Evaluation of recent Protein A stationary phase innovations for capture of biotherapeutics

We describe a comprehensive evaluation of 12 Protein A stationary phases for capture of biotherapeutics. We first examine the morphological properties of the stationary phases using a variety of orthogonal techniques including electron microscopy, particle sizing, pressure-flow behavior, and isocratic pulse response.

jou rn al h om ep a g e : w w w e l s e v i e r c o m / l o c a t e / c h r o m a

MedImmune, Purification Process Sciences, One MedImmune Way, Gaithersburg, MD 20878, USA

a r t i c l e i n f o

Article history:

Received 21 December 2017

Received in revised form 26 March 2018

Accepted 29 March 2018

Available online 7 April 2018

Keywords:

Protein A

Antibody

Affinity chromatography

Dynamic binding capacity

Bioprocessing

a b s t r a c t

1 Introduction

Monoclonalantibodies(mAbs)continuetobethemost

preva-lentclassofapprovedbiotherapeutics[1 Inaddition,theuseof

Fc-fusionproteinsandmAb-likemolecules,suchasbispecific

anti-bodiesandantibody-drugconjugates(ADCs),continuestoincrease

[2–4].SincethefirstmAbwasapprovedinthe1980s,ProteinA

hasbecomethemostwidelyusedcapturestepforFc-containing

moleculesduetoitshighlyspecificnature,easeofuse,andstrong

regulatorytrackrecord.Forthesepurificationprocesses,ProteinA

chromatographyisroutinelyutilizedaspartofaplatformapproach

whereitisplacedfirstinthepurificationtraintocaptureproduct

fromclarifiedcellculturebroth[5–13].Thisconfigurationallows

forrobustprocessingofsimilarmolecules.Evenwiththese

advan-tages,therehavebeeneffortstoidentifyalternativestoProtein

∗ Corresponding author.

E-mail address: pabstt@medimmune.com (T.M Pabst).

A, such as cation exchange or multimodal capture chromatog-raphy, toovercome the burden of high stationary phase costs [14–19];however,thesetechniquesmaynotbeasselectiveand maylacktheabilitytobeemployedaspartofaplatformapproach Therehasalsobeeninterestinnon-chromatographictechniques, suchasprecipitation[20–22]andaqueoustwo-phaseextraction [23–25],butthesestrategieshavenotgainedwidespreadusefor industrial-scalebioprocessing.Thus,itseemsunlikelythatProtein

Achromatographywillbesupersededasthedominantplatform approach for antibodyand Fc-fusionprotein purificationin the foreseeablefuture,anditsuseislikelytocontinuetoincreasewith thegrowthinthemarket[26]

StaphylococcalProteinAisa42kDasinglechainpolypeptide locatedontheoutersurfaceofStaphylococcusaureus[27–30].Early Protein A affinity stationaryphases consisted of native Protein

A coupledtoabase matrixmostoften throughcovalent bond-ing to amines Since then, dramatic improvements have been madeinProteinAchromatographystationaryphases,mostnotably increasedstabilityunderalkalineconditionsrealizedthroughpoint

https://doi.org/10.1016/j.chroma.2018.03.060

0021-9673/© 2018 MedImmune Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.

c interstitialproteinconcentration(mg/mL)

C proteinconcentrationinsolution(mg/mL)

CF columnproteinfeedconcentration(mg/mL)

cp proteinconcentrationintheporefluid(mg/mL)

CCCB clarifiedcellculturebroth

dp volumemeanparticlediameter(␮m)

Dax axialdispersioncoefficient(cm2/s)

De effectiveporediffusivity,=εpDp(cm2/s)

Dp porediffusivity(cm2/s)

D0 freesolutiondiffusivity(cm2/s)

DBC dynamicbindingcapacity(mg/mL)

EBC equilibriumbindingcapacity(mg/mL)

K adsorptionconstant in Langmuirisothermmodel

(mL/mg)

kf filmmasstransfercoefficient(cm/s)

KD distributioncoefficient,=

VR/CV−εb



/ (1−εb)

M dextranmolecularweight(Da)

q stationaryphaseproteinconcentration(mg/mL)

qd percentoftotalinahistogramparticlesizebin(%)

qm maximumbindingcapacityinLangmuirisotherm

model(mg/mL)

r radialcoordinate(cm)

rp volumemeanparticleradius(␮m)

rpore poreradius(nm)

rs dextranhydrodynamicradius(nm)

S standarderrorofregression(unitsoftheresponse

variable)

u superficialmobilephasevelocity(cm/s)

v interstitialmobilephasevelocity(cm/s)

VR retentionvolume(mL)

V10% volumeatwhich10%breakthroughoccurs(mL)

z axialcoordinate(cm)

Dimensionlessnumbers

NBi Biotnumber,=kfrp/De

NRe Reynoldsnumber,=dpu/εb

NSc Schmidtnumber,=/D0

Greeksymbols

P pressuredropinapackedbed(MPa)

εb extraparticleporosity (interstitialcolumnvolume

fraction)

εp,x intraparticleporosityforsolutex

εT totalcolumnporosity,=εb+ (1−εb)εp

 dynamicviscosity(cP)

mutationsintheBandCdomainsofProteinA[31].Manyofthe

commerciallyavailableProteinAstationaryphases nowconsist

ofengineeredligandswithrepeatunitsderived fromtheBorC

domain,and canwithstandmany cyclesof exposuretosodium

hydroxideconcentrationsatorgreaterthan0.1N

Inadditiontotheincreaseinalkalinestability,therehasbeen

arobustdemandtoincreasethebindingcapacityofProteinA

sta-tionaryphasesinresponsetocontinuouslyincreasingcellculture

titers,whichnowroutinelyexceed5g/L.Processmodelsdeveloped

topredictfacilitycapacityandcostofgoodshavebeenshownto

besensitivetoProteinAcapturecolumndynamicbinding

capac-ity(DBC)[32,33].Moreover,theimportanceofincreasedbinding

capacity,bothequilibriumanddynamiccapacity,isnotlimitedto batchprocessingassimilarincreasesinproductivityareexpected

incontinuousmulti-columnprocesssettings.Inresponsetothese marketpressures,numerousProteinAstationaryphaseshavebeen introducedoverthelastseveralyearstoachieveeverhigherDBCs andincreasedproductivity.Insomecases,anincreaseinDBCis real-izedbyimprovingstaticbindingcapacitythroughmodificationof theligand[34,35]orincreasingtheliganddensity[7,36].Inother cases,DBCcanbeincreasedbyreducingmasstransferresistance suchthattheavailablestaticcapacityisutilizedmoreefficiently [37–40]

