Impact of stationary-phase pore size on chromatographic performance using oligonucleotide separation as a model

http://www.diva-portal.org

This is the published version of a paper published in Journal of Chromatography A.

Citation for the original published paper (version of record):

Bagge, J., Enmark, M., Lesko, M., Lime, F., Fornstedt, T. et al. (2020)

Impact of stationary-phase pore size on chromatographic performance using

oligonucleotide separation as a model

Journal of Chromatography A, 1634: 1-10

https://doi.org/10.1016/j.chroma.2020.461653

Access to the published version may require subscription.

N.B. When citing this work, cite the original published paper.

Permanent link to this version:

Journal of Chromatography A 1634 (2020) 461653

ContentslistsavailableatScienceDirect

Journal

of

Chromatography

A

journalhomepage:www.elsevier.com/locate/chroma

Impact

of

stationary-phase

pore

size

on

chromatographic

performance

using

oligonucleotide

separation

as

a

model

Joakim

Bagge

a

_,

_Martin

_Enmark

a

_,

_Marek

_Le

_´sko

a

_,

_Fredrik

_Limé

b

_,

_Torgny

_Fornstedt

a,∗

_,

Jörgen

Samuelsson

a,∗

a Department of Engineering and Chemical Sciences, Karlstad University, SE-651 88 Karlstad, Sweden b Nouryon, SE–445 80 Bohus, Sweden

a

r

t

i

c

l

e

i

n

f

o

Article history: Received 21 July 2020 Revised 7 October 2020 Accepted 26 October 2020 Available online 29 October 2020

Keywords: Pore size Ion-pair RPLC Oligonucleotides Gradient elution Adsorption isotherm Surface area

a

b

s

t

r

a

c

t

Acombinedexperimentalandtheoreticalstudywasperformedtounderstandhowtheporesizeof pack-ingmaterialswithpores60–300 ˚Ainsizeaffectstheseparationof5–50-mer oligonucleotides.Forthis purpose,wedevelopedamodelinwhichthesolutesweredescribedasthinrodstoestimatethe accessi-blesurfaceareaofthesoluteasafunctionoftheporesizeandsolutesize.First,ananalytical investiga-tionwasconductedinwhichwefoundthattheselectivityincreasedbyafactorof2.5whenseparating 5-and15-meroligonucleotidesusingpackingwith300 ˚Arather than100 ˚A pores.We complemented theanalyticalinvestigationbytheoreticallydemonstratinghowtheselectivityisdependentonthe col-umn’saccessiblesurfaceareaasafunctionofsolutesize.Inthepreparativeinvestigation,wedetermined adsorptionisothermsforoligonucleotidesusingtheinversemethodforseparationsofa9-anda10-mer. Wefoundthatpreparativecolumnswitha60 ˚A-pore-sizepackingmaterialprovideda10%increasein productivityascomparedwitha300 ˚Apackingmaterial,althoughthesurfaceareaofthe60 ˚Apacking isasmuchasfivetimelarger.

© 2020TheAuthor(s).PublishedbyElsevierB.V. ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/)

1. Introduction

High-performanceliquidchromatography(HPLC)hasbecomea standard technique forseparating biopharmaceutical compounds, with ion-pair reversed-phase liquid chromatography (ion-pair-RPLC) commonly being used [1,2]. Several papers have studied the separation of nucleic acids using ion-pair-RPLC [3–6]. A use-ful strategy to improve separation performance inRPLC is to al-ter thecolumn properties.The columnproperties that can affect the separation include the surface chemistry, particle size, pore size, andcolumndimensions, whichcan be scaled to enable im-proved separation efficiency or higher sample loads [7–10]. We recently published a study on the separation of 5-20 long ho-momer oligonucleotides wherewe compared the performance of three alkylphases(C4, C8,C18)andone arylphase(phenyl). Our findings showed that C4 was preferable in terms of selectivity when separatingnativeoligonucleotides,using triethylammonium acetate(TEtAA)asion-pairingreagent[10].

∗_{Corresponding authors.}

E-mail addresses: torgny.fornstedt@kau.se (T. Fornstedt),

jorgen.samuelsson@kau.se (J. Samuelsson).

Theporesizeofthepackingmaterialscanbe usedtoseparate biomoleculesbytheir size,inatechniqueknownassize-exclusion chromatography(SEC).Historically,chromatographicseparation us-ingaporousmatrixwasgreatlyinfluencedbythepioneeringwork ofPorathandFlodin[11],inwhichbiomoleculeswereseparatedby their sizeusingcross-linked dextrangels. Fundamentalaspects of theprocess werelater clarifiedby Giddingsetal.,who presented an equilibrium theory for adsorption chromatography in which thepartition ofasolute betweentheinter-particle spaceandthe porous space within the particles was in constant equilibrium [12]. Later, Casassa introduced stochastic models for expressing thenon-accessibleporevolumeofasolute ofdefinedgeometrical form[13,14]. Theseessential models havebeen acknowledged in otherworks[15–17].Foroligonucleotides andmanyothercharged polymers, themolecular geometry dependsstronglyon the envi-ronmentalconditions. In thecase ofion-pairing chromatography, the ion-pairing reagents will increase the ionic strength of the medium. Simet al.showedthat the radius ofgyration of homo-mericoligonucleotidesdecreaseswiththeincreasingionicstrength ofthemedia[18].Forexample,thepresenceof125mMNaCl re-sulted in a radius of 19.1 ˚A fora 16-mer oligonucleotide, versus 22 ˚Awith25 mM NaCl. Studies have shown that single stranded oligonucleotides best can be described using a worm-like chain modeloramodifiedfreelyjointedchainmodel[19]withalength

https://doi.org/10.1016/j.chroma.2020.461653

that isalsodependenton theionicstrength,temperatureandpH ofthesolution.

In adsorption chromatography,totally porousparticles provide a large bindingsurfacearea per mass,the relationshipwithpore sizebeinginverselyproportional[20].Thelargersurfacearea pro-vides an increasednumberofbinding sites,increasing the reten-tion of the adsorbing solutes [21–24]. However, the accessibility ofthe surfacearea diminishes withtheincreasing sizeofthe so-lute molecule. From inverse size-exclusion measurements, Gritti andGuiochondemonstrated thatfora totallyporousC18 column withaporesizeof90 ˚A,theaccessiblesurfaceareawasonly40% forinsulinduetoexclusion,andthataporesizeof200–250 ˚Awas recommendedfortheseparation[25].Theuseoflargerporesizes wasalsoinvestigatedby Gétazetal., who examinedimpurity se-lectivityforinsulinandanunknowncrudepeptideonC18columns withporesizesof90–300 ˚A[8].Whilelargerporesizesproduced nobenefitsforoverloadedinjectionsofinsulin,aporesizeof300 ˚Aimproved theselectivityforthecrudepeptide.Theyalsofound that masstransfer increasedforlarger poresizes[8]. Sandsetal. characterized andinvestigated a totally porous C18 column with a pore size of100–300 ˚A [23],finding that the resolution gener-allydecreasedwithincreasingporesizeforsmallermoleculesand largersolutes(i.e.,mixturesofsmallerproteins)andthatthe reso-lutionimprovedwithincreasingporesize.

