anion exchange membrane adsorbers for flow-through polishing steps: part i. clearance of minute...

ARTICLE

Anion Exchange Membrane Adsorbers forFlow-through Polishing Steps: Part I. Clearanceof Minute Virus of Mice

Justin Weaver,1 Scott M. Husson,2 Louise Murphy,1 S. Ranil Wickramasinghe1,3

1Department of Chemical and Biological Engineering, Colorado State University,

Fort Collins, Colorado 805232Department of Chemical and Biomolecular Engineering, Clemson University, Clemson, SC3Ralph E. Martin Department of Chemical Engineering, University of Arkansas,

Fayetteville, AR; telephone: þ1-479-575-8475; fax: þ1-479-575-4937;

e-mail: [email protected]

ABSTRACT: Membrane adsorbers may be a viable alterna-tive to the packed-bed chromatography for clearance ofvirus, host cell proteins, DNA, and other trace impurities.However, incorporation of membrane adsorbers intomanufacturing processes has been slow due to the significantcost associated with obtaining regulatory approval forchanges to a manufacturing process. This study has investi-gated clearance of minute virus of mice (MVM), an 18–22 nm parvovirus recognized by the FDA as a model viralimpurity. Virus clearance was obtained using three com-mercially available anion exchange membrane adsorbers:Sartobind Q1, Mustang Q1, and ChromaSorb1. Unlikeearlier studies that have focused on a single or few operatingconditions, the aim here was to determine the level of virusclearance under a range of operating conditions that couldbe encountered in industry. The effects of varying pH, NaClconcentration, flow rate, and other competing anionicspecies present in the feed were determined. The removalcapacity of the Sartobind Q andMustang Q products, whichcontain quaternary ammonium based ligands, is sensitive tofeed conductivity and pH. At conductivities above about20mS/cm, a significant decrease in capacity is observed. Thecapacity of the ChromaSorb product, which contains pri-mary amine based ligands, is much less affected by ionicstrength. However the capacity for binding MVM is signifi-cantly reduced in the presence of phosphate ions. Thesedifferences may be explained in terms of secondary hydro-gen bonding interactions that could occur with primaryamine based ligands.

Biotechnol. Bioeng. 2012;xxx: xxx–xxx.

� 2012 Wiley Periodicals, Inc.

KEYWORDS: biopharmaceuticals; bioseparations; mem-brane chromatography; monoclonal antibody; proteinpurification; validation of virus clearance

Introduction

Biopharmaceuticals, and in particular monoclonal anti-bodies (mAbs), represent an increasingly large fraction ofthe overall pharmaceutical market. Since downstreampurification costs can account for up to 80% of themanufacturing cost (Gottschalk, 2005), there is a strongdemand for new technologies that reduce the overallmanufacturing cost. Here we focus on the chromatographicpolishing steps in the production of mAbs as these stepshave become amajor production bottleneck (Langer, 2009).The impurities and contaminants to be removed duringthese processing steps are orders of magnitude lower inconcentration than the mAb.

Anion-exchange chromatography generally uses packedcolumns with porous chromatographic beads (Riordanet al., 2009a). However, packed-bed chromatography suffersfrom a number of disadvantages: the pressure drop acrossthe bed is usually high and may increase during operationdue to media deformation or blockage; pore diffusion isslow and often leads to degradation of the protein product,and scale up of packed beds is difficult. In addition, packedbeds have been shown to display a very low dynamiccapacity for virus particles at common process flow rates of150–450 cm/h, where binding is restricted to the surface ofresin particles, as slow pore diffusion prevents the virusparticles from entering the resin pores (Ghosh, 2002;Wickramasinghe et al., 2006; Yao and Lenhoff, 2006). Thus,most of the binding sites in the resin pores are not used,leading to underutilized beds (Han et al., 2005).

Correspondence to: S. R. Wickramasinghe

Contract grant sponsor: NSF Industry/University Cooperative Research Center

Received 9 April 2012; Revision received 7 August 2012; Accepted 20 August 2012

Accepted manuscript online 4 September 2012;

Article first published online in Wiley Online Library

(wileyonlinelibrary.wiley.com).

DOI 10.1002/bit.24720

� 2012 Wiley Periodicals, Inc. Biotechnology and Bioengineering, Vol. xxx, No. xxx, 2012 1

Membrane chromatography or membrane adsorption,where a macroporous membrane is used as a supportmaterial and the ligands are bound to the pore surface,was first described by Brandt et al. (1988). Membraneadsorbers can be run at much lower pressure drops and areeasy to scale up. Importantly, since the feed is pumpedthrough the membrane pores, transport of the solute to thebinding sites occurs mainly by fast convective flow.Consequently, the dynamic capacity is independent offlow rate over a much larger range of flow rates compared topacked beds (Curling and Gottschalk, 2007; Specht et al.,2004).