In this work, we examine recent innovations in Protein A stationaryphasessuitableforindustrial-scalecaptureof biophar-maceuticals.Theworkincludescomprehensivecharacterizationof thestationaryphasemorphologicalproperties,aswellasbinding andelutionbehaviorforapanelofantibodiesandFc-fusion pro-teins.Inaddition,wequantitativelydescribeproteinmasstransfer usingthegeneralratemodeltodetermineeffectiveporediffusivity Finally,theProteinAstationaryphasesweretestedwithclarified cellculturebrothtoassessprocessperformanceandproduct qual-ityunderrealisticbioprocesssettings

2 Materials and Methods

2.1 Bufferreagentsandproteinpreparations Chemicalsusedforbufferpreparation anddextranstandards were obtained from Sigma (St Louis, MO, USA) and JT Baker (Phillipsburg, NJ, USA) Antibodies and Fc-fusion proteinswere expressedinChinesehamsterovary(CHO)cellsusingstandardcell culturetechniques.Togeneratepurifiedmaterial,clarifiedcell cul-turebrothwaspurifiedbyProteinAchromatographyandthenby ionexchangechromatography.Table1summarizestheantibodies andFc-fusionproteinsusedinthiswork

2.2 ProteinAstationaryphases Protein A stationary phases used in this work are summa-rizedinTable2alongwithpubliclyavailabledataobtainedfrom manufacturerliterature.Allstationaryphaseswerecommercially available,exceptforMabSelectPrismA,whichwasobtainedasa pre-commerciallaunchsamplefromGEHealthcare(Marlborough,

MA,USA).TheProteinAstationaryphaseswereflowpackedin 1.1cmdiameterVantageL11columnsfromMillipore(Billerica,MA, USA)toa10cmbedheightandcompressionfactorsof∼1.2were achievedforallresins.Packedbedqualitywasevaluatedby calcu-latingasymmetryfactorsandreducedHETPfrompulsesofsodium chlorideinTrisbufferatpH7.4.Valuesforasymmetryfactorranged from1.0–2.2andreducedHETPvaluesrangedfrom3.6–9.2,which areconsistentwiththereducedHETPvaluesrecentlyreportedfor lab-scalecolumns[41]

2.3 Proteinconcentrationdetermination Proteinconcentrations of purified samplesweredetermined using a Nanodrop 2000c from Thermo Scientific (Wilmington,

DE, USA) with themicrovolume pedestal and measurement at

a wavelength of 280nm Concentrations of antibodies and Fc-fusion proteins in clarified cell culture brothwere determined

byanalyticalProteinAhigh-performanceliquidchromatography (ProA-HPLC)usingaPOROSA20(4.6mmID×10cm,20␮m) col-umnobtainedfromThermoFisher(GrandIsland,NY,USA)with

anAgilent1200HPLCsystem(PaloAlto,CAUSA).TheHPLC sys-temwasoperatedat3.5mL/minwithbindingandelutionmobiles phases consistingof phosphate bufferedsaline (PBS) atpH 7.2 and pH 2.2, respectively Samples were applied tothe column

Table 1

Protein properties.

Molecule Class pI a MW (kDa) b rh(nm) c D0(x10−7cm 2 /s) d

a Measured by capillary isoelectric focusing.

b Measured by mass spectrometry including contribution from glycosylation.

c Calculated by Stokes-Einstein equation with D0measured by dynamic light scattering at infinite dilution.

d Measured by dynamic light scattering at infinite dilution.

Table 2

Protein A stationary phase properties as reported by manufacturers.

Name Abbreviation Manufacturer Base matrix Protein A ligand progenitor domain

(number of repeat units)

d p (␮m) Alkaline stabilized

MabSelect SuRe MSS GE Healthcare Agarose B (4) 85 Yes

MabSelect SuRe LX LX GE Healthcare Agarose B (4) 85 Yes

MabSelect SuRe pcc PCC GE Healthcare Agarose B (4) 50 Yes

MabSelect PrismA PrismA GE Healthcare Agarose B (6) ∼60 Yes

Amsphere A3 A3 JSR Life Sciences polymethacrylate – a 50 Yes

AF-rProtein A HC-650F 650F Tosoh Biosciences polymethacrylate C (6) 45 Yes

a Not disclosed by the manufacturer.

neatandtheelutionprofilewasmonitoredat280nmusingthe

systemspectrophotometer Elutionpeak areawasconverted to

proteinconcentrationusingastandardcurvegeneratedwith

puri-fiedmaterial

2.4 AggregatedeterminationbysizeexclusionHPLC

Aggregatelevelsinpurifiedproteinsamplesweredetermined

by analytical high-performance size exclusion chromatography

(SEC-HPLC)usingaTSKgelG3000SWXL(7.8mmID×30cm,5␮m)

columnobtainedfromTosohBiosciences(KingofPrussia,PA,USA)

withanAgilent1260HPLCsystem.TheHPLCsystemwas

oper-atedat1mL/minwithamobilephaseconsistingof100mMsodium

phosphate,200mMsodiumsulfate,pH6.8.Themobilephaseused

toanalyzemAb1samplesalsoincluded10%isopropanol.Samples

(250␮g)wereappliedtothecolumnneatandtheelutionprofile

wasmonitoredat280nmusingthesystemspectrophotometer

Aggregatelevelsweredeterminedasaratioofpeakareasofthe

early-elutingaggregatepeak(s)andthemonomerpeak

2.5 DynamicLightScattering

Dynamiclightscattering(DLS)measurementsweremadewith

aDynaProPlateReaderIIfromWyattTechnology(SantaBarbara,

CA,USA)withpurified proteinsamplespreparedat 2–10g/L in

25mM Tris, 150mM NaCl,pH 7.4 35␮Lprotein sampleswere

measuredin triplicate in 384well plates at 20◦C Diffusivities

wereobtainedfrom theDynamics software (version7.4.0) and

hydrodynamic radii were calculated within the software using

theStokes-Einsteinequation.Averaged(n=3)valueswereplotted

versusconcentrationandextrapolatedtoobtaindiffusivityvalues

andhydrodynamicradiiatinfinitedilution

2.6 Electronmicroscopy Transmissionelectronmicroscopy(TEM)ofstationaryphases wasperformed at Charles RiverPathology Associates (Durham,