Theeffectofporesizeonresolutionofvariousnucleotide poly-mers such as double-stranded DNA [26], DNA oligonucleotides [6]andnon-coding RNA[27]haspreviously beeninvestigated.Of relevanceforthisstudyaretheresultsofCloseetal[6]who inves-tigated the useof80,150and400 ˚Aporesizeswhen separating various short oligonucleotides using superficially porous particles withtriethylammoniumacetatecontainingmobilephases.

Although important research has been done concerning oligonucleotide separation problems and pore size [6,26,27], no fundamentalsystematicinvestigationhasyetbeenperformedover alargeenoughrangeofsizesofsimilarsolutemoleculesand sep-arated with columns ofdifferent pore sizes, butwith everything elseidenticalinordertoobtainamorecompletefundamental un-derstanding which pore size to select for a certain analytical or preparativeseparationproblems.Ourgoalwiththisstudyisto per-formsuchstudiesonafirmphysicalchemicalbasisandtoprovide aclearconclusionfoundedonsolidchromatographicexperiments, theoryandsimulationsusingwell-acknowledgedmodels.

Thisstudyexperimentallyandtheoreticallyinvestigatesthe ef-fect of pore size on analytical as well as preparative chromato-graphicseparationmodesforalargerangeofoligonucleotidesizes previously investigated by us in a recent study [10]. Separation wascarriedoutwithion-pairchromatographyonbutylsilane(C4) reversed-phase columnswithporesizesof60–300 ˚A.Toestimate the accessible surfacearea of the oligonucleotides on the differ-entcolumns,ageometricmodelwasderived.Oligonucleotide lad-derswillbeusedtoexperimentallyinvestigatetheselectivity. Us-ing a model based on adsorption theory, the selectivity trends willbefundamentallyanalyzedandoverloadedpreparative separa-tionswillbeinvestigated.Theadsorptionisothermsofan oligonu-cleotidewillbedetermined onthe60and300 ˚Aporesize mate-rials andtheacquired parameterswill be used toinvestigatethe effect of pore size on chromatographic productivityby means of simulations.

2. Theory

2.1. Calculationofsurfaceareafromtheintra-particleequilibrium constant

The theory of the equilibrium partitioning of different sized molecules in porous media considers the ratio between the

so-lute’saccessiblevolume andtheaccessiblevolumeofthe station-ary phase. The ratio between thesevolumes is describedby the solute-specificequilibriumpartition coefficient,K_SEC .Solute parti-cles are often considered to be spherical; instead, here we have chosen toimplement amodelin whichthe solute isrepresented asa thinrodwitha length ofb. We stronglybelieve that a thin rod represents the oligonucleotide geometry better than does a sphere.Previousstudieshaveshownthatsinglestranded oligonu-cleotidesbehaveslikeachainwhichismorerigidthan for exam-plepolystyreneandpolyethylene[19]andthereforea rod geome-trywillbeanapproximationoftheactualshape.Itiswellknown that oligonucleotides can form secondary structures, for example hair-pinloops.Thepropensityoftheoligonucleotidetoformsuch structures depends onthe sequence and length andthe solution composition. In thecurrent study,we have chosen to investigate deoxythymidinehomomers whichhas an insignificantprobability offorminghair-pin loops,especially attheelevated temperatures andhighconcentration ofdenaturingalkylamineusedduring ex-periments[28].Wehaveapproximatedlengthoftherodby6.76 ˚A perunit[19].Fortherodmodel,K_SEC canbeexpressedas:

K_SEC =

⎧

⎪

⎨

⎪

⎩

1− 4 b π·a + b 2 2 a2 a≥ b −2 πcos−1



_a b



+1+ 3 b π·a



−a 2 b2 +1−_π4 b_·a − b2_{·co s} −1

(

a b

)

π·a 2 +b2 2 a2 a<b (1)

whereaisthestationaryphaseporeradius. Theavailablesurface area,henceforth,“accessiblesurfacearea” (A),iscalculatedby mul-tiplyingKSECbythetotalsurfaceareaofthecolumn.Eq.(1)is

jus-tifiedintheSupplementaryMaterial. 2.2. Adsorptionisothermsandgradienttheory

Ingradientelution,thefractionofmodifier(inthiscaseMeCN) changesover time. Linearsolventstrength (LSS)theory expresses theretention factor,k,ofa givencompound asa function ofthe elutionstrengthofthemodifier:

k=kw · exp

(

−S

φ

)

(2)

where

φ

isthevolumefractionoforganicmodifier,kwisthe

reten-tionfactorforsoluteinneatweakeluent(inthiscasewater),and Sisasolute-specificconstant thatdescribeshowtheretentionof thesolutesvarieswiththevolume fractionofmodifier.For gradi-entseparations,theretentionfactorcanbe derivedbyintegrating Eq.(2) over the time domain to obtain the following expression [29]:

k=_S tG

· t0 ·



φ

· ln

(

t0 · S·



φ

/tG · k0 +1

)

+

tD

t₀ (3)

where



φ

isthechangeinmodifieroverthegradientrun,t_G isthe gradienttime,k₀ istheretentionfactoratthestartofthegradient, t0isthecolumndeadtime,andtDisthedwelltime.Thefinalterm

ofEq.(3)correctsfortheisocraticholdduetothesystem-related extravolumefromtheeluentmixertothecolumninlet.

For overloaded injections, one needs to use an adsorption isotherm todescribe howthesolute interactswiththestationary phase.Anadsorptionisotherm describestherelationshipbetween theconcentrationofthesoluteinthemobilephase,C,andinthe stationaryphase,q.OnesuchsimplemodelistheLangmuirmodel [30]:

q=qS ₁₊KC_KC, (4)

whereqs is thesaturationcapacityandKis theassociation equi-libriumconstant.TheadsorptionisothermpresentedinEq.(4)can alsobemodifiedtocoverthemodifierdependency[30]:

q=qs ₁₊Kw _K· C· exp

(

−S·

φ

)

J. Bagge, M. Enmark, M. Le ´sko et al. Journal of Chromatography A 1634 (2020) 461653

whereKw correspondstotheassociationequilibriumconstant

de-termined when no organic modifier is present, i.e.,

φ

= 0. This adsorption model hasbeenapplied in other studiesto character-ize the adsorption of peptides on reversed-phase octadecylsilane phases[31,32].Theretentionfactorisproportionaltotheslopeof the adsorptionisotherm, inthiscasecalculatedastheproduct of the phase ratio, F, and the derivative with respect to the solute concentration

∂

q/

∂

C.

