• Aucun résultat trouvé

Negative gradient slope methods to improve the separation of closely eluting proteins

N/A
N/A
Protected

Academic year: 2022

Partager "Negative gradient slope methods to improve the separation of closely eluting proteins"

Copied!
10
0
0

Texte intégral

(1)

Article

Reference

Negative gradient slope methods to improve the separation of closely eluting proteins

FEKETE, Szabolcs, et al.

Abstract

In the present work, we describe the fundamental and practical advantages of a new strategy to improve the resolution of very closely eluting peaks within therapeutic protein samples. This approach involves the use of multiple isocratic steps, together with the addition of a steep negative gradient segment (with a decrease in mobile phase strength) to "park" a slightly more retained peak somewhere along the column (at a given migration distance), while a slightly less retained compound can be eluted. First, some model calculations were performed to highlight the potential of this innovative approach. For this purpose, the retention parameters (log k 0 and S ) for two case studies were considered, namely the analysis of a mixture of two therapeutic mAbs (simple to resolve sample) and separation of a therapeutic mAb from its main variant (challenging to resolve sample). The results confirm that the insertion of a negative segment into a multi-isocratic elution program can be a good tool to improve selectivity between critical peak pairs. However, it is also important to keep in mind that this approach only works with large [...]

FEKETE, Szabolcs, et al . Negative gradient slope methods to improve the separation of closely eluting proteins. Journal of Chromatography. A , 2021, vol. 1635, p. 461743

DOI : 10.1016/j.chroma.2020.461743 PMID : 33260022

Available at:

http://archive-ouverte.unige.ch/unige:146721

Disclaimer: layout of this document may differ from the published version.

1 / 1

(2)

ContentslistsavailableatScienceDirect

Journal of Chromatography A

journalhomepage:www.elsevier.com/locate/chroma

Negative gradient slope methods to improve the separation of closely eluting proteins

Szabolcs Fekete

a,b,

, Amarande Murisier

a,b

, Jennifer M. Nguyen

c

, Matthew A. Lauber

c

, Davy Guillarme

a,b

aSchool of Pharmaceutical Sciences, University of Geneva, CMU-Rue Michel Servet 1, 1211 Geneva 4, Switzerland

bInstitute of Pharmaceutical Sciences of Western Switzerland, University of Geneva, CMU-Rue Michel Servet 1, 1211 Geneva 4, Switzerland

cWaters Corporation, 34 Maple Street, Milford, MA 01757-3696, United States

a rt i c l e i nf o

Article history:

Received 26 August 2020 Revised 16 November 2020 Accepted 19 November 2020 Available online 23 November 2020 Keywords:

therapeutic proteins monoclonal antibody method development gradient elution negative gradient step

negative segmented multi-isocratic mode

a b s t r a c t

Inthepresentwork,wedescribethefundamentalandpracticaladvantagesofanewstrategytoimprove theresolutionofverycloselyelutingpeakswithintherapeuticproteinsamples.

Thisapproachinvolvestheuseofmultipleisocraticsteps,togetherwiththeadditionofasteepnegative gradient segment (withadecrease inmobile phasestrength) to "park"aslightlymore retained peak somewherealongthecolumn (atagivenmigrationdistance),whileaslightlyless retainedcompound canbeeluted.

First,somemodelcalculationswereperformedtohighlightthepotentialofthisinnovativeapproach.For thispurpose,theretentionparameters(logk0 andS)fortwocasestudieswereconsidered,namelythe analysisofamixtureoftwotherapeuticmAbs(simpletoresolvesample)andseparationofatherapeutic mAbfromitsmain variant(challenging toresolvesample).Theresultsconfirmthattheinsertion ofa negativesegmentintoamulti-isocraticelutionprogramcanbeagoodtooltoimproveselectivitybetween criticalpeakpairs.However,itisalsoimportanttokeepinmindthatthisapproachonlyworkswithlarge solutes,whichmoreorlessfollowan“on-off” typeelutionbehavior.

Tworealapplicationsweresuccessfullydevelopedtoillustratethepracticaladvantageofthisnewap- proach,includingtheseparationofatherapeuticmAbfromitsmainvariantpossessingverycloseelution behavior,andtheseparationofacarrierproteinfromanintactmAbasmightbeencounteredinaquan- titativebioanalysisassay.Thesetwoexamplesdemonstratethatimprovedselectivitycanbeachievedfor proteinRPLCthroughtheinclusionofanegativegradientslopethatselectivelybifurcatestheelutionof twoormorepeaksofinterest.

© 2020TheAuthor(s).PublishedbyElsevierB.V.

ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/)

1. Introduction

Liquid chromatographic separations ofproteins are performed in gradient elution mode. In general, simple linear gradients are performed since they are easy to generateand control,and con- sequentlythosemethodscanbeeasilytransferred [1].However,a simple linear gradient is often unable to provide sufficient chro- matographicresolution.Therefore,segmentedgradientscanbeap- pliedto improveseparationquality [2].Two-segment(“bi-linear”) gradients areoftenusedtoshortentheanalysistimewhenasep-

Corresponding author.

E-mail address: szabolcs.fekete@unige.ch (S. Fekete).

aration includes a few well-resolved late-eluting peaks. Then, a steeper gradient segment can be set for the late eluting peaks [1]. For more complex samples, multiple gradient segments can becombinedtoattainsuitableseparation.Inordertofacilitatethe elutionofthepeaks,itiscommonknowledgethatgradientslopes shouldalwaysbepositive; howeveroccasionally,oneormoreiso- craticstepscanbeinsertedtoobtainthemostoptimalseparation [2].

Besidesmulti-linear gradients,non-linear gradientsmight also providesome benefits.Power functionbasedgradients havebeen successfullyappliedfortherapeuticproteinseparationsforbothre- versedphase(RPLC)andion-exchange(IEX)chromatography[3,4].

Anothertype of non-lineargradient can alsobe usefulwhen the

https://doi.org/10.1016/j.chroma.2020.461743

0021-9673/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )

(3)

S. Fekete, A. Murisier, J.M. Nguyen et al. Journal of Chromatography A 1635 (2021) 461743

compounds ofinterestbelong toaseriesofincreasinglymorere- tained analytes(e.g.membersofhomologuesseries).Inthiscase, a logarithmicshape gradient profile provides thebest overall se- lectivity [5].Customizednon-lineargradients (includingbothcon- cave and convexsegments)were alsodeveloped by Kall etal. to separate complex peptidemixturesandwere successfullyapplied in shotgun proteomics[6].Some other complex gradient profiles were alsoappliedbasedon consideringeithertheelutionofonly thefirstandlastelutingpeaks[7]ortheelutionofeachindividual peak[8].

Forapreparativescaleseparation(purification)ofpeptidesand proteins,steporstep-wisegradientshavebeenextensivelyusedin flash chromatographyandcounter-current chromatography[9,10].