NC, USA) Stationary phases in their shipping solutions (∼20% ethanol)weredehydratedwithgradedstepsto100%ethanol, sol-ventexchangedinto100%acetone,andthenembeddedinSpurr’s resintogeneratesolidsampleblocks.Thinsections(∼90nm)were cutusingadiamondknife,stainedwithmethanolicuranylacetate andReynold’sleadcitrate,andthenexaminedwithatransmission electronmicroscopefromJEOL(Peabody,MA,USA;model JEM-1011).TEMwasperformedatmagnificationsbetween800–6000x andhigh-resolutionimageswerecapturedwithanAMTXR16M digitalcamera

2.7 Particlesizemeasurements Volumemeandiametersweredeterminedbylightscattering using thePartica laser scattering particlesizedistribution ana-lyzer(modelLA-950V2)fromHoribaInstruments(WestChicago,

IL,USA).Stationaryphasesampleswerepreparedat15%slurries

in50mMTris,150mMNaCl,pH7.4andthen1mLofthesample slurrywasaddedtothedetectionchamberandmeasuredatroom temperature.Horibasoftwarereportedthevolumemeandiameter, thestandarddeviationoftheparticlesizedistribution,andthe dis-tributionhistogram.Samplesweremeasuredintriplicateandthe volumemeandiameterwasdeterminedfromtheaverageofthree samplemeasurements

2.8 Pressure-flowcurves Pressuredropacrosspackedbedswasmeasuredatlinear super-ficialvelocities between0–400cm/husing a digital manometer

(model3462)purchased fromTraceableProducts (Webster,TX,

USA).ColumninletswereconnectedtoanAKTAPure25(GE

Health-care,Marlborough,MAUSA)systempumpandthecolumnoutlet

leftopentoatmosphericpressure.Thedigitalmanometerwas

con-nectedtotheinletandoutlettubingwithT-connections.Pressure

dropmeasured in anempty column withthe adapterspushed

togetherwassubtractedfromthepackedbedmeasurements

2.9 Particleporositymeasurements

Particleporosities,εp,wereobtainedfromisocraticpulses

injec-tionsofmAb3,bsAb1,andsodiumchloridecarriedoutat150cm/h

usinganAKTAPure25.Forproteinpulses,themobilephasewas

50mMacetate,100mMNaCl,pH3.0forallproteinandstationary

phasecombinationsexceptproteinsinjectedontotheToyopearl

AF-rProteinAHC-650F column, which used 100mM acetic acid

topreventinteractionswiththecolumnatpH3.0.Protein

sam-ples (final concentration ∼2.5mg/mL) were injected on to the

columnsusinga100␮Lsampleloop.Forsodiumchloride

injec-tions,the mobilephase was50mM Tris,150mMNaCl, pH7.4,

and100␮Lsamplesof50mMTris,1MNaCl,pH7.4wereinjected

usingasampleloop.PeakretentionwasmonitoredusingtheAKTA

instrumentation(absorbanceforproteins,conductivityforsodium

chloride)andthedatawasexportedtoExcelsoftwareto

deter-minethepeakretentionvolumeusingthefirststatisticalmoment

Thesystem volume, determinedin an empty column withthe

adapterspushedtogether,wassubtractedfromthevaluesobtained

inpackedbeds

2.10 Inversesizeexclusionchromatographyusingdextranpulse

injections

Dextraninjectionswerecarriedoutat1mL/minusingan

Agi-lent1260HPLCand thepeakretention wasmonitoredwithan

Agilentrefractiveindex(RI)detector(ModelG1362A).Themobile

phase(50mMTris,pH7.4)wascontrolledto26◦Cwitha

recir-culating water bath Temperature of the autoinjector, column

block,andRIdetectorwerecontrolledat26◦CusingChemstation

software.Dextransamples,purchasedfromSigma-Aldrich(with

nominalmolecularweights of 10kDa (catalog#D9260),40kDa

(31389),70kDa(31390),270kDa(00894),410kDa(00895),and

670kDa (00896), and dextrose, purchased from Thermo Fisher

Scientific(catalog#D16-1),weredissolvedinthemobilephase

buffer at a final concentration of 5mg/mL and injected neat

(10␮L) on to the ∼10mL Protein A columns (packed in

Van-tage L11 columnsas described previously).Data wasexported

toExcel software todetermine thepeak retention volume,VR,

usingthefirststatisticalmoment,andthedistributioncoefficient,

KD=

VR/CV−εb



/ (1−εb).Peakswerenormalizedbypeakarea forplotting.Thesystemvolume,determinedinanemptycolumn

withtheadapterspushedtogether,wassubtractedfromthevalues

obtainedinpackedbeds

2.11 Adsorptionisotherms

Equilibrium adsorption isotherms were constructed from

1.25mLbatchbindingexperiments.Proteinstocksolutionswere

preparedat∼5mg/mLinequilibrationbuffer(50mMTris,150mM

NaCl,pH7.4)and dilutedtoknownconcentrations with

equili-brationbufferin1.5mLEppendorftubes.Stationaryphaseslurries

werepreparedat∼15%inequilibrationbuffer,addedtothediluted

protein samples, and allowed to gently mix on a rotator for

24–28hours.Afterequilibration,theEppendorftubeswere

cen-trifugedbriefly to pelletthe stationaryphases and the protein

concentrationsintheliquidphasesweremeasured.Amass

bal-ancewasusedtodeterminetheamountofproteinthatwasbound

tothestationaryphaseatequilibrium

2.12 Breakthroughbehaviorandequilibriumbindingcapacity measurements

BreakthroughbehaviorwasdeterminedusingtheAKTAPure

25ata residencetimeof 4minuteswitha ∼10mLpackedbed Purifiedprotein loadwasadjusted topH7.4±0.2 and conduc-tivity wasadjustedto15±2 mS/cmwithsodiumchloride.The columnwasequilibratedwith50mMTris,150mMNaCl,pH7.4 andthenloadedwithprotein(feedconcentration,CF∼5mg/mL) untilthecolumnwassaturated(i.e.theoutletconcentrationwas

∼99%ofthefeedconcentration).Theexactproteinconcentration

ofthefeedwasdeterminedbyofflineA280measurementusing theNanodrop2000c.Bedexhaustionoccurredwithaproteinload

of150–225mg/mLcolumnvolume,dependingontheproteinand stationaryphase.Instantaneousconcentrationatthecolumn out-letwasmeasuredonlineat280nmusingtheAKTAUVmeter.The

UVabsorbancewasmeasuredwiththecolumninbypassto deter-minethemaximumabsorbanceoftheproteinfeedsolution.To testforlinearity,theAKTAUVmeterwascalibratedwithmAb1, whichhadthegreatestUVabsorbanceoftheproteinsevaluatedin thisstudy.Ahighlylinearcalibrationcurvewasobtainedoverthe proteinconcentrationrangeusedforbreakthroughanddynamic bindingcapacityexperiments