Bysubstituting k₀ andincorporating itintoEq.(3)we obtain:

k= tG

t0 · S·



φ

·

(

ln

(

t0 · S· qS · Kw ·



φ

/tG · F

)

− S·

φ

0

)

+

t_D t0

(6)

wheretheadditionofoneinthelogarithmicexpressionhasbeen neglected. Given that two solutes are separated under the same gradientconditions,Eq.(6)canbeexpressedintermsofthe selec-tivityandthereforebeapproximatedas:

α

=k_k2 1 ≈ ln

(

qS 2

)

+ ln



t0 · S2 · Kw ,2 ·φtG · F



−

φ

0 ln

(

qS 1

)

+ ln



t0 · S1 · Kw ,1 ·φtG · F



−

φ

0 (7)

The system-related t_D can also theoretically be neglected, as-sumingthatnoelutionwilloccuratthestartofthegradient.The valuesqS₁andqS₂willgenerallybeconsideredequalwhen

consid-eringcloselyrelatedcompounds. 3. Materialandmethods 3.1. Chemicalsandmaterial

Lyophilized phosphodiester oligonucleotides were obtained from Integrated DNA Technologies (Leuven, Belgium) and used withoutfurtherpurification.Themobilephasewaspreparedusing MilliQ waterwitha conductivityof18.2MΩcm–1 froma Milli-Q ultrapurewatersystem(Darmstadt,Germany)andHPLC-grade ace-tonitrile(MeCN) fromVWR(Radnor, PA,USA).Triethylammonium acetate (TEtAA) was prepared from triethylamine (≥99.5%) and aceticacid(≥99.5%),bothpurchasedfromSigma-Aldrich(St.Louis, MO, USA). Methyl mandelate (99%) was obtained from Sigma-Aldrich.

FourKromasilcolumnswithabutylsilane(C4)stationaryphase andpore sizesof60,100, 200,and300 ˚A were usedthroughout thisstudy. Allcolumnshadthe dimensions150× 3.0 mmanda particlesizeof5

μ

m;detailedcolumninformationispresentedin Supplementary Material Table S.1.All columnsused in thisstudy wereobtainedfromNouryon(Bohus,Sweden).

3.2. Instrumentation

Experimentswereconductedontwodifferentsystems.Thefirst system was an Agilent 1200 HPLC system (Agilent Technologies, Palo Alto, CA, USA) configuredwith abinary pump,an autosam-pler equipped witha 100-

μ

L sampleloop, a column thermostat, andadiode-arrayUV detector.Thesecondsystemwasan Agilent 1260 Infinity IIBio-inert LC systemconfiguredwitha degasser,a quaternarypump,amulticolumnthermostat,amultisamplerwith a100-μLsampleloop,andadiodearraydetector.Bothinstruments wereoperatedat50°Catasetflowrateof0.5mLmin–1 .

The flow rate changes during gradient elution were mea-sured usingaMiniCori-FlowM12 Coriolismass-flow meterfrom Bronkhorst High-Tech B.V. (Ruurlo, Netherlands). The measure-mentsweremadeattwolocations:1)downstreamfromthepump and2)directlyafterthemixer.

The column void volume was determined using uracil asthe dead-volume marker.The system’s extra-columnvolume was

de-termined by injecting diluted TFA samples with the column re-moved;thedeterminedvolumewassubsequentlysubtractedfrom all generated retentionanddead-volume data.The dwell volume wasdeterminedusingamobilephasesimilartothat usedforthe screeningexperimentbutwiththeadditionof0.15v%TFAintheB phase.Twolineargradients withdifferentgradientslopes(i.e.,30 to90%B in30minand30to 80%Bin50min) andwithout any isocraticholdswereused.

3.3. Procedures

3.3.1. Oligonucleotidesampleandeluentpreparation

Stock oligonucleotide solutions were prepared inMilliQ water to a final concentration of 10 mg mL–1_{, followed by brief heat}

treatmentat50°C.Thefollowingnamingconventionisusedforthe oligonucleotides:forexample,10-merdeoxythymidine monophos-phatewiththesequence5’-TTTTTTTTTT-3’iscalledT10.AT5–T50 sample for analytical screening was prepared from T5, T10,T15, T20, T25, T30, T40, and T50 stock solutions, diluted with MilliQ water to a final concentration of 0.5 mgmL–1 for each oligonu-cleotide.Foroverloadedinjections,aT10samplewaspreparedby diluting stocksolutionwith MilliQwater toa final concentration of5mgmL–1 .

AlloligonucleotideseparationswerecarriedoutwitheluentsA andBconsisting of3.3and16.7(v/v%)MeCN, respectively,with theproportionsbeingdeterminedbymass.Tobotheluents,TEtAA was added to a final concentration of 100 mM. All experiments were performed at 50°C which minimizes the risk that an arbi-traryoligonucleotidewouldformahair-pinloop.Since homopoly-mersareusedinthisstudy,theycannotformhair-pinloopsatthe investigatedtemperature.

3.3.2. Determinationofadsorptionisothermparametersformethyl mandelateinisocraticelutionmode

Theadsorptionisothermparametersweredeterminedusingthe elutionby characteristicpointsmethodinslopemode,previously usedtoinvestigatetheadsorptionofthismodelcompound,methyl mandelate [33–35].ALangmuir adsorptionisotherm, Eq.(4),was usedforfittingadsorptionparameters.Theoverloadedelution pro-files were generated with injections of 150 mM methyl mande-late using 20 (v/v %) MeOH in MilliQ water as the eluent; 900-μLfull-loopinjectionswere usedforthe200 and300 ˚Acolumns and900-μLpartial-loop injections (cutinjections) forthe 60and 100 ˚Acolumns.Manualinjections wereperformedusing a Rheo-dyne 7725 injector (Rheodyne, Cotati, CA, USA) with a 4980-μL loop andinjectionvolumesof2700μL(100 ˚A column)and3600 μL (60 ˚A column). For more details about how these injections were conducted, see Samuelsson andFornstedt [33]. The adsorp-tionisothermparametersarepresentedinFig.2.

3.3.3. Analyticalseparationof5–50-meroligonucleotides

Twodifferentanalyticalgradientexperiments wereconducted. The first set of experiments was conducted usingthe same gra-dient (i.e., G1) on all columns (see Table 1); the resulting chro-matograms are presented in Fig. 3a–d. In the second set of ex-periments,thegradientwasalteredsothatsimilarretentiontimes wereobservedonallcolumns.GradientsG2,G1,G3,andG4 were used on the 60, 100, 200, and300 ˚A columns,respectively. The resulting chromatograms for the normalized retention are pre-sentedin Fig.3e–h. The correspondingselectivity is presentedin Fig.4forthesamegradient(Fig.4a)andforthenormalized reten-tion(Fig.4b).