Theideabehindsuchstepgradientsisthatonlyoneoraveryfew numberofcomponentshavetobeseparated,whiletheothersam- plecomponentsarejustwashedoutfromthecolumn.

Timperman and co-workers demonstrated that a “saw-tooth”

gradient program allows small subsets of proteins to be eluted fromthecolumnintermittentlyby usingshortgradientstepssep- aratedbyanegativeslopeandisocraticholdingsegments.Theiso- cratic holding periods can be used to perform additional sample processing (e.g. online fractioning fora second dimension analy- sis)[11].Thesaw-toothgradientwasfoundsuperiortoacommon segmentedlineargradient/isocraticmode,sincethenegativesteps preventsbandbroadeningthattakesplaceduringisocraticelution steps.Thissaw-toothgradientwassettoachievecompletesample transfer betweenthefirst-andsecond-dimensionforprotein and peptideidentification[12].Saw-toothgradientsarealsoappliedfor polymerseparations[13].

Armstrongetal. discussedthepossibilityto runasimpleneg- ativelineargradient inRPLCforproteinseparations [14].Theau- thorsreferredtotheirseparationas“non-traditionalreversegradi- ent”. Unusual convex logk

ϕ

plots with global minima were re-

ported for ribonuclease, insulin and myoglobin, and the authors explainedthatthoseobservationswerenotinagreementwithpre- viously reported results. Nevertheless, it was pointed out that if thereisaminimaonalogk

ϕ

plot,thenretentionandelutioncan

beattainedeitherwithpositiveornegativegradients.Theauthors alsoexplainedthatsuchunusualbehaviorwasprobablyrelatedto solubility-based phenomenon.Despitethattheconditionsusedin the study were not ideal (narrow pores of60 ˚A–RPLC phase, 1%

TFAasmobilephaseadditiveandambienttemperature),thethree proteinsweresuccessfullyseparatedwithalinearreversegradient.

Recently,aninnovativestrategytermedas“multi-isocratic” elu- tion has beenshown to provideexponential increase inselectiv- ity betweenprotein variants. It wasdemonstrated that the com- binationofmulti-isocraticstepsandveryshort,yetsteepgradient segments(withsteepnessclosetoinfinity)atsoluteelutionallows onetosettheselectivityasdesiredwhilemaintainingsharppeaks duetosignificantbandcompressioneffects[15].Thismethodwas successfully applied for the analytical scale separation of intact andsubunitdigested samplesofmonoclonalantibodies(mAbs)as well as antibody-drug conjugates (ADCs). Uniform peak distribu- tion (equidistantband spacing)andmuchhigherresolutioncould be achieved than with common linear, multilinear, or nonlinear gradients. Inafollowing study,thisapproachwascombinedwith column coupling to further improve separation power. In such a setup, if aprotein peak is trapped atthe inlet ofa later column segment-ofaseriallycoupledsystem-thenitsbandisrefocused anditelutesasanunprecedentedsharppeak[16].Furthermore,it becamepossible toperform onlineon-columnfractioning ofpro- teinspecieswithinaveryshortanalysistime(~1min)andwithout sampledilution.Similarideahasalreadybeenreportedtosharpen peaksutilizingpost-columnrefocusingandremobilizationontrap- pingcolumns[17,18,19].

Inthe presentwork,the goalwasto furtherimprovethe res- olutionofRPLCseparationsofverycloselyelutingpeaksofthera- peuticproteins.Therefore,therecentlydeveloped“multi-isocratic”

elution mode was upgraded by adding a steep negative gradient segment between the “eluting” and “non-eluting” isocratic seg- ments.Asaresultofthisnegativeslopesegment,itbecomespos- sibleto"park"aslightlymoreretainedpeakalongthecolumnbed (at a given migration distance) witha condition that still allows the slightly lessretained compound to be eluted.Therefore, this approachhasthe potentialto resolvecompounds possessingvery similar retention properties,which are difficultto separate, even withthemulti-isocratictechnique.Here,wepresentsometheoret- ical considerations andillustratethe capabilitiesof thisapproach forlarge solutes.Theexpectedbenefitofinsertingnegativegradi- entsteps(shortsegments)intoa“multi-isocratic” programisalso discussed.Twoapplicationswere developedtoshowthepractical advantage ofthisnewapproach. Thesetwo separations were not feasiblebyapplyingthemulti-isocraticelutionmode(letalonelin- earormulti-lineargradients).

2. Materialsandmethods 2.1. Equipmentandsoftware

ChromatographicexperimentswereperformedonaWatersAc- quity UPLCI-Class system equipped with a binarysolvent deliv- erypump,an autosampler,afluorescence(FL)detectorandaflow throughneedleinjectionsystemwith15 μLneedleanda 2μLFL flow-cell. The overall extra-column volume was about 8.5 μL as measuredfromtheinjectionseatoftheauto-samplertothedetec- torcell.ThedwellvolumewasmeasuredasVd=0.09mL.Dataac- quisitionandinstrumentcontrolwereperformedbyEmpowerPro 2software(Waters).Calculationanddataprocessingweredoneby usingDrylab(4.2)andExcel(Microsoft)software.

2.2. Chemicalsandcolumns

LC-MS grade acetonitrile (ACN) and LC-MS grade water were purchased from Fisher Scientific (Reinach, Switzerland). ULC/MS gradeformicacid(FA)andULC/MSgradetrifluoroaceticacid(TFA) werepurchasedfromBiosolve(Dieuze,France).

Intact mAb Mass Check Standard (murine anticitrinin IgG1) wasobtainedfromWaters.Bovine SerumAlbumin(BSA)waspur- chasedfromSigma-Aldrich (Buchs,Switzerland).TherapeuticmAb (eculizumab) was obtained as European Union pharmaceutical- gradedrugproductfromitsrespectivemanufacturer.

Prototype columns (15 × 2.1 mm)packed with3 μm 2000 ˚A polystyrene divinylbenzene (PS-DVB) particles aswell as a com- mercialBioResolveRPmAbPolyphenylcolumn(50× 2.1mm,2.7 μm,450 ˚A)were provided by Waters (Milford,USA). To preparethe prototype PS-DVB column, a specialized guard column was con- structedwithalowclearanceendnutandalowdispersioncoupler togiveastandardfemaleinlet/femaleoutletconfiguration.

2.3. Sampleandmobilephasepreparation

Intacteculizumabwasdiluted to1 mg/mLwithwaterandin- jectedwithoutfurtherpreparation.WatersIntactmAbMassCheck Standardwasdilutedto0.025 mg/mLwithwaterandmixedwith BSAdilutedto0.25mg/mLwithwater.

For the separation of intact eculizumab variants, the mobile phaseA was0.1% TFA(v/v) inwaterandBwas0.1% TFA(v/v)in ACN.