Nevertheless one of the major perceived disadvantages ofmembrane adsorbers is that the ligand density is generallyhigher for porous resin particles compared to macroporousmembranes. Consequently for smaller protein species thatare not excluded from the internal resin pores, the dynamiccapacity is higher for packed beds than for membranes(Curling and Gottschalk, 2007). Membrane adsorbers aretherefore ideally suited for removal of large molecules andvirus particles present at low concentrations (Zhou andTressel, 2006).

Presently, membrane adsorbers are used in the biophar-maceutical industry almost exclusively in ‘‘flow-through’’polishing steps (Boi, 2007; Zhou et al., 2008a). Anion-exchange membranes are used to bind impurities such asviruses, host cell proteins, and DNA while allowing the mAbto flow through. The feed pH is usually greater than 7.0where the impurities to be removed are negatively chargedbut the generally higher pI mAb is positively charged.Riordan et al. (2009a, b) indicate, that the ionic strength ofthe feed stream is usually low, as high ionic strength candisrupt electrostatic interactions and reduce the removal ofcontaminants.

Here we focus on clearance of minute virus of mice(MVM) an 18–22 nm parvovirus by three commerciallyavailable anion-exchange membranes: Sartobind Q1

(Sartorius AG, Gottingen, Germany), Mustang Q1 (PallCorporation, Port Washington, NY) and ChromaSorb1

(Millipore Corporation, Billerica, MA). Commercialmembrane adsorbers have pore sizes ranging from 0.5 to3mm. Since the feed is pumped through the membranepores, the kinetics of binding is the rate-limiting step(Teeters et al., 2002). Thus, compared to resin particles,much higher flow rates may be used. This results in theability to process a much larger load volume or to decreasethe process time required for a standard load volume; bothcases realize an economic benefit. Current commerciallyavailable anion exchange membrane adsorbers are sold aspre-sterilized, disposable devices, negating the need forcostly cleaning validation. Zhou et al. (2008a) havedemonstrated that an economic benefit can be realized ata product load of 2 kg/m2 membrane surface area, with costsavings coming predominantly from decreased buffer/WFIusage.

A few investigators have reported the use of membraneadsorbers for removal of low concentration high molecular

weight impurities and virus particles during the polishingsteps in the manufacture of mAbs (Knudsen et al., 2001; vanReis and Zydney, 2001; Zhou et al., 2006, 2008b). Brownet al. (2010) have investigated the use of membraneabsorbers as a prefiltration step prior to virus filtration.Zhou and Tressel (2006) used a Sartobind Q scale-downmodel (Sartobind Q 125) to obtain greater than 6.03 LRVat pH 7.2, conductivity, 4.0mS/cm, at a monoclonalantibody capacity of 3,000 g/m2. However, for manyindustrial manufacturing processes the feed stream requiresa large WFI dilution and/or pH adjustment to reach theseconditions, which can be both costly and time consuming.Furthermore, operating at low salt concentration mayincrease the likelihood of aggregation of the mAb (Ahreret al., 2006).

Frequently, ligands containing quaternary ammoniumions are grafted onto the surface of a macroporousmembrane to produce a strong anion exchange membraneadsorber, for example, Sartobind Q andMustang Q. The saltconcentration is kept low to prevent screening of theelectrostatic interactions between the ammonium ion andnegatively charged impurities. High salt concentrations canseverely limit capture of negatively charged impurities byquaternary ammonium based ligands (Curtis et al., 2003;Phillips et al., 2005).

The low binding capacities of quaternary ammoniumbased ligands at high salt concentration has led to thedevelopment of alternative anion-exchange ligands thatexhibit high capacities at high salt concentrations (Burtonand Harding, 1998; Riordan et al., 2009a). Recent studiesindicate that increased capacity at high salt concentrationsis due to secondary hydrophobic and hydrogen bondinginteractions between the target species and the ligand(Yang et al., 2007). Johansson et al. (2003) indicatethat non-aromatic anion-exchange ligands based onprimary and secondary amines display high capacities athigh salt concentrations. The presence of hydroxylgroups near the ionic group (primary amine) leads tosecondary hydrogen bonding interactions. Unlike theSartobind Q and Mustang Q, the ChromaSorb membraneadsorber contains primary amine based ion-exchangeligands.