3.3.4. OverloadedinjectionsofT10andsimulations

Overloadedelutionprofilesforallcolumnsweredeterminedby injecting16and64μLof5mgmL–1 concentratedT10,carriedout

Table 1a-b

Summary of gradient specifications used for analytical and overloaded experiments. a) Analytical elution, T5–T50

Gradient

abbreviation t G [min]

Column pore

size [ ˚A] Gradient start [v% MeCN] Gradient slope [v% MeCN min –1 ]

G1 70 all 7.3 0.08 G2 70 60 6.8 0.09 G3 100 200 7.3 0.07 G4 120 300 7.3 0.05 b) Overloaded elution, T10 G5 22 60 8.8 0.09 G6 22 100 8.5 0.09 G7 22 200 7.7 0.09 G8 22 300 7.1 0.09 G9 17 300 7.1 0.18 G10 44 300 7.1 0.05 G11 18 60 8.8 0.18 G12 22 60 8.8 0.11 G13 30 60 8.8 0.05

on theAgilent1200 HPLCsystem (seeFig. 6).The gradientslope wassetto0.09v%MeCNmin–1 andtheinitialfractionofmodifier wasadjustedto elutesoluteswithin approximately20minonall columns(seegradientsG5–G8inTable1).

TheadsorptionisothermsofT10on60and300 ˚Acolumnswere determined usingtheinversemethod.First,the concentrationsof T9andT10wereestimatedby injecting5μLof0.5mgmL–1 T10 using gradientG8. Wethen assumedthat thetotal peakareas of T9 and T10constituted the total concentrations.Overloaded elu-tion profiles were generatedby injecting 8, 16,32, and64 μL of 5 mg mL–1 T10using gradients G8 to G13. The chromatographic data were fitted using the equilibrium-dispersive model andthe modifier-dependentadsorptionisotherminEq.(5);thesystemwas numericallysolvedusingorthogonalcollocationonfiniteelements. Theparametersweredeterminedusingapreviouslydescribed ap-proach[36,37].Theexperimentalresultsandmodelpredictionsare comparedinFig.7for8-and16-μLinjectionsof5mgmL–1 ofT10 usinggradientsG5andG8for60and300 ˚Acolumns,respectively. 4. Resultsanddiscussion

The results are presented in two sections. In section 4.1, in which thefocus ison analytical-scale separations,the adsorption isothermparametersformethylmandelateonallcolumnsare pre-sented and discussed. We continue by discussing the analytical-scale separationsofT5–T50oligonucleotidesintermsofretention and selectivity to find out how selectivity is dependent on pore size and how this is related to the change in accessible surface area. In Section 4.2, we consider overloaded preparative separa-tions.Here,theresultsofexperimentalandnumericallysimulated overloaded injectionsofT10arepresentedtoinvestigatehowthe pore size/accessible surface affects the productivity. All units of lengthforoligonucleotidesreferstotheirpolymericchainlengths, approximatedasarod[13,14]unlessotherwisestated.

4.1. Analytical-scaleinvestigations

The larger the oligonucleotide, the more it will be excluded from inter-particle space. Fig. 1 presents the accessible surface area (A) foreach poresize, calculatedusingEq.(1) asafunction of oligonucleotide length. A smallerpore size initially provides a larger accessible surface area, but the decrease in area, i.e., the slope ofthefunction, isinitially muchhigher.Witha largerpore size,theaccessiblesurfacearearemainsmuchmoreconstantwith increasing solutelength,atleastinourinvestigatedrange.Weare mostly interestedinassessing thismodelwithin the rangeup to

Fig. 1. Predicted accessible surface area ( Eq. 1 ) for C4 Kromasil columns with pore sizes of 60, 10 0, 20 0, and 30 0 ˚A as a function of solute length. The vertical (dashed) lines indicate the rod model lengths: T5, T10 and T20.

the longest studied oligonucleotide, T50, which corresponds to a maximumlengthof338 ˚A.

Weexpectthelargeraccessiblesurfacearea,asfoundin pack-ingwithsmallerpores,torelatetoanincreaseinmonolayer satu-rationcapacitythataffectsthestrengthoftheadsorption(i.e., as-sociationequilibriumconstant)notatalloronlyslightly.To inves-tigatethis,theadsorptionisothermformethylmandelate(asmall uncharged probe) wasdetermined using elution by characteristic pointsforallcolumns.Here,weassumethatthemethylmandelate molecule issufficiently smalltohaveaccessto theentiresurface area of all columns C4 layers. The monolayersaturation capacity (qs )andequilibriumconstant(K)asafunctionofavailablesurface area(A)arepresentedinFig.2,whiletheadsorptionisotherm pa-rametersarefoundinSupplementaryMaterialTableS.2.InFig.2a, onecanseethatthesaturationcapacityincreaseslinearlywith in-creasing surfacearea. In other words, the smaller the pores, the largerthesurfaceareaandsaturationcapacity.Ontheotherhand, theassociationequilibriumconstant(seeFig.2b)remainsmoreor lessunchangedovertherangeofinvestigatedcolumns.Theresults

J. Bagge, M. Enmark, M. Le ´sko et al. Journal of Chromatography A 1634 (2020) 461653

Fig. 2. Adsorption isotherm parameters ( Eq. 4 ) determined for methyl mandelate on the 60, 10 0, 20 0, and 30 0 ˚A pore size columns and these parameters relationship with the (left) the columns monolayer saturation capacities ( q s ) and with (right) the columns association equilibrium constants ( K ).

Fig. 3. Separation of the T5–T50 ladder on the 60, 100, 200, and 300 ˚A columns with (a-d) identical gradient elution conditions and with (e)–(h) adjustments of the initial volume fraction of modifier to normalize the retention times to approximately 60 minutes. For the exact gradients operational conditions, see Table 1 : (a)–(d) and (g) G1; (e) G4; (f) G3; (h) G2.

strongly suggest that the columns’ saturation capacity increases withtheaccessiblesurfaceareaandthatnoorverysmallchanges should be observedintheequilibriumconstant underunchanged eluotropicconditions.