For the separation of anti-citrinin mAb (Waters Intact mAb MassCheckStandard)andcarrierprotein(BSA),themobile phase Awas0.1%FA(v/v)inwaterandBwas0.1%FA(v/v)inACN.

2

(4)

2.4. Chromatographicconditions

To improve theselectivity of closely elutingproteins, a linear gradient separation was compared to separations achieved with a multi-isocratic elution technique that either contained or did not contain a negative gradient segment (“negative segmented multi-isocraticelution”).First,theparametersofthelinearsolvent strength(LSS)modelweredeterminedfromtwolineargradients.

For eculizumab,the flow rate wasset at0.5 mL/min, column temperaturewassetat85°C,and0.5μLofintacteculizumabsam- ple wasinjected. For the initial linear gradient experiments, the gradient times were set as tG1 = 4 andtG2 = 10 minutes anda 25–50%Bgradientwasrun.

Fortheseparation oftheanti-citrininmAbandcarrierprotein (BSA), theflow ratewassetat 0.4mL/minandthe columntem- peraturewassetat80°C.Sampleinjectionvolumewas0.1μL.For theinitiallineargradientexperiments,thegradienttimeswereset astG1=4andtG2 =10minutesanda20–50%Bgradientwasrun.

Forallmeasurements,datawereacquiredat280nmexcitation and350nmemissionwavelengths(FL).

Theoptimizedconditionsformulti-isocraticand“negativeseg- mented” programs are detailedin the Resultsand Discussion sec- tion.

2.5. Calculations

InLC,theLSSmodel-sometimescalledexponentialmodel-is commonlyusedtodescribetherelationshipbetweensolutereten- tion(k)andmobilephasecomposition(

ϕ

)[20]:

logk=logk0S

ϕ

(1)

where k is the solute retention factor,

ϕ

is the volume fraction

of mobilephase “B” (stronger eluent), Sisa constant fora given solute (it describes how sensitive is the solute retention to mo- bilephasecomposition)andk0 isthe(extrapolated)valueofkfor

ϕ

=0(i.e.,theretentionfactorobservedinpuremobilephase“A”).

The migrationvelocity (u) ofasolute alongthe column(mea- suredatagiven

ϕ

)dependsontheinterstitialmobilephaseveloc- ity(u0)andretentionfactor:

u= u0

1+k (2)

ExpressingkfromEq.(1)andsubstituting toEq.(2)enablesone todescribetherelativesolutemigrationspeed(urel)as:

urel= u

u0 = 1

1+k010(Sϕ) (3)

Thetimespenttoreachpositionzcanbeexpressedas[21]: t

(

z

)

=t0

z

L+1 blog

1+kinbz L

(4)

Wherekinistheretentionfactoratthestartingmobilephasecom- position (inlet retention factor), b is the gradient steepness (b= S·

ϕ

·ttG0),t0isthecolumndeadtime,Listhecolumnlengthand z isthesolutepositionalongthecolumn.Then,thetimetotravel the entire column(z = L) corresponds to the retention time (tr) andcanbewrittenas:

tr=t

(

L

)

=t0

1+1

blog

(

1+kinb

)

(5)

PleasenotethatEqs.(4)and(5)assumelineargradient.Forour calculations,theparameterslogk0 andSwerederivedfromexper- imentallymeasuredretentiontimesdataoftwopreliminarylinear gradientexperiments–performedwithdifferentgradienttimes(tG) (corresponding to different gradient steepness) - using DryLab 4 software. (Pleasenote that, deviationsfromthe LSSmodelmight be observed, especiallywhen workingina broad%Brange–e.g.

ϕ

>0.5, then other non-linear models canbe used - e.g.Neue- Kuss orquadraticmodels -instead oftheLSS model[22]. Inour practice, forpeptides and proteins, lessthan 5% deviationis ob- served in reversed phase chromatography. However here in this study, very narrow

ϕ

ranges were set (< 0.1) and LSS models werefoundtobeappropriate.)

2.5.1. Studyingtheevolutionofselectivity

Then,fromlogk0 andS,theretentionfactor(k) wasestimated fora given mobile phase composition (

ϕ

) forany setof %Bpro-

gram (e.g. linear-, multi-isocratic or negative segmented multi- isocratic program). The solute relative velocity and the travelled distancecan be calculated foranytime point of a given%Bpro- gram. To illustrate solute migration and study selectivity, plots of (1) B fraction vs tG, (2) u/u0 vs z and (3) distance travelled vst were constructed. Simulatedchromatograms were also plot- ted√ assumingthecommongradientbandcompressionfactor;G=

1+p+p2/3

1+p ,withp=2.3b[23].Please notethatthis factorGisonly valid for linear gradient. For our calculations for multi-isocratic separations,we assumedconsecutivelineargradientandisocratic segments.

2.5.2. Optimizationofmulti-isocraticandnegativesegmented separations

It is known that the retention oflarge solutessuch asthera- peuticproteins isverysensitivetothemobilephase composition.

Aminorchangeinthemobilephasecompositioncanindeeddras- ticallyaffecttheirretention(veryhighSvalue intheLSSmodel).

Snyder explainedthisphenomenon by the fact that large solutes areeitherfullycapturedatthecolumninletorcompletelyreleased fromthecolumn[20,24,25].Thisbehavioristodayoftentermedas an “on-off” or “bind-and-elute” mechanism. Very recently,it was indeedshownthattheretentionoflargeproteinscanonlybecon- trolledinaverynarrowmobilephasecompositionrange(e.g.with gradientsapplyingonly3.5–5%B forintactmAbs)[15,16].Their relativemigrationspeedvarieswithinthe0<urel <1rangeonly inthisverynarrow%Bwindow, otherwiseitiseither0or1(cor- respondingto“on”–fullycaptured–stateorto“off”–released–state).

Therefore,themobilephasecomposition requiredtostartthemi- gration of a large molecule (

ϕ

(urel=0.01); when the solute starts travelingwithonly1% ofthemobile phase velocity, u/u0 =0.01) canbeestimatedas:

ϕ

(urel=0.01)= − log

1

0.011 k0

S (6)

Similarly,themobilephasecompositiontoreachthe“off” state (

ϕ

(urel=0.99);unboundstatewithu/u0=0.99)canbewrittenas:

ϕ

(urel=0.99)= − log

1

0.991 k0

S (7)

Ontheotherhand,elutingmobilephasecompositionwithvery low retention factor(

ϕ

k<0.1) andbinding mobile phase composi- tion withhighretention factor(

ϕ

k>100) fora multi-isocratic elu- tionseparationcanbeestimatedas[15]:

ϕ

k<0.1>logk0+1

S (8)

ϕ

k>100< logk0−2

S (9)

The

ϕ

k<0.1 and

ϕ

k>100 can be good starting points for the optimization of a multi-isocratic protein separation. However, in practice,thereissometimes onlya minordifference betweenthe modelparametersofcloselyrelatedproteins(e.g.variantsofintact mAbs)andthusitishardtopredictwhethertheycanbeseparated

(5)

S. Fekete, A. Murisier, J.M. Nguyen et al. Journal of Chromatography A 1635 (2021) 461743

ornot. Forsucha situation,wefound that performinga “screen- ing multi-isocratic gradient” can be very helpful.To realizesuch, a 5-segmentedmulti-isocraticconditionwasset(pleasenote that anynumberofsegments can beset). Themobile phasecomposi- tion forthe initial step (

ϕ

in) wasset to be retentive enough(eg.