The equilibrium binding capacity (EBC, in mg/mL of solid support)wasdeterminedfromafullbreakthroughcurveby numer-ically integrating theareaabove thecurve andbelow the feed concentrationaccordingto:

EBC=

 0

∞

(CF−C)dV−εTCFCV

WhereεT=εb+ (1−εb)εP isthetotalporosity(outsideof parti-cleandinsidepores),εbistheextraparticleporosity,andCVisthe columnvolume(orpackedbedvolume).Inthisequation,the inte-gralinthenumeratorrepresentsthemassofproteinaccumulated

inthecolumnduringtheentirebreakthroughexperimentandthe secondterminthenumeratorrepresentsthemassofproteinthat remainsunboundintheliquidwithinthecolumn;takentogether theyrepresentthemassofproteinboundtothesolidphaseduring theexperiment.Thedenominatorrepresentstheamountof sta-tionaryphase(inunitsofmLofsolidsupport).Thesystemholdup volume(determinedbyproteinpulsesinemptycolumnswiththe adapterspushedtogether)wasalsosubtractedfromthevolume loaded.EBCswereconverted tomg/mLofparticleormg/mL of packedbedusingεpandεb,asappropriate

2.13 Dynamicbindingcapacitymeasurements Dynamicbindingcapacity(DBC),definedastheamountof pro-teinloadedat10%breakthrough,wasdeterminedforstationary phasespackedin∼10mLcolumnsusinganAKTAPure25ina man-nersimilartofullbreakthrough curvesasdescribed above.The columnwasequilibratedwith50mMTris,150mMNaCl,pH7.4 andthenloadedwithpHandconductivityadjustedpurifiedprotein

atCF∼5mg/mLuntil>10%breakthroughwasobserved.DBCwas determinedatresidencetimesof2.4,4,and6minutesaccording to:

DBC= (V10%−εTCV)CF

whereV10%isthevolumeatwhich10%breakthroughoccurs

AscanbeseenfromEq.(2),theunboundproteinthatremainsin theliquidinsidethecolumnissubtractedfromtheboundprotein forpurposesofcalculatingDBC,whereastheamountofproteinthat

Table 3

Summary of Protein A stationary phase morphological properties.

Name Particle size a (␮m) rporeb (nm) Extraparticle

porosity, εbc

Intraparticle porosity

a As measured by laser light scattering dpis the volume mean diameter; SD is the standard deviation of the frequency distribution.

b Determined by fit of Eq (4) to isocratic elution data obtained with dextran probes ranging from 10–670 kDa.

c Determined from fit of Eq (3) to pressure drop data in a packed bed.

d Determined from injections of mAb or bsAb under non-binding (acidic) conditions described in Section 2.9.

e Determined from NaCl pulse injections.

breaksthroughthecolumn(upto10%breakthrough)isincluded

intheDBCvalue.Amorerigorouscalculation,likeEq.(1)where

thelimitsofintegrationgofrom0toV10%,couldbeimplemented

toaccountforproteinthatbreaksthroughuptoV10%,butthis

con-tributionisnegligibleandthuswasnotaccountedforintheDBC

values.ThetotalporositydeterminedbymAb3injectionswasused

inDBCcalculationsformAb1-3,bsAb1,and Fc1(i.e.thesmaller

proteins),whiletotalporositydeterminedbybsAb1injectionswas

usedforthelargerproteins

2.14 LinearpHgradientelutionchromatography

ForlinearpHgradientelutionexperiments,columnswere

equi-libratedwith50mMTris,150mMNaCl,pH7.4andthenpurified

proteinwasloadedonthecolumnto5mg/mLpackedbedusing

anAKTAPure25.Thecolumnwasre-equilibrated,washedwith

50mMcitrate,pH6.7,andthenelutedinalineargradientto50mM

citrate,pH2.7over10columnvolumesat150cm/h.Equilibration

andelutionbufferpHwasmeasuredofflineusingaSevenMultipH

meterfromMettlerToledo(Columbus,OH,USA)equippedwithan

InLabExpertPropHprobefromMettlerToledoandcalibratedwith

pH2,4,7,and10standards.Peakretentionvolumewasdetermined

bypeakmaximumabsorbance(fromtheAKTA

spectrophotome-terat280nm).ElutionpHatthepeakmaximumwascalculated

throughlinearinterpolationbetweentheofflinepHvaluesofthe

two buffersusedtoformthegradient TheAKTA pHtracewas

adjustedtoaccountforthetotalcolumnporosity,εT,andthe

sys-temdelayvolume.ElutionpHatpeakmaximumasdetermined

bylineargradientelutionwasusedtoselectanappropriate

elu-tionpHforthestepelutionphaseinthecapturechromatography

experimentsdescribedinthefollowingsection

2.15 Capturefromclarifiedcellculturebroth

Captureofproteinsfromclarifiedcellculturebroth(CCCB)was

evaluatedinpackedcolumnsusinganAKTAPure25.Columnswere

equilibratedwith50mMTris,150mMNaCl,pH7.4andthenCCCB

wasloadedonthecolumnataresidencetimeof4min.Theload

wascalculatedas85%oftheDBCmeasuredataresidencetime

of4min.Columnswerere-equilibratedwith50mMTris,150mM

NaCl, pH7.4 and then eluted in stepwisefashion with 25mM

acetate,pH3.4–3.5forallstationaryphasesexceptforToyopearl

AF-rProteinAHC-650F(whichusedpH3.2),andMabSpeedrP202

(pH3.8).Elutionpoolswerecollectedfrom100-100mAU(using

AKTAspectrophotometerwitha2mmpathlengthat280nm).Step yieldwasdeterminedusingmassofproductintheload (deter-minedbyProA-HPLC)andpool(determinedbyA280)

2.16 HostcellproteinandDNAmeasurements Hostcellprotein(HCP)concentrations weremeasuredusing thebioaffysandwichimmunoassayontheGyrolabxPworkstation fromGyrosAB(Uppsala,Sweden).Captureanddetection antibod-ieswerein-housereagentsraisedagainstHCPfromthecellline usedtoproducetheantibodiesandFc-fusionproteinsusedinthis work