The analyticalseparation oftheT5–T50 oligonucleotideladder produced the expected retention order, indicating increased re-tention withdecreased pore size (Fig. 3a–d). The chromatograms show theseparationwitha70-mingradientinwhichallcolumns provided acceptable baseline separation for n – 1 up to about n = 20.The 60 ˚A columnprovided almost baseline resolution of shorter andlonger oligonucleotides up to T23,i.e. beingbaseline

resolvedfromT22andT24(T23elutesataround55minutes).The 100 ˚Acolumn gave similar separationbaseline resolution upto T28 (elutes at around 49 minutes) and the 200 ˚A columnup to T30 (elutes at around 39minutes). Broad and tailingpeaks were ob-served forT20 andlonger oligonucleotideson the300 ˚Acolumn. Incontrast,Wagneretalcouldshowthatthepeakwidthand res-olution betweenlarge double strandedDNA decreased when in-creasingtheporesizefrom400to1000 ˚A[26].Ourobservedband profiletailing wasunexpectedandfurther experimentswouldbe required to elucidate its origin, for example by using other ion-pairingreagentsandsolutes.However,themainfocusofthisstudy

Fig. 4. The impact of pore size on selectivity between pairs of oligonucleotides and the relationship with accessible surface area ( Fig. 1 ). In (a) the selectivity is calculated from the chromatograms in Fig. 3 a–d, while in (b) and (d) the selectivity is calculated from Fig. 3 e–h. In (c), the results can be compared with the difference in accessible surface area between pairs of oligonucleotides.

is to investigate retentionas well as theselectivity and not effi-ciencyandresolution.Therefore,thisunexpectedpeaktailingwill probablynotaffecttheconclusionconcerningselectivityor reten-tion.Byinspectingthechromatograms,wecanseethatonlythe60 ˚Acolumnusesthewholeseparationgradient.Bychangingthe gra-dientso thattheretentiontime ofT50 becomessimilar (60min) on all columns,all columns cantake full advantage ofan identi-calseparationwindow,asshowninFig.3e–h.Notethattheflatter gradientslopesforthe200and300 ˚Acolumnsgenerateimproved separationoflongeroligonucleotides.Investigatingthepeak-width at halfheight forthe chromatogramsinFig.3e-h we find thatit remains almostconstantusing60,100and200˚Aporesizebutfor the 300 ˚A pore size it increases withthe length of the oligonu-cleotide (SupplementaryMaterial FigS.3). T10 has approximately the samewidthonall poresizes. BeyondT20on the300 ˚A pore size the peak-widths using thehalf-height-measure becomesnot applicable dueto thehighpeak asymmetry.The shorter oligonu-cleotides between the three last main peaks, i.e., T30, T40, and T50, are hardto distinguishonall poresizessince thedifference insequencelengthandthustheadsorptionenergycontributedby eachadditionalnucleotidebecomesmallerwithincreasing molec-ularlength.Theabsolutedifferenceinretentiontimebetweenthe longer oligonucleotidesisgreater forthe200and300 ˚Acolumns thanthe60and100 ˚Acolumns,implyingthatthesewillhave bet-ter separationperformanceforlongeroligonucleotides.Thisresult isinlinewiththeresultobtainedbyCloseetal[6]who demon-strated that150 ˚Aporesincreasedresolutionbetween19-24mers butthat80 ˚Awaspreferableforshorter7-10mers.

To further investigate the separations from the analytical in-jections, the selectivity was calculated for pairs of neighboring oligonucleotides from all chromatograms, shown in Fig. 4a for equal-gradientconditionsandFig.4bfornormalizedretention.The difference betweenthesolute pairs isonlyone nucleotide, corre-spondingtoanapproximatedifferenceintheirlengthof6.76 ˚A.As observed, theselectivity isgreateron 200and300 ˚Acolumns in

bothequal-gradientandnormalized-gradientconditions,whilethe normalizationprovides improvedseparationbecauseoftheflatter gradientslope.

The observed improvement in selectivity between columns should depend solely on their individual accessible surface ar-eas,i.e., the pore size. As shownin Fig.1, the accessiblesurface areadecreasesconsiderablyfasterinthesmaller-pore-sizecolumns withincreasingoligonucleotidesize. Theratioofavailablesurface areafortheoligonucleotidepairsT6/T5uptoT20/T5ateachpore size is plotted in Fig. 4c. Here, we have assumed that each ad-ditionalnucleotide increasesthe length ofthe oligonucleotideby 6.76 ˚A.The ratioof accessiblesurface areadecreasesdrasticallyfor smallpore sizes,asthe smallerpores drastically restrict oligonu-cleotides from entering the inner pore volume. When the pores arelarger,thedifference issmaller,meaningthatthesoluteshave accesstomoreorlessequalsurfaceareasandthattheir selectiv-ityshouldimprove.Experimentalselectivityforthesame oligonu-cleotidepairsobtainedusingnormalizedretentiondata(Fig.3e–h) isshowninFig.4dandsupportsthesurfacearearatiocalculations. Here,thebenefitsofusingalargerporesizeare evenmore obvi-ous.Theselectivityimprovesbyafactorofupto3.45betweenthe largestand smallestpore sizes, andby afactor ofnearly 2.5 be-tween the300 and100 ˚A pores.This resultclearlydemonstrates the importance of utilizing columns withsufficient pore size for analyticalpurposes.

Theobserved benefitofusinglargerporesizesto increase se-lectivity isfurthersupported byadsorption theory.The accessible surfacearea correlatesto thesaturation capacityofeachcolumn, as we showed when studying methyl mandelate adsorption (see Fig.2). Inview ofpreviousfindings, itwouldthereforebe logical toassumethattheselectivitycanalsobeexplainedbythechange in saturation capacity. Eq. (7) expresses how the selectivity’s of twocompounds arerelatedtotheir adsorptionisotherm and gra-dientparameters.Whenseparatingtwooligonucleotidesofsimilar length,theirmonolayersaturationcapacitiescanbeassumedtobe

J. Bagge, M. Enmark, M. Le ´sko et al. Journal of Chromatography A 1634 (2020) 461653

Fig. 5. The selectivity between T10 and T9, theoretically determined from Eqn. 7 , plotted versus the monolayer saturation capacity (here, the saturation capacity is assumed to be equal for T10 and T9).

moreorlessequal.Sincethegradientconditionsobviouslyremain thesameinasinglechromatographicrun,theselectivitywilldiffer onlyasafunctionofSiandKw, i.Here,weassume thatKw is530

and1130andthatSis92.9and98.8forfirst-andsecond-eluting compounds, respectively.Theseparameters wereselectedbecause theygive closetoexperimentalselectivitybetweenT10andT9on 300 and60 ˚Acolumns.Interestingly,bycalculatingtheselectivity withvaryingvaluesofqs ,wefoundthatselectivityimprovesasthe adsorptioncapacityapproachessmallervalues(Fig.5).

If thesaturation capacities are insteadunequal andq_S

1 > qS 2,

selectivitywilldecreaseaccordingtoEq.(7).Thisisexpectedfrom comparingselectivitybetweenoligonucleotidesofincreasingly dif-ferent sizesonpacking materials withdecreasingpore sizes, and thisisindeedwhatcanbeobserved(cf.Fig4bandd).Toconclude, all thefindings implythat largerporesizesshouldprovidebetter selectivitybetweenoligonucleotidesofbothsimilarandespecially differentsizes.