ϕ

k>100),whilethecompositionofthelastsegment (

ϕ

last)wasset to beableto eluteallcompounds (eg.

ϕ

k<0.1).Then, fiveequidis- tantsegments (2minutelongintervals)weresetbetweentheini- tial and final compositions. The difference (%Bsegment) between themobilephasecompositionsoftheconsecutivesegmentsforthe caseofnisocraticsegmentscanbedeterminedas:

%Bsegment=

ϕ

last

ϕ

in

n−1 (10)

After performing the first screening run, one can fine-tune the number of steps and the mobile phase composition of the isocratic steps to improve the separation by performing a so- called“stretched” multi-isocraticrun.(Supplementaryfig.1shows a schematic view of the optimization procedure, including (1) a preliminary linear gradient run, (2) a “screening” multi-isocratic run and (3) a “stretched” multi-isocratic run). As a generic sug- gestion, forintactmAbs, an effectivescreeningrun mayconsider a5% Bdifferencebetween

ϕ

last and

ϕ

in(which isduetothehigh S value,typicallyranging between90and150underRPLC condi- tions).

4. Resultsanddiscussion

4.1. Modelcalculations-potentialofinsertinganegativegradient step

It wasrecently shown that a so-called multi-isocratic elution technique could produce uniform peak distribution (equidistant band spacing) fora separation ofprotein species(assuming they obey an on-off typeelution mechanism) [15]. Ideally, theelution distancebetweenpeakscanbeadjustedarbitrarilybychangingthe lengthoftheholdingisocraticsegments.However,thisisonlyfea- sible if just one of the peaks of interest starts migrating within a givenelution gradient–or isocratic hold - segment. In practice, it may happen that not only one, but two or more compounds start tomigratealong thecolumnatagivensegment(because of a lack ofselectivity, their retention behavior andthus modelpa- rameters are verysimilar). Such cases representthelimitsofthe multi-isocratic approach. In caseofco-migration (even if theso- lutesmigrationspeedisslightlydifferent),theselectivitycannotbe increasedanymorewithoutboundaries[15].Accordingly, wewere motivatedto findabettersolution toresolvesuch closelyeluting protein compounds. The idea was to park (“freeze”) a migrating -moreretained-compoundsomewhere alongthecolumn,while lettingalessretainedcompoundcompleteitselutionthroughthe column. Forthis, weexplored theuseofanegativegradientseg- ment along witha “holding” isocraticsegment immediatelyafter theelutionofthelessretainedcompound.

First,somemodelcalculationswereperformedtohighlightthe potential ofamulti-isocraticseparationandtheeffectofinserting anegativegradientsegmentintoamulti-isocraticelutionprogram.

Twosetsofcompoundswerestudied,namelya“simpletoresolve sample” and a “challenging to resolve sample”. For the simple- sample, amixtureofintactrituximabandramucirumabwascon- sidered, sincethere is enough difference betweentheir retention to separate them either with a linear gradient ormulti-isocratic elutiontechnique.Forthechallenging-sample,intactatezolizumab andits mainvariantwere chosensince thissamplealreadyfaced the limitsof the multi-isocratic elution mode in a former study [15].(The parametersof retentionmodels usedforthesecalcula- tionsweretakenfromourpreviousstudies[15,16].)

Fig. 1 A-D showsthe evolution of solute migration, the trav- elleddistance along a columnandthe calculatedchromatograms forthesimple-samplewhenperforminglineargradientandmulti- isocratic separations. Based on Fig. 1 B and 1C, it is clear that once the two compounds start their migration (switching to “off” mode), their relative migration speed and acceleration are nearly the same along the entire column. However, peak 1 startsmigrating(

ϕ

(urel=0.01)= 0.335)farearlierthanpeak 2does (

ϕ

(urel=0.01) = 0.366). With a 10 min long linear gradient (32 to 42%B), peak1 startsmigratingafter aparking time (tpark) of1.5 min, while peak 2 parks at the column inlet until 4.6 min. Af- tertheirreleasefromtheheadofthecolumn,theytravelthrough thechromatographicbedwithin nearly thesameamountoftime (ttrav = tr–tpark,gives 2.67 and2.52 min,respectively). Therefore, theselectivityismostlydeterminedbythedifferenceoftheirpark- ing times (tpark = +3.1 min) and not by their travelling time (ttrav=-0.15min).Duetothelargedifferencebetweenthepark- ing times, the separation of those peaksis easy to achieve with a linear gradient. Figs. 1 E-Hshow the caseof a multi-isocratic separation. Setting 28%B for the starting isocratic binding seg- ment resulted in very high initial retention for both compounds (k1 = 2.03107 and k2 = 1.111011). Over a 2 min isocratic seg- ment,thetwopeakspracticallydonotmovefromtheheadofthe column.Changingto36.8%Bmobilephasecompositionresultedin theimmediateelutionofthelessretainedpeak(k1=0.07).Mean- while, peak 2 remained to be strongly retained (k2 = 61.15),al- beitwithanindicationofsomeveryslowmigration(urel =0.016).

Holding thissecond isocratic segment for4 min (6 min of total run time) resulted in z = 0.9 cm travelled distance for peak 2 (Figs.1FandG,redcurve).Subsequentlysettingthemobilephase composition to45%Bresultedintheimmediateelutionofpeak 2 (k2 =1.4410−7). Inconclusion,dueto thelargedifference ofre- tention betweenthe two mAbs(determined by logk0 values), ei- therlineargradient ormulti-isocraticseparations areeasy toim- plement. In the end, the latter technique has the advantage to drasticallyimproveselectivity. Withtheconditionssetinthisex- ample (36.8%Bforthesecond isocratic segment), theelution dis- tance(selectivity)betweenthetwo peakscan beincreasedup to

~40min.(With36.8%B,ittakesabout44minforpeak2-migrat- ing with urel = 0.016 - to travel the entire column length of 10 cm.)