Host cell DNA concentrations were measured by an in-house method employing a sodium iodide/sodium dodecyl sulfate/ProteinaseKsampletreatmentfollowedbyanisopropanol DNA extraction coupled with a quantitative Polymerase Chain ReactiontargetingtheShortInterspersedNuclearElementDNA sequencerepeatedacrosstheCHOgenomewithSYBRGreenbased detection

3 Results and discussion

3.1 Stationaryphasemorphologicalproperties Table3 summarizesProteinA stationaryphase morphologi-cal properties Ascan be seen in thetable, particlesizes were

inagreementwithdataprovidedbythemanufacturerand stan-darddeviationsmeasuredweretypicallysmallcomparedtothe meanparticlesize(13–24%ofthevolumeaveragedmeanvalue) ParticlesizedistributionscanbefoundinSupplementalmaterial (Fig.S3).Extraparticleporositiesofthestationaryphasespackedin chromatographycolumnsweredeterminedbyfittingtheKozeny equation[42]topressuredropdata(byminimizingtheresidual sumofsquares):

P

L =150

d2 P

(1−εb)2

Whereanduarethemobilephasedynamicviscosityand super-ficialvelocity,respectively,anddPistheaverageparticlediameter Forthiswork,dP-valuesdeterminedfromparticlesizing(volume meandiameters;seeTable3)wereusedinplaceofmanufacturer’s data.Pressuredropcurveswerefoundtobelinearovertherange tested(upto400cm/hin1.1cm×10cmpackedbeds)andthe mag-nitudeofthepressuredropwasconsistentwithexpectationsbased

diametermanufacturing-scalecolumnwilllikelybehigherdueto

alossofwallsupport;nonetheless,thedatapresentedisusefulto

assessrelativedifferencesbetweenthestationaryphases.Thefitted

εb-valuesareshowninTable3andpressuredropcurveswiththe

fittedKozenyequationareshowninSupplementalmaterial(Fig

S4)

Particleporositiesmeasuredusingsodiumchlorideweregreater

than0.9 for theless rigid (agarose andcellulose) particles and

between0.8–0.9forthemorerigid(syntheticpolymer)particles

suggestingahigherpercentageofsolidsinthebasematricesforthe

syntheticpolymersupports.Particleporositywasalsoestimated

withmAb3andbsAb1undernon-binding(acidic)conditions.As

expected,thelargerproteinscouldnotaccesstheentirepore

vol-umeduetostericrestrictions,andresultedinporositiesthatwere

lowerthanobtainedusingsodiumchloride,withonlyminor

dif-ferencesobservedbetweenthemAb(εp,mAb3=0.34–0.76)andthe

slightlylargerbsAb(εp,bsAb1=0.30–0.70)asshowninTable3

Inertdextanprobes,rangingfrom10–670kDa,wereusedto

fur-therelucidateparticleporosity(datainSupplementalmaterial,Fig

S5).Forallthestationaryphasestested,thelargest670kDa

dex-tranprobecouldaccessalimitedfractionoftheparticlevolume

whilesmallerdextranprobesgainedaccesstoalargerportionof

theparticlevolume.Toestimatetheporeradius,rpore,and

intra-particleporosity,εp,dex,usingthedextranelutiondata,acylindrical

poremodelwasappliedaccordingtoHageletal.[43]:

KD=εp,dex



1− rs

rpore

2

(4)

WhereKDisthedistributioncoefficientandrsisthedextran

hydro-dynamicradius.Forthisworkrs-valueswereestimatedfromthe

dextranviscosityradiicorrelationofSquire[44]:

where M is the dextran molecular weight A plot of 

KD vs

rs was used to obtain the fitted values of rpore, and εp,dex (by

linear least squares) The εp,dex-values obtained, as shown in

Table3,wereslightlyhigherthanthoseofnon-retainedproteins

and slightly lower than those obtained using sodium chloride

injections,suggestingthatthedextranprobesexperiencea

macro-porousstructurefreeofverysmallporesthatonlysodiumchloride

canaccess.Therpore-valuesobtainedfromthemodelfit,asshown

inTable3,weregenerallyconsistentacrossstationaryphases,with

valuesintherangeof30–60nm.Thesedatasuggestthatthe

par-ticleshavehighlyporousstructureswhichwassupportedbyTEM

ofthestationaryphases(datainSupplementalmaterial,Fig.S2)

TEM micrographs showed spherical particles with

well-connectedporousnetworks,regardlessofthebasematrixmaterial

Somedifferences wereobservedinthemorphologyofthebase

matrixstructures,wherethemorerigidsyntheticpolymer

mate-rials appeared to have structures that include dense nodes

surrounded by pore networks, but in general the pore sizes

appearedtobesimilar,whichisconsistentwiththecylindricalpore

modelresultsdiscussedabove

Substantialvariationinstainingwasobservedacrossthe

sta-tionaryphasesasshowninFig.S2.Forexample,MonofinityAand

MabSelectPrismAappearedtobeheavilystainedwhileMabSelect

SuRepccandPraestoAPshowedlittlestaining.Thestainsemployed

inthisstudy,uranylacetateandReynold’sleadcitrate,are

primar-ilyutilizedforTEMimagingofbiologicalspecimens.Exactlyhow

theyinteractwithembeddedstationaryphasespecimensofthis

natureremains amatterofconjecture Therefore,whileheavier

stainingimpliesaregionwithahigherconcentrationofelectron

denseatoms, it is difficulttodraw conclusionsbeyond this In

particular,wewouldcautionagainstinterpretationsthatlink

stain-ingvariationtoProteinAliganddensityordistributionunderthe assumptionthatthesestainsprimarilyinteractwiththe polypep-tidechainoftheProteinAligand

3.2 Adsorptionisotherms Fig.1showsadsorptionisothermsformAb1onProteinA sta-tionary phases Adsorption isotherms for all other protein and stationaryphasecombinationsareavailableinSupplemental mate-rial (Fig S6) The isotherm data were fit with the Langmuir isotherm:

q=1qm+KC

whereqmisthemaximumbindingcapacityandKisthe adsorp-tionconstant.Fittedvalues oftheseparameters (determinedby minimizing the residual sum of squares) are available in Sup-plementalmaterial(Table S2).Sincethere arechallengesinthe isothermexperimentation,particularlywithaccurate determina-tion of stationary phase volume added to the isotherm batch binding experiment, we adopted a methodology that relies on breakthrough experiments to estimate qm-values in a manner similartoNgetal.andRecketal.[45–47].Theqm-value(and corre-spondingexperimentalq-valuesshowninFig.1andFig.S6)were normalizedtotheEBCsascertainedwithEq.(1)usingfull break-throughcurvessuchthatqmissetequaltotheEBC.Thisapproach assumesforEq.(6)thatq=qmatC=CF,whichisvalidforthecaseof

anearlyrectangularisothermprovidedthatCFliesintheflat por-tionofthecurve.Inmostinstances,theqm-valuechangedbyless than15%whennormalizingtotheEBCvalue