4.2. Overloadedpreparative-scaleinvestigations

In Section 4.1we showedthat the selectivity oftwo arbitrary oligonucleotides is highest on the column with the larger-pore-sizepacking.Thiscouldbeattributedtothelargeravailablesurface area, whichis manifestedin decreasingcolumnsaturation capac-ity (Fig. 5). In this section we thereforeexamine how saturation capacity mightinfluencethe loadingcapacitywhen conductinga preparative separation of T9 and T10. These oligonucleotides are eluted usingeach stationary phase poresize, withoutthe degree ofpeakasymmetryobservedusingthe300 ˚Aporesizematerial.

OverloadedelutionprofilesoftheT10sampleforinjection vol-umesof16and64μLcanbeseeninFig.6usinggradientsG5–G8 in Table 1. Although the differences are small betweenthe four chromatograms, “touching bands” (i.e. precisely adjacent but not overlappingelutionprofiles)areonlyseenforthe60 ˚Acolumn be-tween T9andT10while partialoverlapoccursfortheotherpore sizes (see Fig. 6). This suggests slightly better separation perfor-manceforsmallerporesizes.

Forssén etal.previously usedsimulations toshow that inthe caseof isocraticseparation,increasing theselectivity affectedthe productivitymuchmorethandidincreasingthesaturation capac-ity [38]. They investigated 1000 different theoretical separation systems,inspiredbychiralisocraticseparations.

To investigate whether the Langmuir model as well as the equilibrium-dispersivecolumnmodelcanbeusedtopredict over-loaded elution profiles for oligonucleotides, the inverse method,

Eq.(5),wasusedto estimatetheadsorption isotherm parameters onthe 60and300 ˚Acolumns.The equilibriumconstant was de-terminedtobe341.6× 102 _{and4.817 × 10}2 _Lg–1 _{,thesaturation}

capacity31.24and33.53gL–1 ,andthemodifier-dependent param-eterS106.5and77.53for60and300 ˚A,respectively.Inaddition, theadsorptionisothermparametersforT9wasalsoestimated, as-suming identicalsaturation capacity. The equilibrium constant of T9wasdeterminedto be88.92× 102 _{and4.775× 10}2 _Lg–1 _and

the modifier-dependentparameter S 95.45 and80.55 for60 and 300 ˚A,respectively.

Large difference was observed between the equilibrium con-stants, butsince the estimated parameters were numerically de-termined merely to ensure good agreement betweenthe experi-mentalandmodelled profiles,the resultsshould insteadbe eval-uated fromtheshape oftheadsorption isotherm. Theadsorption isothermsforT10areshowninFig.7aandbforcolumnporesizes of60 and300 ˚A, respectively, atthe start andend ofthe gradi-ent(



0=0.088and



end=0.104).Numericalsimulations(dashed

lines)basedontheparametersareshowninFig.7canddforthe separationofT9andT10comparedwiththeexperimentalprofiles (solidlines),showinggoodcorrelation.Althoughonlypartial over-lapsareseenbetweenthesimulatedandexperimentalprofilesdue toaminoroffsetalongthetimeaxis,thegoodagreementinpeak shapes strongly suggests that the Langmuir adsorption isotherm (Eq.5) andthe equilibrium-dispersivecolumnmodel aresuitable choices.FromtheshapeoftheadsorptionisothermsinFig.7aand b,we can clearlyseethat the amountof solute adsorbedon the column is much higher on the 60 than the 300 ˚A column. This confirmsthatthe 60 ˚Apackinghasamorefavourableadsorption isotherm underoperational conditionsthandoesthe300 ˚A pack-ing,asexpectedfromanabsorbentwithahighermonolayer satu-rationcapacity.However,theacquiredelutionprofileswillalsobe affectedbyhowtotheadsorptionisothermchangeswithmodifier concentrationduringgradientelution,whichpotentially diminish-ingtheinitialdifferences.

UsingtheadsorptionmodelparametersdeterminedforT9and T10 discussed above, overloaded elution profiles were simulated forthe T9/T10sample. Fig. 8 presentsthe predictedelution pro-filesforthe touching-bandoptimization resultson 60 ˚A(Fig.8a) and300 ˚A (Fig.8b) columns. Touching bands were identified on the300 ˚Acolumnwitha58-μLinjectionandonthe60 ˚Acolumn witha64-μL injection,an increase inproductivityofabout10 %. Thiscorrelateswellwiththedifferenceinavailablesurfaceareaof T10whichis27.7m2 and30.6m2 onthe60and300 ˚Acolumns, respectively, according to the rod model prediction (Fig. 1). The

Fig. 6. Overloaded injection profiles of T10 on columns with pore sizes of (a) 60 ˚A (b) 100 ˚A (c) 200 ˚A and (d) 300 ˚A. The gradient slope was kept constant for all pore sizes; changes were only made to the initial volume fraction of MeCN to normalize the retention times (see gradients G5 – G8 in Table 1 ). The injection concentration was 5 g L –1 and the elution profiles are shown for the 16- and 64-μL injections.

Fig. 7. The relationship between the predicted adsorption isotherms derived using the inverse method and the experimental and predicted chromatographic profiles of T10 on 60 and 300 ˚A columns. The adsorption isotherms are shown for the (a) 60 ˚A and the (b) 300 ˚A columns at the start (red, top isotherm) and at the end (green, lower isotherm) of the gradient elution. In (c) and (d), the experimental (blue, solid lines) and predicted (orange, dashed lines) profiles are shown for the 60 and 300 ˚A columns, respectively. The injection concentration was 5 g L –1 and the injection volumes were 8 and 16 μL.

J. Bagge, M. Enmark, M. Le ´sko et al. Journal of Chromatography A 1634 (2020) 461653

Fig. 8. Simulated elution profiles for touching band injection volumes of T9 and T10 based on the adsorption isotherms presented in Fig. 7 a and b. The sample concentration was set to 5 g L –1 .

results demonstratethat the productivitydifference between dif-ferent pore sizesdecreasewithincreasing oligonucleotide length. Simulations also support the slight differences in the touching-bandseparationexperimentallypresentedinFig.6.

5. Conclusions

Experimental data from the separation of T5–T50 oligonu-cleotide ladders were used to demonstrate that the analytical separations of oligonucleotides benefit from using columns with largerporesizes.WhenexaminingtheselectivityforT5andlarger oligonucleotides uptoT20,the300 ˚Acolumnclearlyprovided se-lectivitythatwasbetterbyafactorof2.5thanthatofthe conven-tional100 ˚Acolumn.

Determining the adsorption isotherm for methylmandelate (a small uncharged probe) confirmedthat thesaturation capacityis linearlyproportionaltotheaccessiblesurfacearea.Thisledtothe investigation ofwhethertheselectivitycan bedescribedinterms oftheaccessiblesurfaceareaforaparticularoligonucleotide.With this findingit wasshownthat columns withlarge saturation ca-pacitieshavelowerselectivity.Itwasalsoshownthatifthe acces-siblesurfaceareasdifferbetweentheinvestigatedsolutes,the se-lectivitydecreases.Alltheseresultsindicatethatthe300 ˚Acolumn providedsufficientlylargeporesforsoluteswithintheinvestigated rangenottobesubstantially excludedfromtheporespace.These resultssuggestthata goodpracticeinordertoobtainoptimal se-lectivityistochoosealargeporesize.

