheretentionoflarge proteinscanbeaccuratelymodeledasafunctionofgradient steep-nessand mobilephasetemperature[41].Becausetheretention modelswerealwayslinear,onlyfourinitialexperiments(2 gradi-entstimesat2temperatures)wererequiredtomodelthebehavior inCEXpH-gradient.Then,only∼6hwererequiredtofindoutthe optimalconditionsona100×4.6mmcolumn[41].

In document Ion-exchange chromatography for the characterization ofbiopharmaceuticals Journal of Pharmaceutical and Biomedical Analysis (Pldal 4-7)