Figs. 2 A-D illustrate the challenge to separate intact ate- zolizumab andits main hydrophobic variantby applyinga linear gradient. Peak 1 startsmigrating at

ϕ

(urel=0.01) = 0.335 (33.5%B), whilepeak 2begins travellingat

ϕ

(urel=0.01)= 0.338 (33.8%B).By settinga 10min long lineargradient(32 to 42%B), peaks1and 2willstartmigratingaftertpark=1.50 and1.76min,respectively.

Followingtheirreleasefromthecolumninlet,theirtravellingtimes arealsonearly thesame(ttrav= 2.66and2.58min,respectively).

Since both their parking times(tpark = +0.26 min) and travel- ling times (ttrav = -0.08 min) are almost identical,it is hardly possibletoaffordselectivitythroughtheapplicationoflineargra- dients.However, by runninga multi-isocratic program,the selec- tivity can be slightly increased(Figs. 2 E-H). By setting 33%B as initialisocraticsegment,bothcompoundswerefoundtobehighly retained (k1 = 312and k2 = 590). After two minutesof holding time andaswitch to 34.5%B, bothcompounds startedmigrating withurel =0.09and0.05(Fig.2F).Attheendofthesecondseg- ment,peak1traveledz=2.5cmwhilepeak2traveledz=1.6cm (Fig.2G).Then,duringthethirdsegment(35.3%Bholdfor2min), peak1acceleratedandleftthecolumn, whilepeak2approached z=9.1cm.Finally,thelastsegment(35.9%B)quicklyelutedpeak2 fromthecolumn.Itisimportanttonoticethatoncepeak1started migrating(switchesto“off” mode) sotoodidpeak 2,albeitwith aslightlylowervelocity.Sincethedifferencebetweentheirmigra- tion speed is limited (a factor of 1.2–1.6 difference can be real- 4

(6)

Fig. 1. Evolution of selectivity (“simple-sample”) with linear gradient 32–42 %B in 10 min (A,B,C,D), and with multi-isocratic (E,F,G,H) elution. The %B program for the multi-isocratic run was 28%B (0–2 min), 36.8%B (2.01–6 min) and 45%B (6.01–10 min). F = 0.3 mL/min, 100 ×2.1 mm column, ε= 0.62, rituximab peak (1): S = 96.4, log k 0= 36.3, ramucirumab peak (2): S = 105.2, log k 0= 40.5.

ized), it was not possible to find conditions that simultaneously yieldedhighvelocityforone compound(“off” mode)andlow ve- locityfortheother (closeto“on” mode).Incontrast,forthecase ofthesimple-sample,afactorof58wasobtainedbetweenthemi- gration speeds ofthe two solutesduring the second segment of themulti-isocraticprogram,asshownonFig.1F.

Thechallenging-samplecanbeusedtoillustratethelimitations of the multi-isocraticelution techniqueandthe beneficial effects of inserting a negative gradient segment. Fig. 2 G shows an ex- ample whereina short,negativeandsteep gradientsegment was insertedjustaspeak1leftthecolumn(tG=5.5min,greydashed line in Fig. 2 G). With this, the migration of the more retained compound was stopped at the position to which it had traveled up until that point intime(z = 7.2 cm,thecrossing point of the

greydashed lineandred curve onFig.2 G).Fig.3illustrates the separationfor thecasewhen at5.5 min,the mobile phasecom- positionwassetback from35.3to 33%Bandheld until8min(to give a2.5min holding(parking)time).As suggestedby Figs.3 B and3C,peak2didnotmovetoanyappreciableextentduringthis negativestep. Upon setting a strongermobile phase composition (e.g.38%B)peak2wasmadetoimmediatelyelute(k=5.210−3).

Based onthesemodelcalculations, it isproposed that anegative segment(decrease ofmobilephasestrength)canbe insertedinto a multi-isocratic elution program to improve selectivity between criticalpeakpairsfoundinlargebiopharmaceuticaldrugproducts.

Tothebestofourknowledge,thisisthefirsttimethat sucha combinationofpositiveandnegativegradients(andisocratichold- ing)segmentsarecombinedandapplied.

(7)

S. Fekete, A. Murisier, J.M. Nguyen et al. Journal of Chromatography A 1635 (2021) 461743

Fig. 2. Evolution of selectivity (“challenging-sample”) with linear gradient 32–42 %B in 10 min (A,B,C,D), and with multi-isocratic (E,F,G,H) elution. The %B program for the multi-isocratic run was 33%B (0–2 min), 34.5%B (2.01–4 min), 35.3%B (4.01–6 min) and 35.9%B (6.01–10 min). F = 0.3 mL/min, 100 ×2.1 mm column, ε= 0.62, main peak (1, atezolizumab): S = 97.53, log k 0 = 34.68, minor peak (2): S = 101.2, log k 0 = 36.17.

4.2. ApplicationtotheseparationofintactmAbvariants

One of the mostchallenging tasks in the field of therapeutic protein analysis isthe RPLC separation ofprotein variants atthe intactlevel.Hence,we triedournewapproachforsuchchalleng- ingsample.Eculizumab(humanizedtherapeuticmAbproduct)has been selected asan example,since it containshydrophobic vari- ants. Wehavealreadymadeseveralattemptstoseparate thetwo mainvariantsofthismAbbylineargradientseparationsbuthave always failed. Our new method development approach (see de- scriptioninSection2.5.2.)hasbeenappliedtodefine theoptimal conditions for a multi-isocratic or negative step inserted multi- isocratic separation. LSS parameters were derived from two ini- tial lineargradients andthen ascreening multi-isocraticrun was

performedto estimate thebinding (parking)and elutingcompo- sitions (Supplementary Figure 2). It was found that 33.8%B mo- bilephaseprovidedsufficientlyhighretentionforbothcompounds to be parked at the column inlet (k1 = 856, k2 = 1062). Then, fourdifferent%Bcompositions wereemployedaseluting compo- sitions (37.3, 36.8, 36.3 and35.8 %B). Whatever the composition, thetwopeaksco-migrated withonlyaminordifferencebetween theirvelocities(SupplementaryFigure3,leftpanel).Thelowerthe

%B-duringtheelutionstep/hold-thehighertheselectivitywas, butsensitivitydecreaseddrasticallyduetobandbroadeningofthe macromolecules during isocratic migration with k ≥ 1. Baseline separationwasthereforenotfeasiblewithamulti-isocraticelution mechanism. A mobile phase composition of 36.3%B was selected asthefirstelutingsegmentoftheprogram-asitshowedagood

6

(8)

Fig. 3. Evolution of selectivity (“difficult-sample”) with multi-isocratic elution, when inserting a negative (“parking”) gradient step. Conditions and samples as de- scribed in Fig. 2 , except the %B program was: 33%B (0–2 min), 34.5%B (2.01–4 min), 35.3%B (4.01–5.5 min), 33%B (5.51–8 min) and 38%B (8.01–10 min).

compromise betweenselectivityandsensitivity–andthennegative gradientandholdingsegmentswereaddedwithanattempttoim- prove the separation.After theeluting segment, a 0.01 minlong steep negativeramp-from36.3%Bto33.8%B-wasaddedtostop the migration of the second peak andthis composition (33.8%B) washelduntil4min.Thenat4.01min,apositivestepwasadded to resetto36.3%Bandresumeelutionofthemoreretainedpeak.