Overloaded experimentswithT10indicatedthat the60 ˚A col-umnhada slightadvantage over columnswithlarger poresizes. Tofurtherinvestigatetheimpactofthestationaryporesizeunder overloaded operationconditions,we conductednumerical simula-tion usingparameters obtainedfromtheexperimental separation ofT9andT10using60and300 ˚Acolumnsdemonstratingthatthe 60 ˚A column provides 10 % times higher productivity than does the 300 ˚Acolumn. This isindeed interesting considering that he surface area of the 60 ˚A packing is fivetime largerthan that of the 300 ˚A material. The difference inproductivity instead corre-latesverywellwiththerelativedifferenceinavailablesurfacearea ofT10onthe60 ˚Aand300 ˚Acolumnwhichprovidesareasonable explanationtotheobservation.

DeclarationofCompetingInterest

Theauthorsdeclarethattheyhavenoknowncompeting finan-cialinterestsorpersonalrelationshipsthatcouldhaveappearedto influencetheworkreportedinthispaper.

CRediTauthorshipcontributionstatement

JoakimBagge:Conceptualization, Investigation,Writing- orig-inal draft, Writing - review & editing, Visualization. Martin En-mark: Conceptualization, Methodology, Validation, Formal anal-ysis, Writing - original draft, Writing - review & editing, Vi-sualization, Investigation. Marek Le´sko: Conceptualization, Soft-ware, Methodology, Resources. Fredrik Limé: Conceptualization, Resources.TorgnyFornstedt:Conceptualization,Methodology, Val-idation,Formalanalysis,Datacuration,Writing-review&editing, Supervision, Project administration, Funding acquisition. Jörgen Samuelsson:Conceptualization,Methodology,Software,Validation, Formal analysis, Investigation, Data curation, Writing - original draft,Writing-review&editing,Supervision.

Acknowledgementsandfundinginformation

This work was supported by the Swedish Knowledge Foun-dation via the project “BIO-QC: Quality Control and Purification for New Biological Drugs” (grant number 20170059) and by the SwedishResearchCouncil(VR)viatheproject“Fundamental Stud-iesonMolecularInteractionsaimedatPreparativeSeparationsand BiospecificMeasurements” (grantnumber2015–04627).

Supplementarymaterials

Supplementary material associated with this article can be found,intheonlineversion,atdoi:10.1016/j.chroma.2020.461653. References

[1] H. Toll, H. Oberacher, R. Swart, C.G. Huber, Separation, detection, and iden- tification of peptides by ion-pair reversed-phase high-performance liquid chromatography-electrospray ionization mass spectrometry at high and low pH, J. Chromatogr. A. 1079 (2005) 274–286, doi: 10.1016/j.chroma.2005.03.121 . [2] S. Studzi ´nska, Review on investigations of antisense oligonucleotides with the

use of mass spectrometry, Talanta 176 (2018) 329–343, doi: 10.1016/j.talanta. 2017.08.025 .

[3] K.H. Hecker, S.M. Green, K. Kobayashi, Analysis and purification of nucleic acids by ion-pair reversed-phase high-performance liquid chromatography, J. Biochem. Biophys. Methods. 46 (20 0 0) 83–93, doi: 10.1016/S0165-022X(00) 00133-0 .

[4] N.E. Hoffman, J.C. Liao, Reversed phase high performance liquid chromato- graphic separations of nucleotides in the presence of solvophobic ions, Anal. Chem. 49 (1977) 2231–2234, doi: 10.1021/ac50022a030 .

[5] A. Azarani, RNA analysis by ion-pair reversed-phase high performance liquid chromatography, Nucleic Acids Res. 29 (2001) 7e–77, doi: 10.1093/nar/29.2.e7 . [6] E.D. Close, A.O. Nwokeoji, D. Milton, K. Cook, D.M. Hindocha, E.C. Hook,

H. Wood, M.J. Dickman, Nucleic acid separations using superficially porous sil- ica particles, J. Chromatogr. A. 1440 (2016) 135–144, doi: 10.1016/j.chroma.2016. 02.057 .

[7] B.M. Wagner, S.A. Schuster, B.E. Boyes, J.J. Kirkland, Superficially porous silica particles with wide pores for biomacromolecular separations, J. Chromatogr. A. 1264 (2012) 22–30, doi: 10.1016/j.chroma.2012.09.052 .

[8] D. Gétaz, N. Dogan, N. Forrer, M. Morbidelli, Influence of the pore size of reversed phase materials on peptide purification processes, J. Chromatogr. A. 1218 (2011) 2912–2922, doi: 10.1016/j.chroma.2011.03.008 .

[9] H. Oberacher, A. Premstaller, C.G. Huber, Characterization of some physical and chromatographic properties of monolithic poly(styrene–co-divinylbenzene) columns, J. Chromatogr. A. 1030 (2004) 201–208, doi: 10.1016/j.chroma.2004. 01.009 .

[10] M. Enmark, J. Bagge, J. Samuelsson, L. Thunberg, E. Örnskov, H. Leek, F. Limé, T. Fornstedt, Analytical and preparative separation of phosphorothioated oligonucleotides: columns and ion-pair reagents, Anal. Bioanal. Chem. (2019), doi: 10.10 07/s0 0216- 019- 02236- 9 .

[11] P. Flodin, Methodological aspects of gel filtration with special reference to desalting operations, J. Chromatogr. A. 5 (1961) 103–115, doi: 10.1016/ S0021- 9673(01)92827- 4 .

[12] J.C. Giddings, K.L. Mallik, Theory of gel filtration (permeation) chromatography, Anal. Chem. 38 (1966) 997–10 0 0, doi: 10.1021/ac60240a014 .

[13] E. Casassa, Theoretical models for peak migration in gel permeation chro- matography, J. Phys. Chem. (1971) 75, doi: 10.1021/j10 0695a0 03 .

[14] E. Casassa, Equilibrium distribution of flexible polymer chains between a macroscopic solution phase and small voids, J. Polym. Sci. [B]. 5 (1967) 773– 778, doi: 10.1002/pol.1967.110050907 .

[15] R. Tijssen, P.J. Schoenmakers, M.R. Böhmer, L.K. Koopal, H.A.H. Billiet, Lat- tice models for the description of partitioning/ adsorption and retention in reversed-phase liquid chromatography, including surface and shape effects, J. Chromatogr. A. 656 (1993) 135–196, doi: 10.1016/0021-9673(93)80801-E . [16] Y. Wang, I. Teraoka, F.Y. Hansen, G.H. Peters, O. Hassager, A theoretical study

of the separation principle in size exclusion chromatography, Macromolecules 43 (2010) 1651–1659, doi: 10.1021/ma902377g .