The purpose of this study was to determine the level ofvirus clearance over a range of operating conditions forthree commercial anion exchange membrane adsorbers:Sartobind Q, Mustang Q, and ChromaSorb. This is the firstof a two part study. Here we focus exclusively on clearance ofMVM. In part 2 we investigate the effects of host cellproteins and DNA as well as a ‘‘model’’ antibody on MVMclearance. Since the anion exchange polishing step canfollow anion or cation exchange packed bed chromatogra-phy, two different sets of experiments, were conducted usingan anionic buffer set and a cationic buffer set to represent thefeed stream after anion- and cation-exchange chromatogra-phy respectively. The results of this study will help determinethe effect of changes in operating conditions on the level ofMVM clearance.

2 Biotechnology and Bioengineering, Vol. xxx, No. xxx, 2012

Experimental

Experimental Design

The anionic buffer set of experiments was a full-factorialdesign containing three variables: NaCl concentration (50and 200mM), pH (7.5 and 9.0), and flow rate (4, 20membrane volume/min (MV/min)). One centerpoint runwas also included at 125mMNaCl, pH 8.25, and 12MV/minflow rate. The cationic buffer set of experiments was a half-factorial design containing four variables: NaCl concentra-tion (0 and 200mM), pH (6.0 and 9.0), phosphateconcentration (0 and 50mM), and flow rate (4 and20MV/min). The effect of phosphate in the feed for thecationic buffer set of experiments was investigated asnegatively charged buffer species are commonly present inthe eluate from a cation exchange chromatography step. Acenterpoint run was also conducted at 100mM NaCl, pH7.5, 12MV/min flow rate, and 25mM phosphate. Oneadditional run was conducted using solution conditions thatresulted in poor binding for the ChromaSorb (200mMNaCl, pH 9.0, and 50mM phosphate) where phosphate wasreplaced by 50mM acetate. All runs were conducted induplicate. Table I gives the feed buffers tested.

Materials

Sartobind Q Nano 1mL devices, Mustang Q coins withstainless steel housing, and ChromaSorb 0.08mL deviceswere obtained through gracious donations from SartoriusAG, Pall Corporation, and Millipore Corporation, respec-tively. All buffer chemicals (tris-base, tris–HCl, NaCl,NaH2PO4 (monohydrate), Na2HPO4 (anhydrous), glacialacetic acid, and sodium acetate trihydrate) were purchasedfrom JT Baker (Phillipsburg, NJ). Table I gives the variousfeed buffers used in order of increasing conductivity. Thecolumn labeled ‘‘design of experiments’’ indicates whetherthe buffer was used for the anionic or cationic buffer set of

experiments. Three pairs of stock buffers were prepared: (1)20mM tris–HCl, 20mM tris-base; (2) 20mM tris–HCl and50mM monobasic phosphate, 20mM tris base and 50mMdibasic phosphate; (3) 20mM tris–HCl and 50mMacetic acid, 20mM tris base and 50mM sodium acetate.The buffers given in Table I were prepared by titrating therequired volume of each pair of buffers to obtain thedesired pH and adding the appropriate mass of NaCl.

pMVM (minute virus of mice prototype strain) andmouse A9 fibroblasts were purchased from ATCC(Manassas, VA). High glucose DMEM media, DMEMcontaining trypsin/EDTA and fetal bovine serum wereobtained from HyClone, a division of Thermo FisherScientific (Waltham, MA). A9 cells were thawed andexpanded into multiple T-150 culture flasks using highglucose DMEM media with 10% FBS and 100mg/mLpenicillin. Cell culture was performed in jacketed incubators(non-infected and infected cultures in separate incubators)at 378C with 10% CO2. Cultures were expanded and grownto �80% confluence at which point infection with MVMwas performed. Infection was accomplished by discardingold growth media from culture (cells were adherent) andincubating cells at 378C for 10min in 1mL DMEMcontaining 1� 1010–1� 1011 virus particles/mL for anMOI (multiplicity of infection) of greater than 1,000assuring efficient infection. An additional 34mL of DMEMmedia (as described above) were added to the cultures afterthe 10min incubation for a total 35mL culture volume. A 6-day infection propagation period was found to be suitablefor complete infection/propagation, which consistentlyyielded MVM titers of 1� 1010–1� 1011 particles/mL.