Thepurposewastoelutetheentirepeakofthelessretainedcom- pound during the first eluting step while parking the more re- tainedcompoundatagivenmigrationdistance(negativestep),and thenfinallytoelutetheparkedcompoundbyreturningtotheelu- tioncondition(lastpositivestep).Torealizethis,thelengthofthe first eluting isocratic step (36.3%B) has to be optimized. Various holding times were tried ranging from 0.25to 0.5min (Supple-

Fig. 4. Separation of intact eculizumab variants with a linear gradient (A), multi- isocratic elution mode technique (B) and multi-isocratic mode including negative gradient step technique (C). F = 0.5 mL/min, column: BioResolve RP 50 ×2.1 mm, mobile phase A: water + 0.1% TFA, mobile phase B: acetonitrile + 0.1% TFA, tem- perature = 85 0C, main peak (1, eculizumab): S = 129.46, log k 0= 46.69, minor peak (2): S = 124.42, log k 0= 45.08. The %B program for the negative step inserted multi-isocratic run (panel C) was 33.8%B (0–2 min), 36.3%B (2.01–2.4 min), 33.8%B (2.4–3.5 min) and 36.3%B (3.51–6 min). Red dashed lines correspond to the sum of column dead time and gradient delay time. The blue curves (%ACN) were corrected for the total (system + column) delay time.

mentaryFigure3,middlepanel).Inthecaseofholdingtimesthat weretooshort,peak1wasnotcompletelyeluted,whileincaseof atoolongholdingtime,afractionofpeak2elutedtogetherwith peak 1. At the end, 0.4 min was found to be the optimal hold- ing time since the entirepeak of thefirst compound waseluted withoutallowing throughafractionofthesecondcompound.For onelaststepofoptimization,theselectivitybetweenpeaks1and 2 was changed by adjusting the length of the negative isocratic parkingsegment(SupplementaryFigure3,rightpanel).Ultimately, addinganegativegradientandholding stepintoamulti-isocratic program enabledusto achieve arbitraryselectivity betweencrit- ical peakpairs, which wasnot possiblewitha linear gradientor multi-isocratic elutiontechnique. Fig. 4 showsthe comparisonof experimentally measured chromatogramsobtainedby performing an optimized linear gradient (A), multi-isocratic elution (B) and negativestepinsertedmulti-isocraticelutionmode(C)separation.

It isworth mentioning that the shapeof peak 1is morefronted withtheoptimizednegativestepgradient comparedtothelinear gradientelution.Thereasonisprobablythatthe pre-peakvariant (minorpeakelutinginthefrontpartofthemainpeak1)isbetter separatedfrompeak1 (Fig.4 AvsC)since bothpeakseluteiso- craticallywithverysmallkvalues.Whatisimportanttosayisthat

(9)

S. Fekete, A. Murisier, J.M. Nguyen et al. Journal of Chromatography A 1635 (2021) 461743

Fig. 5. Separation of anti-citrinin mAb and carrier protein (BSA) with a linear gra- dient (A), multi-isocratic elution mode technique (B) and with multi-isocratic mode including negative gradient step technique (C). F = 0.4 mL/min, column: PS-DVB, 15 ×2.1 mm, mobile phase A: water + 0.1% FA, mobile phase B: acetonitrile + 0.1%

FA, temperature = 80 0C, peak 1 (mAb): S = 54.48, log k 0= 18.87, peak 2 (BSA):

S = 50.87, log k 0 = 18.29. The %B program for the negative step inserted multi- isocratic run (panel C) was 28%B (0–0.5 min), 34.6%B (0.51–0.65 min), 28%B (0.66–

1.05 min), 37%B (1.06–1.5 min) and 50%B (1.51–2 min). Red dashed lines correspond to the sum of column dead time and gradient delay time. The blue curves (%ACN) were corrected for the total (system + column) delay time.

nosignallossisobserved,theentirequantityofsolute1elutesin peak1–itissupportedbySupplementaryFigure3.

4.3. ApplicationtotheseparationofanintactmAbfromacarrier protein

RPLC analysis of mAbs often suffers from the loss of re- covery and, in addition, solutes may undergo some non-desired on-column aggregation or degradation [26]. To prevent intact mAbsfromself-aggregationandnon-specificbinding,albumin(like bovine serum albumin,BSA) can be added into the sample asa so-calledcarrierprotein.Thecarrierproteinhelpstoimprovepro- tein recoveryandcanbeespeciallyusefulwhenverylowconcen- trationsof proteinsneedto beanalyzed [27,28]. However, itmay happen thattheseparationofthemAbofinterestandthecarrier proteinischallenging.

In the example reported in Figs. 5 A and 5 B, we could not achieve appropriate resolution betweenan anti-citrinin mAb and BSApeak,neitherwithlineargradientnorwithmulti-isocraticelu- tion separationtechniques.Weagainfound thatoncethe slightly lessretainedmAbstartsits migration,BSAalsobeginstomigrate.

Despite the large difference between the molecular structures of

thetwoproteins,theypossessverysimilarretentionmodelparam- eters(anti-citrininmAb:S=54.48,logk0=18.87,BSA: S=50.87, logk0= 18.29) (as measured for the utilized PS-DVB stationary phase and selected mobile phase). The only chance to separate thesecompoundswastoaddanegativesegmentimmediatelyafter theelutionofthemAbpeakinordertostop themigrationofthe BSA, andthereby improve overall selectivity. The sameoptimiza- tionprocedurewasappliedasinSection4.2.Fortheinitialbind- ingstep,mobilephasecompositionwassetto28%B(k1=4.1103, k2=1.1104)andheldfor0.5min.Switchingto34.6%Belutedboth themAbpeak (k1 ~ 0.9andBSA (k1~ 3.9)withnoticeablybroad- enedpeakshape.Addinganotherpositivestep(37%B)at1minre- sulted inthe prompt elution of the remaining portion of BSA in asharp (compressed)peak.Thisinteresting behavior isportrayed inFig.5B,whereBSAwassplitintotwo peaks;thefirstfraction experienced isocraticelution,while thesecond fractionelutedby a very steep gradient segment (24%B/min). Finally,going back to 28%B after holding the elution segment (34.6%B) until 0.15 min (0.65minintheelution program)madeitpossible tocompletely elutethe mAbpeak and topark theentirety ofBSA. At the1.05 minutemark,themobilephasewasthenchangedto37%Btoyield immediateelutionofBSAinasinglesharppeak(Fig.5C).