AscanbeseeninFig.1(andSupplementalmaterialFig.S6)the isothermsarehighlyfavorableforallstationaryphases,whichis consistentwithProteinAbehaviorpreviouslyreportedinthe lit-erature[37,39,40] Equilibriumcapacities arequitehighforthe ProteinAstationaryphases,greater than140mg/mLparticlein someinstances.Tomakecomparisonswithinthelargedataset, EBCsofstationary phaseswere averagedfor a givenstationary phaseacrossvariousproteingroupings(e.g.allofthemAbs,orallof theproteinswithmolecularweightsintherangeof147–159kDa) ComparingstationaryphasesintermsofaverageEBCvaluefor pro-teinswithmolecularweightssimilartomAbs,MabSelectPrismA andMabSelectSuRe pccperformedbest,averaging>126mg/mL particle KanCapA3G and MabSelect SuRe LXshowedthe next highestequilibriumcapacities(107–112mg/mLparticle),followed

byToyopearlAF-rProteinAHC-650F,MonofinityA,andPraestoAP (92–96mg/mLparticle),andthenAmsphereA3,MabSpeedrP202, and Eshmuno A (81–86mg/mL particle) Similarcapacities and trendswereobservedforthelargerbsAbs(∼200kDa);however, averageEBCsforthelargerFc-fusionproteins(280–310kDa)were 50-65%ofthethosedeterminedforthesmallerproteins

3.3 Dynamicbindingcapacity Dynamicbindingcapacitiesat10%breakthroughwere deter-minedforallcombinationsofproteinsandstationaryphasesfor2.4,

4,and6minresidencetimes.Fig.2showsDBCplottedasafunction

ofresidencetime.DBCvaluesarealsoavailableintabularformatin Supplementalmaterial(TableS1).Forthe4minresidencetime con-dition,fullbreakthroughcurveswereobtainedforcolumnsloaded

to150–220mg/mLpackedbed,andareshowninFig.3formAb1 Fullbreakthroughcurvesat4minresidencetimeforthe remain-ingmoleculeandstationaryphasescombinationsareavailablein Supplementalmaterial(Fig.S7)

AscanbeseeninFig.2,theProteinAstationaryphaseshave highdynamic bindingcapacities, greaterthan70mg/mL packed bedinsomeinstancesat4–6minresidencetimes.Therangeof

Fig 1. Equilibrium isotherms for mAb1 on Protein A stationary phases Open circles are experimental data and the solid line is a fit of the Langmuir isotherm, Eq (6), with parameters given in Table S2.

dynamicbindingcapacitiesthatwereobservedwassimilarforthe

mAbsandbsAbsexceptformAb1,whichwashigherforall

sta-tionaryphases.ForFc1,whichissimilarinsizetothemAbs,the

DBCrangewasbroaderacrossstationaryphases,butlargely

sim-ilartomAbsandbsAbs.ForthelargerFc2andFc3molecules,the

bindingcapacitieswereconsiderablylower,reaching30-40%ofthe

capacitydeterminedforthesmallermAbsandbsAbs

SimilartotheanalysisofEBCintheprevioussection,

compar-isonsweremadebetweenstationaryphasesintermsofaverage

DBCat4minresidencetimeforproteinswithmolecularweights

similartomAbs(147–159kDa).Thestationaryphasesgenerallyfell

intogroupswithMabSelectPrismAandMabSelectSuRepcc

hav-ingthehighestDBCsforallmoleculesat4minutesresidencetime

ThenextgroupofstationaryphasesincludedKanCapA3G,

Mab-SelectSuReLX,ToyopearAF-rProteinAHC-650F,PraestoAP,and

AmsphereA3,whiletheremainingstationaryphasesbehaved

sim-ilarlytoeachotherinathirdgroup.ForlargerbsAbs(∼200kDa)and

thelargerFc-fusionproteins(∼300kDa)theaverageDBC

compari-songroupingsremainedthesamebutAmsphereA3andToyopearl

AF-rProteinAHC-650Fmovedtothetopofthesecondgroupwhile

MabSelectSuReLXandKanCapA3Gfellslightlywithinthesecond

group.ThisobservationsuggeststhatAmsphereA3andToyopearl

AF-rProteinAHC-650Fmaybeabletobetteraccommodatethe

largerproteins

In all cases, theDBC data follow the expected trend where

shorterresidencetimesresultedinlowerdynamicbinding

capaci-ties.ThedecreaseinDBCbetween6minand4minresidencetimes

wasconsistentformAbsandbsAbsandwastypicallysmallerthan

thedecreasefrom4minto2.4min.Comparingstationaryphases

foragivenmoleculerevealedthatMabSelectPrismA,Monofinity

A,MabSelectSuReLX,KanCapA3G,andPraestoAPwereslightly moresensitivetoresidencetime,asthesestationaryphasesshowed steeperdeclinesinDBCbetween4minand2.4minresidencetimes acrossmostproteinstested.Thelargerimpactofresidencetime suggeststheseresinsmayexperiencehighermasstransfer resis-tance.Thistopicisexploredinthenextsection

3.4 Proteinmasstransport Masstransferplaysakeyroleintheperformanceofmodern Pro-teinAstationaryphases[36–39].Therefore,examinationofprotein masstransportiscrucialtogainacomprehensiveunderstanding

ofthesematerialsandmakeinformedjudgements.Tocharacterize proteinmasstransportinpackedbeds,wechosethe chromatogra-phygeneralratemodel,givenby:

∂c

∂t =−v ∂c

∂z+Dax∂2c

∂z2−1−εb

εb

3

rpkf(c−cp|r=rp) (7)

∂cp

∂t =Dp



∂2cp

∂r2 +2 r

∂cp

∂r



−1−εp

εp

∂q

Eq.(7)representsamassbalanceontheinterstitialcolumn vol-umewithtermsforconvection,dispersion,andfilmmasstransfer

Eq (8)representsa massbalanceonthestationaryphase with termsforporediffusionandadsorption

Danckwertsboundaryconditionswereappliedatthecolumn inletandoutlet[48].Asymmetryconditionwasassumedatthe beadcenter,andstagnatefilmmasstransferwasappliedatthe boundaryofthestationaryphaseandtheinterstitialcolumn vol-ume[48].Localequilibriumwasassumedforproteinadsorptionas