[17] P.L. Dubin, J.I. Kaplan, B.-S. Tian, M. Mehta, Size-exclusion chromatography dimension for rod-like macromolecules, J. Chromatogr. A. 515 (1990) 37–42, doi: 10.1016/S0021-9673(01)89299-2 .

[18] A.Y.L. Sim, J. Lipfert, D. Herschlag, S. Doniach, Salt dependence of the radius of gyration and flexibility of single-stranded DNA in solution probed by small- angle x-ray scattering, Phys. Rev. E. 86 (2012) 021901, doi: 10.1103/PhysRevE.86. 021901 .

[19] Q. Chi, G. Wang, J. Jiang, The persistence length and length per base of single- stranded DNA obtained from fluorescence correlation spectroscopy measure- ments using mean field theory, Phys. Stat. Mech. Its Appl. 392 (2013) 1072– 1079, doi: 10.1016/j.physa.2012.09.022 .

[20] C. Horváth , High-performance liquid chromatography, Adv. Perspect. Vol. 2 (1980) Vol. 2 .

[21] T. Cecchi, Ion pairing chromatography, Crit. Rev. Anal. Chem. 38 (2008) 161– 213, doi: 10.1080/10408340802038882 .

[22] K.K. Unger , Porous Silica, its properties and Use as support in Column Liq- uid Chromatography, Elsevier Scientific Pub. Co.; distributors for the U.S.A. and Canada, Elsevier, North-Holland, Amsterdam; New York: New York, 1979 .

[23] B.W. Sands, Y.S. Kim, J.L. Bass, Characterization of bonded-phase silica gels with different pore diameters, J. Chromatogr. A. 360 (1986) 353–369, doi: 10. 1016/S0 021-9673(0 0)91685-6 .

[24] T. Undin, J. Samuelsson, A. Törncrona, T. Fornstedt, Evaluation of a combined linear–nonlinear approach for column characterization using modern alkaline- stable columns as model, J. Sep. Sci. 36 (2013) 1753–1761, doi: 10.1002/jssc. 201201132 .

[25] F. Gritti, G. Guiochon, Comparison between the loading capacities of columns packed with partially and totally porous fine particles: what is the effective surface area available for adsorption? J. Chromatogr. A. 1176 (2007) 107–122, doi: 10.1016/j.chroma.2007.10.076 .

[26] B.M. Wagner, S.A. Schuster, B.E. Boyes, T.J. Shields, W.L. Miles, M.J. Haynes, R.E. Moran, J.J. Kirkland, M.R. Schure, Superficially porous particles with 10 0 0 ˚A pores for large biomolecule high performance liquid chromatography and poly- mer size exclusion chromatography, J. Chromatogr. A. 1489 (2017) 75–85, doi: 10.1016/j.chroma.2017.01.082 .

[27] Y. Yamauchi, M. Taoka, Y. Nobe, K. Izumikawa, N. Takahashi, H. Nakayama, T. Isobe, Denaturing reversed phase liquid chromatographic separation of non- coding ribonucleic acids on macro-porous polystyrene-divinylbenzene resins, J. Chromatogr. A. 1312 (2013) 87–92, doi: 10.1016/j.chroma.2013.09.021 . [28] E. Stellwagen, J.M. Muse, N.C. Stellwagen, Monovalent cation size and DNA

conformational stability, Biochemistry 50 (2011) 3084–3094, doi: 10.1021/ bi1015524 .

[29] L.R. Snyder, J.W. Dolan, High-Performance Gradient Elution: The Practical Ap- plication of the Linear-Solvent-Strength Model, John Wiley & Sons, Inc., Hobo- ken, NJ, USA, 2006, doi: 10.1002/0470055529 .

[30] G. Guiochon , D.G. Shirazi , A. Felinger , A.M. Katti , Fundamentals of Preparative and Nonlinear Chromatography, 2nd ed., Academic Press, Boston, MA, 2006 .

[31] M. Zoubaïr, E. Fallah, G. Guiochon, Prediction of a protein band profile in preparative reversed-phase gradient elution chromatography: protein band profile in reversed-phase chromatography, Biotechnol. Bioeng. 39 (1992) 877– 885, doi: 10.1002/bit.260390810 .

[32] D. ˚Asberg, A. Langborg Weinmann, T. Leek, R.J. Lewis, M. Klarqvist, M. Le ´sko, K. Kaczmarski, J. Samuelsson, T. Fornstedt, The importance of ion-pairing in peptide purification by reversed-phase liquid chromatography, J. Chromatogr. A. 1496 (2017) 80–91, doi: 10.1016/j.chroma.2017.03.041 .

[33] J. Samuelsson, T. Fornstedt, Injection technique for generating accurate adsorp- tion isotherm data using the elution by characteristic points method, Anal. Chem. 80 (2008) 7887–7893, doi: 10.1021/ac8010999 .

[34] M. Enmark, J. Samuelsson, T. Undin, T. Fornstedt, Characterization of an unusual adsorption behavior of racemic methyl-mandelate on a tris-(3,5- dimethylphenyl) carbamoyl cellulose chiral stationary phase, J. Chromatogr. A. 1218 (2011) 6688–6696, doi: 10.1016/j.chroma.2011.07.064 .

[35] J. Samuelsson, T. Undin, A. Törncrona, T. Fornstedt, Improvement in the gen- eration of adsorption isotherm data in the elution by characteristic points method - the ECP-slope approach, J. Chromatogr. A. 1217 (2010) 7215–7221, doi: 10.1016/j.chroma.2010.09.004 .

[36] M. Le ´sko, D. ˚Asberg, M. Enmark, J. Samuelsson, T. Fornstedt, K. Kaczmarski, Choice of model for estimation of adsorption isotherm parameters in gradient elution preparative liquid chromatography, Chromatographia 78 (2015) 1293– 1297, doi: 10.1007/s10337-015-2949-0 .

[37] D. ˚Asberg, M. Le ´sko, M. Enmark, J. Samuelsson, K. Kaczmarski, T. Fornstedt, Fast estimation of adsorption isotherm parameters in gradient elution preparative liquid chromatography. I: the single component case, J. Chromatogr. A. 1299 (2013) 64–70, doi: 10.1016/j.chroma.2013.05.041 .

[38] P. Forssén, J. Samuelsson, T. Fornstedt, Relative importance of column and ad- sorption parameters on the productivity in preparative liquid chromatography II: investigation of separation systems with competitive Langmuir adsorption isotherms, J. Chromatogr. A. 1347 (2014) 72–79, doi: 10.1016/j.chroma.2014.04. 059 .