4.4. Robustnessofthemeasurements

Since very minor changes need to be set in the %B program when running multi-isocratic and negative step inserted multi- isocratic separations, it is essential to consider the repeatability of the technique. To this end, five consecutive replicates of the eculizumab sample were injected and the same replicates were re-injected again on the next day. The relative standard devia- tions (RSD) of the retention times obtained for the two peaks over 2dayswere lower than 0.1%. Consequently,the resultssug- gest that the accuracy and precision of current modern UHPLC instrumentation–atleast in termsof mobile phase delivery - are sufficient to perform these negative step inserted multi-isocratic separations.

Inmanycases,proteinsamplesneed onlybeseparatedwitha change inorganic modifier contentof no greater than 5–10%. As a result,it mightbe preferredto preparemobilephasesA andB aspremixedsolvents(e.g.A:70%aqueous+ 30%organicsolvent, and B: 50% aqueous + 50% organic solvent). The multi-isocratic separations(includingthosewithnegativeslopesegments)canbe performedwithbroaderabsoluterangesforpumpoperation.This contributestoimprovetherepeatabilityofthemeasurements.

5. Conclusion

.SeparationoftherapeuticproteinsbyRPLCismostcommonly performedbyusinglineargradients.However,inmanycases,com- monlineargradientsdonotoffersufficientselectivity andresolv- ing power. For this reason, a recently developed multi-isocratic elution mode should be considered to enhance theseparation of challengingsamples. Evenstill, unsatisfactorylevels of resolution might be encountered in protein RPLC separations. In thisstudy, we explored the use of a negative gradient segment along with a “holding” isocratic segment to beneficially affect critical pairs of peaks within a protein sample. Withthe insertion of a nega- tive gradient, themigration of themore retained compoundwas stopped(“parked”)somewherealong thecolumn,whilealessre- tainedcompoundwassuccessfullyeluted.Notethatthisapproach onlyworksforlargesolutes,whichapproachan“on-off” typeelu- tion behavior. Forthe separation ofchallenging protein samples, we suggest the combination of a so-called (1) binding isocratic segment with (2) elutingshort steep gradients and holding seg- mentsalongwith(3)“parking” segmentsconsistingofshortsteep 8

(10)

negative gradients and holding steps. Please note that true “on- off” mechanismdoesnotexist,largesolutesjustapproachthisbe- havior. In ourpractice,we saw that solutes possessingmolecular weights of MW > 20 - 25 kDa are close enough to a retention behavior whichcanbenefita lotfromthe multi-isocraticandthe negativegradientslopemethods.

The theoretical benefits of the negative segmented multi- isocratic elution mode have been demonstrated in comparison with common linear and optimized multi-isocratic separations.

Two real applications have also been developed and we have proven their utility and the significance of this new separation mode.

We have also reported a fast and efficient optimization pro- cedure to develop multi-isocratic and negativesegmentedmulti- isocraticseparations.Theproposedprocedureincludes(1)twoini- tial lineargradients, followed by a (2) “screening” multi-isocratic run and one (or few) (3) “stretched” multi-isocratic runs. In the end,thelengthoftheelutingandparkingisocraticsegmentsneed to be empirically determined.The total time ofmethod develop- mentonlytakesafewhours.

This negativesegmentedmulti-isocratic elution mode canpo- tentially be applied to improve the separation of notoriously heterogeneous biopharmaceutical samples (e.g. intact mAb vari- ants,Fc-fusionproteins,bispecific-mAb,antibodymixtures,orADC species). Moreover, auniformpeak distribution (equidistant band spacing)canbeachievedifsodesired.

DeclarationofCompetingInterest

Theauthorsdeclarethattheyhavenoknowncompetingfinan- cialinterestsorpersonalrelationshipsthatcouldhaveappearedto influencetheworkreportedinthispaper.

CRediTauthorshipcontributionstatement

Szabolcs Fekete: Writing - original draft, Methodology, Inves- tigation.AmarandeMurisier:Conceptualization,Writing-original draft.JenniferM.Nguyen:Writing-review&editing.MatthewA.

Lauber: Resources, Writing - review & editing. Davy Guillarme:

Supervision,Writing-review&editing.

Acknowledgements

The authorswish to thankJean-Luc Veutheyfromthe Univer- sityofGenevaforfruitfuldiscussions.

Supplementarymaterials

Supplementary material associated with this article can be found,intheonlineversion,atdoi:10.1016/j.chroma.2020.461743. References

[1] T. Jupille , L. Snyder , I. Molnar , Optimizing multilinear gradients in HPLC, LCGC Europe 15 (2002) 596–601 .

[2] V. Concha-Herrera, G. Vivó-Truyols, J.R. Torres-Lapasió, M.C. García-Alvarez- Coque, Limits of multi-linear gradient optimisation in reversed-phase liquid chromatography, J. Chromatogr. A 1063 (2005) 79–88, doi: 10.1016/j.chroma.

20 04.12.0 01 .

[3] V.S. Joshi, V. Kumar, A.S. Rathore, Role of organic modifier and gradient shape in RP-HPLC separation: analysis of GCSF variants, J. Chromatogr. Sci. 53 (2015) 417–423 http://dx.doi.org/10.1093/chromsci/bmu222 .

[4] V.S. Joshi, V. Kumar, A.S. Rathore, Rapid analysis of charge variants of mono- clonal antibodies using non-linear salt gradient in cation-exchange high per- formance liquid chromatography, J. Chromatogr. A 1406 (2015) 175–185 http:

//dx.doi.org/10.1016/j.chroma.2015.06.015 .

[5] B. Bobály, G.M. Randazzo, S. Rudaz, D. Guillarme, S. Fekete, Optimization of non-linear gradient in hydrophobic interaction chromatography for the analyt- ical characterization of antibody-drug conjugates, J. Chromatogr. A 1481 (2017) 82–91 http://dx.doi.org/10.1016/j.chroma.2016.12.047 .

[6] L. Moruz, P. Pichler, T. Stranzl, K. Mechtler, L. Kall, Optimized Nonlinear gra- dients for reversed-phase liquid chromatography in shotgun proteomics, Anal.

Chem. 85 (2013) 7777–7785 http://dx.doi.org/10.1021/ac401145q .

[7] E. Tyteca, A. Liekens, D. Clicq, A. Fanigliulo, B. Debrus, S. Rudaz, D. Guillarme, G. Desmet, Predictive elution window stretching and shifting as a generic search strategy for automated method development for liquid chromatography, Anal. Chem. 84 (2012) 7823–7830 https://doi.org/10.1021/ac301331g . [8] E. Tyteca, K. Vanderlinden, M. Favier, D. Clicq, D. Cabooter, G. Desmet, En-

hanced selectivity and search speed for method development using one- segment-per-component optimization strategies, J. Chromatogr. A 1358 (2014) 145–154 http://dx.doi.org/10.1016/j.chroma.2014.06.097 .