Fig 2.Dynamic binding capacities as a function of residence time for mAbs, bispecific antibodies, and Fc fusion proteins on Protein A stationary phases.

describedbyEq.(6).Equations(7),(8)andtheboundaryconditions

werespatiallydiscretizedusingfinitevolumesandtheweighted

essentiallynon-oscillatory(WENO)method[48].Theresulting

sys-temofordinarydifferentialequationswasnumericallyintegrated

usingCADETsoftwareversion2.3.2(64bit)runningonaWindows

7PCwithMATLABversionR2014bandanIntelCOREi7CPU[49]

Langmuirisothermparameters,Kandqm,aregivenin

Supple-mentalmaterials(TableS2).Axialdispersion,Dax,wasneglected

[50,51].Theparticleradius,rp,extraparticleporosity,εb,and

parti-cleporosity,εp,wereobtainedfromTable3.εp,mAb3wasusedforthe

smallerproteins(mAb1-3,bsAb3,andFc1),andεp,bsAb1wasusedfor

thelargerproteins(bsAb1-2,Fc2-3).Thefilmmasstransfer

coef-ficient,kf,wasestimatedfromthecorrelationgivenbyCarberry

[52]:

kf =1.15u

Where NRe=dpu/εb (Reynolds number) and NSc=/D0

(Schmidtnumber)

Usingtheparametervaluesdescribedabove,porediffusivity,Dp,

wasfittoexperimentalbreakthroughdatabyminimizingthe

resid-ualsumofsquares(RSS).Fig.3showsmodelingresultsobtained

formAb1.Ascanbeseenfromthefigure,thegoodnessoffitwas

generallyexcellentinthiscasebutdidvaryfromstationaryphase

tostationaryphase.Otherproteins(showninSupplemental

mate-rialFig.S6)tendedtobehavesimilarly.Perhapsnotsurprisingly,

overallabetterfitwasobtainedwiththemAbsandapoorerfitwas

obtainedwiththelargerFc-fusionproteins

Whenthegoodnessoffitwasinferioritwasassociatedwith

instanceswherethebreakthroughcurvetailedtoagreaterextent,

suggestingaslowapproachtoequilibrium.Thisobservationis con-firmedquantitativelybythestandarderrorofregression,S,also showninFig.3(andFig.S6inSupplementalmaterial).Thetailing behaviorhaspreviouslybeenexplainedbasedonaheterogeneous bindingmechanismwheretherearefastbindingsitesthatare dif-fusioncontrolledandslowsitescontrolledbybindingkinetics[39]

InthecontextofProteinA,thisexplanationintuitivelymakes senseastherearemultiplebindingsitesonasingleProteinAligand [7,34].Wecanspeculatethatthediffusioncontrolledfastbinding sitescorrespondtolowbindingoccupancyofaProteinAligand.On theotherhand,asligandoccupancyincreases,wecansurmisethat theremainingbindingsitesmaybecomelessaccessibleandmore stericallyhindered,correspondingtoslowbindingkinetics There-fore,ininstanceswherethisbehaviorwasobserved,thereported effectiveporediffusivity,De (=εpDp),valueessentiallyrepresents

anaveragethatincorporatesresistancesduetobothporediffusion andbindingkinetics,likely havinglimitedpredictiveusefulness Nonetheless,evenwhenthegoodnessoffitvaried,theDe-values obtainedwereoftenconsistent.Forexample,basedonboth quan-titativemeasures(S)andqualitativecomparison,thegoodnessof

fitspannedabroadrangeforthethreemAbsonMabSelectSuRe

LX.Despitethis, theDe-valuesobtainedbyminimizationofRSS variedrelativelylittle,rangingfrom3.0–3.5×10−8cm2/s,whichis probablywithintheexpectederrorforthistypeofmeasurement Fig.4summarizesDe/D0forallproteinsandstationaryphases evaluated in this study Quantitative values of all results sum-marizedin Fig.4are provided inSupplementalmaterial (Table S2).D0 valuesaresummarized inTable1andDLSdatausedto estimateD0aregiveninSupplementalmaterial(Fig.S1).Where available, the De-values obtained were largely consistent with

Fig 3. Breakthrough curves for mAb1 on Protein A stationary phases Open circles show experimental data and solid lines show fitted model results based on Eqs (7) and (8) using parameters in Table S2 The standard error of regression, S, is given for each stationary phase.

previouslyreportedvaluesforProteinAstationaryphases

eval-uated athigher proteinconcentrations [36,38,39].In particular,

theDe-valuesobtainedin this studyfor monoclonalantibodies

at5mg/mLconcentrationonMabSelectSuRe,whichrangedfrom

4.8–5.7×10−8cm2/s,wereingoodagreementwithvaluesreported

byHahnandcoworkers[38].Intheirdetailedstudy,aDe-valueof

5×10−8cm2/swasobtainedforshallowbeduptakeonMabSelect

SuReat3.0mg/mLproteinconcentration.Withrespectto

exter-nalmasstransfer,theBiotNumber(NBi)valuesobtainedacross

allstationary phasesand proteinstested rangedfrom118-990,

suggestingfilmmasstransferplaysanegligibleroleunderthe

con-ditionsutilizedforthisstudy

3.5 LinearpHgradientelutionchromatography

SinceProteinAchromatographyisoftenemployedaspartofa

platformapproach,itisimportanttoselectanelutionpHthatworks

consistentlyforamajorityofthemodalitiestobepurifiedwiththe

platformprocess.Thismayincludemorethanoneantibodyformat

aswellasFc-fusionproteins.Table4summarizestheelutionpH

determinedbypHgradientelutionforallproteinandstationary

phasecombinations.Ascanbeseeninthetable,elutionpHwas

foundtooccurbetween3.5–3.8andonlyminorvariationswere

observedforallmoleculesonagivenstationaryphase.Therewere

afewexceptionstothisbehaviorthatwerespecifictoaproteinor

stationaryphase.Forexample,ToyopearlAF-rProteinAHC-650F

requiredthelowestpHforelutionofallproteins(0.1–0.3pHunits

lower),whiletheelutionpHforMabSpeedrP202wasfoundtobe

higherthantheotherstationaryphases(0.2–0.5pHunitshigher)

ThisuniquehighpHelutiononMabSpeedrP202couldbebeneficial

forproteinsthataresensitivetoacidicconditions,suchasFc-fusion proteins,whichcanaggregaterapidlyunderacidicconditions[53] TheonlyproteinthatwasoutoftrendwasFc3whichwasfoundto elute0.3–0.4pHunitshigherthanallotherproteins.Thisdataset wasusedtoselectanappropriateelutionpHforthestepelution conditionsusedforpurificationofselectmoleculesfromclarified cellculturebroth.InmostcasespH3.4–3.5wassuitableforstep elution;however,AF-rProteinAHC-650FrequiredalowerpH(pH 3.2)whileMabSpeedrP202couldelutetheselectmoleculeswith

ahigherpH(pH3.8)

Basedonpreviousreportsintheliterature,theunderlyingcause

oftheobservedvariationinelutionpH,andinthemilderelution conditionsseenforMabSpeedrP202inparticular,aremostlikely duetoligandpolypeptidesequencedifferencesasaresultofprotein engineeringtoimproveligandperformance[54].Thisremainsa matterofspeculation;however,asProteinAligandsequence infor-mationforthestationaryphasesevaluatedinthisstudyhavenot beenpubliclydisclosedbythemanufacturers.Nonetheless,similar strategieshavedemonstrateditispossibletoeluteProteinAunder milderpHconditionsthroughdestabilizationoftheliganditselfor theligand-Fcinteraction[55–57]

3.6 Capturefromclarifiedcellculturebroth Capturechromatographyexperimentswereconductedtotest theabilityoftheProteinAstationaryphasestoselectivelypurify threeoftheproteinsinthisstudy.Inallcases,thechromatogram forthecapturestepwaswellbehaved,showingsharpelutionpeaks andlittletonoproteininthe0.1Maceticacidcolumnstrip(Fig.S8

inSupplementalmaterialshowsexamplechromatogramsforthe

Table 4

Elution pH on Protein A stationary phases as determined by pH gradient elution.

Stationary

phase

pH at elution peak max mAb1 mAb2 mAb3 bsAb1 bsAb2 bsAb3 Fc1 Fc2 Fc3

MabSelect SuRe LX 3.7 3.6 3.7 3.7 3.7 3.7 3.6 3.6 4.0 MabSelect SuRe pcc 3.7 3.6 3.7 3.7 3.7 3.7 3.6 3.6 4.0

AF-rProtein A HC-650F 3.4 3.4 3.5 3.5 3.5 3.4 3.2 3.3 3.7

Fig 4.D e /D 0 values for mAbs (top panel), bsAbs (middle panel), and Fc-fusion

pro-teins (bottom panel) measured on Protein A stationary phases D e -values obtained

from fit of breakthrough curves using Eqs (7) and (8) with parameters given in Table

S2 D 0 -values determined by DLS are given in Table 1.

captureofmAb1fromCCCB).Fig.5summarizestheprocess

perfor-manceandresultingproductqualityintheProteinAelutionpool

Ascanbeseeninthefigure,yieldswereconsistentlyabove90%,

withonlyminordifferencesobservedbetweenstationaryphases

Elutionpoolvolumeswereslightlymorevariable,fallingbetween

1.8–3.8columnvolumes;however,thevariabilitywastheresultof

afewproteinandstationaryphasecombinations,andmostofthe

resultsfallinanarrowerrangeof2–3columnvolumes.Insome

cases,higherbindingcapacity stationaryphasesdidhavelarger

elutionpoolscomparedtothelowercapacitystationaryphases,

butnotrendsamonghighcapacityresinswereobserved

ProteinAchromatographyismostoftenemployedasthe

cap-turecolumninapurificationtrainandisexpectedtoprovidehighly

selectivebindingwhileprocess-relatedimpuritiesremaininthe

columnflowthrough.Moreover,ProteinAchromatographyisnot

expectedtoreducethelevelofproduct-relatedimpurities,such

asaggregates,undertypicaloperatingconditions.Ascanbeseen

inFig.5,HCP,DNA,andaggregatelevelsintheelutionpoolwere relativelyconsistentbetweenstationaryphasesforagivenprotein withonlyminorexceptions.WhenconsideringHCPlevelsinthe elutionpool,therewasnostationaryphasethatprovidedthebest HCPremovalforallthreemoleculestested.Forexample,MabSelect SuReLXhadthehighestHCPlevelofallstationaryphasestestedfor mAb1,butthelowestforbsAb3.Similarly,ToyopearlAF-rProteinA HC-650FhadthelowestHCPformAb1,butthehighestformAb3 ForDNAclearance,notrendswereobserved;however,Monofinity

Afaredthebestofallthestationaryphasesforallthreeproteins tested.WhencomparingaggregatelevelsformAb1andmAb3in theelutionpoolacrossstationaryphases,onlysmalldifferencesare seen,whereasmorevariabilitywasobservedfortheaggregatelevel

inthebsAb3pools.Moreover,sometrendsinthebsAb3datacan

beseenwherehigher(AmsphereA3)orlower(MabSpeedrP202) aggregatelevelswereobservedinelutionpoolsthathadhigheror lowerproteinconcentrations,respectively(concentrationdatanot shown)

Whenmakingcomparisonsbetweenproteins,itwasobserved thatsomeproteinsarenotaseasilypurifiedfromCCCB,andhigher levelsofprocess-andproduct-relatedimpuritiesareseen.Aside fromtheexceptionsnotedabove,impuritylevelswerequite con-sistent when comparingstationary phases and no trendswere observedbasedonstationaryphasesproperties(e.g.higherbinding capacity vs lower binding capacity;natural vs synthetic poly-mericbackbonematerials)orprocessperformancedata(e.g.yield, pool volume, pool concentration) It should be noted that the capturepurificationexperimentsinthisworkonlyemployeda re-equilibrationstep(50mMTris,150mMsodiumchloride,pH7.4); however,amorestringentwashisoftenemployedtoreduce non-specificinteractionsofHCPandstationaryphasesand/orHCPand theproteinofinterest[58–60].Itisunclearifamorestringentwash wouldbenefitonestationaryphaseoveranother,butitisplausible thatthiscouldoccurandmaymeritinvestigationwhendeveloping

aProteinAcapturestep

3.7 Bivariatecorrelationanalysis Fig.6showsamatrixofbivariatescatterplotscorrelating sta-tionaryphasemorphologicalpropertieswithproteinbindingand mass transfer The scatterplot matrix was generated using the gplotmatrix function in MATLAB version R2014b The diagonal showsunivariatehistogramswiththehighestfrequencybin nor-malized to full scale on the ordinate axis Plots in the upper left-hand quadrant are sparser as they correlate morphologi-calproperties toothermorphological properties.Similartothe approachdescribedinproteinmasstransportabove,ε was