[9] S. He, Li S, J. Yang, H. Ye, S. Zhong, H. Song, Y. Zhang, C. Peng, A. Peng, L. Chen, Application of step-wise gradient high-performance counter-current chromatography for rapid preparative separation and purification of diterpene components from Pseudolarix kaempferi Gordon, J. Chromatogr. A 1235 (2012) 34–38 http://dx.doi.org/10.1016/j.chroma.2012.01.040 .

[10] D.V. Camper, R.E. Viola, Fully automated protein purification, Anal. Biochem.

393 (2009) 176–181 http://dx.doi.org/10.1016/j.ab.2009.07.009 .

[11] D.L. Morris, J.N. Sutton, R.G. Harper, A.T. Timperman, Reversed-phase HPLC separation of human serum employing a novel saw-tooth gradient: toward multidimensional proteome analysis, J. Proteome Res. 3 (2004) 1149–1154 https://doi.org/10.1021/pr049901n .

[12] C. Guimei, "Evaluating the effective peak capacity of a saw-tooth gradient for reverse-phase high performance liquid chromatography separation of proteins and peptides" (2007). Graduate theses, dissertations, and problem reports.

1819. https://researchrepository.wvu.edu/etd/1819

[13] B. Durner, T. Ehmann, F.M. Matysik, High-resolution polymer high performance liquid chromatography: Application of a saw tooth gradient for the separation of various polymers, J. Chromatogr. A 1587 (2019) 88–100 https://doi.org/10.

1016/j.chroma.2018.11.075 .

[14] R.S. Blanquet, K.H. Bui, D.W. Armstrong, Mechanistic considerations on the re- versed phase liquid chromatographic separation of proteins, J. Liq. Chrom. 9 (1986) 1933–1949 https://doi.org/10.1080/01483918608078753 .

[15] S. Fekete, A. Beck, J.L. Veuthey, D. Guillarme, Proof of concept to achieve in- finite selectivity for the chromatographic separation of therapeutic proteins, Anal. Chem. 91 (2019) 12954–12961 https://doi.org/10.1021/acs.analchem.

9b03005 .

[16] S. Fekete, H. Ritchie, J. Lawhorn, J.L. Veuthey, D. Guillarme, Improving selec- tivity and performing online on-column fractioning in liquid chromatography for the separation of therapeutic biopharmaceutical products, J. Chromatogr. A 1618 (2020) 460901 https://doi.org/10.1016/j.chroma.2020.460901 .

[17] V. Pepermans, J. De Vos, S. Eeltink, G. Desmet, J. Chromatogr. A 1586 (2019) 52–61 https://doi.org/10.1016/j.chroma.2018.11.078 .

[18] J. De Vos, G. Desmet, S. Eeltink, J. Chromatogr. A 1360 (2014) 164–171 https:

//doi.org/10.1016/j.chroma.2014.07.072 .

[19] J. De Vos, G. Desmet, S. Eeltink, J. Chromatogr. A 1455 (2016) 86–92 https:

//doi.org/10.1016/j.chroma.2016.05.046 .

[20] M.A . Stadalius, M.A . Quarry, L.R. Snyder, Optimization model for the gradient elution separation of peptide mixtures by reversed-phase high-performance liquid chromatography: Application to method development and the choice of column configuration, J. Chromatogr. 327 (1985) 93–113 https://doi.org/10.

1016/S0021-9673(01)81640-X .

[21] S. Fekete, S. Codesido, S. Rudaz, D. Guillarme, K. Horváth, Apparent efficiency of serially coupled columns in isocratic and gradient elution modes, J. Chro- matogr. A 1571 (2018) 121–131 https://doi.org/10.1016/j.chroma.2018.08.002 . [22] A. Vaast, E. Tyteca, G. Desmet, P.J. Schoenmakers, S. Eeltink, Gradient-elution

parameters in capillary liquid chromatography for high-speed separations of peptides and intact proteins, J. Chromatogr. A 1355 (2014) 149–157 https://doi.

org/10.1016/j.chroma.2014.06.010 .

[23] L.R. Snyder , Gradient elution, in: C. Horvath (Ed.), HPLC: Advances and Per- spectives, vol. 1, New York, Academic Press, 1980, pp. 208–316 .

[24] L.R. Snyder, M.A. Stadalius, M.A. Quarry, Gradient elution in reversed-phase HPLC-separation of macromolecules, Anal. Chem. 55 (1983) 1412A–1430A https://doi.org/10.1021/ac0 0264a0 01 .

[25] L.R. Snyder , K.M. Goodings , F.E. Regnier , in: HPLC of Biological Molecules, Methods and Applications, Marcel Dekker, New York, 1990, p. 231 .

[26] S. Fekete, S. Rudaz, J.L. Veuthey, D. Guillarme, Impact of mobile phase tempera- ture on recovery and stability of monoclonal antibodies using recent reversed- phase stationary phases, J. Sep. Sci. 35 (2012) 3113–3123 https://doi.org/10.

10 02/jssc.20120 0297 .

[27] J.E. Battersby, B. Snedecor, C. Chen, K.M. Champion, L. Riddle, M. Vanderlaan, Affinity–reversed-phase liquid chromatography assay to quantitate recombi- nant antibodies and antibody fragments in fermentation broth, J. Chromatogr.

A 927 (2001) 61–67 https://doi.org/10.1016/S0021-9673(01)01108-6 . [28] Y. Alelyunas, H. Shion, M. Wrona, High sensitivity intact monoclonal antibody

(mAb) HRMS quantification. Waters Application 720 0 06222en 2018.

Références

Documents relatifs

Given a TCP port number pair, it returns a character string which identifies the owner of that connection on the server’s system.. The Identification Protocol was formerly

Although ISO 2022 specifies special additional escape sequences to indicate the use of revised character sets, it is suggested here not to make use of this special escape

As IBGP learned routes are reflected, it is possible through mis- configuration to form route re-distribution loops.. The Route Reflection method defines the following

In keeping with the principles of rough consensus, running code, architectural integrity, and in the interest of ensuring the global stability of the Internet, the IAB

res de lois rhéologiques, d’examiner les variations de cette résistance en fonction de la pression négative sur un chemin d’humidification à partir de l’OPM

alspgrad: alternating non-negative least squares using the projected gradient method for each sub-problem (Section 4.1).. lsqnonneg: Using MATLAB command lsqnonneg to solve m

For standard complex Hessian equations on compact K¨ ahler manifolds, this C 2 estimate was obtained by Hou-Ma-Wu [22].. It is worth mentioning that such a C 2 estimate combined with

A new phase space localization technique with application to the sum of negative eigenvalues of Schrödinger operators.. Annales scientifiques