Assays

Quantification of MVM was accomplished through aquantitative PCR (QPCR) assay (Ros et al., 2002;Klee et al., 2006; Wickramasinghe et al., 2010). iQSYBR green w/fluorescein master mix was purchasedfrom BioRad (Hercules, CA). Forward (50-GACGCACAGAAAGAGAGTAACCAA-30) and reverse (50-CCAACCATCTGCTCCAGTAAACAT-30) primers were purchasedwith standard desalting purification from IDT (Coralville,IA). RQ1 DNase enzyme and buffer were purchased fromPromega (Madison,WI). QPCRwas performed on a BioRadiQ5 real-time PCR system with iQ5 optical system softwarev2.0.

QPCR runs were performed in unskirted, low-profile 96-well PCR plates (BioRad) with polypropylene microseal ‘‘B’’adhesive sealers with 20mL per reaction. The reaction recipewas taken directly from the iQ SYBR Green master mixinstructions (given per reaction): 10mL SYBR Green mastermix, 8.2mL distilled water, 0.4mL forward primer (final100 nM), 0.4mL reverse primer (final 100 nM), and 1mLsample. MVM standards were created by PCR amplificationof a highly conserved 501 bp portion of the MVM genomeusing the above mentioned primers and capturing the PCR

Table I. Details of buffers.

Conductivity

(mS/cm)

NaCl

(mM) pH

Phosphate

(mM)

Design

of experiments

0.5 0 9.0 0 CEX

2.3 0 6.0 0 CEX

6.0 50 9.0 0 AEX

6.0 0 6.0 50 CEX

7.2 50 7.5 0 AEX

7.6 0 9.0 50 CEX

14.4 125 8.25 0 AEX

14.6 100 7.5 25 CEX

21.4 200 9.0 0 AEX/CEX

22.0 200 7.5 0 AEX

22.6 200 6.0 0 CEX

23.7 200 9.0 50 (acetate) CEX

25.4 200 6.0 50 CEX

26.5 200 9.0 50 CEX

All buffers contained 20mM tris.

Weaver et al.: Virus Clearance by Anion-Exchange Membranes 3

Biotechnology and Bioengineering

product in the pCR2.1-TOPO plasmid using a InvitrogenTOPO TA Cloning kit (Life Technologies, Carlsbad, CA).Standards ranging from 1� 109 to 1� 102 copies/mL werecreated by serial dilution of the Maxi-prepped (Qiagen,Valencia, CA) cloned PCR product. Annealing temperaturewas determined through temperature gradient runs andmelt curve analysis; 578C resulted in a single dominant meltcurve. The initial PCR cycle was 958C for 10min, whichfunctioned to open virus particles, denature DNA, andinactivate RQ1 DNase. Then 45 cycles of the following wererepeated: denaturing at 958C for 15 s, primer annealing at578C for 10 s, elongation at 728C for 45 s, and an additional10 s at 728C to collect the real-time fluorescence data.

The QPCR assay limit of detection was determined byserially diluting the 1� 102 copies/mL standard 2� until asample with 1 copy/mL was reached. Samples were run intriplicate on a single plate following the above protocol. Thisplate was repeated two additional times for three totalreplicate plates and a total of nine replicates per sample.Limit of detection was determined at 95% confidenceinterval by Probit analysis using Minitab statistical software.The limit of detection was determined to be 14 copies/mL.

Prior to transferring to the QPCR plate, virus-containingsamples (1mL) were pipetted into 96-well plates containing9mL of DNase solution (1mL RQ1 DNase, 1mL 10� DNasebuffer, and 7mL of dH2O). The DNase step providedassurance that the QPCR assay would only quantifycomplete viral particles and not naked viral DNA thatwould not be infective. These plates were sealed andincubated at 378C for 40min. One microliter each of theDNase treated samples were pipetted into a BioRad PCRplate with 19mL of iQ SYBR Green master mix with primers(described above). Eight standards were run with every plate(1� 102–1� 109 copies/mL) from which a linear standardcurve was constructed to quantify unknown samples. Allsamples were run in duplicate.

Virus Clearance

All membrane adsorber evaluation runs were performed onan AKTA FPLC (GE Healthcare Bio-Sciences Corp,Piscataway, NJ) with FRAC-950 fraction collector usingthe associated Unicorn software v. 5.1. Conductivity,absorbance at 280 nm, backpressure, and temperaturewere recorded. Prior to testing, all membrane moduleswere wet with running buffer according to the manufac-turers’ instructions. The FPLC system was flushed withwater followed by running buffer prior to attachment of themembrane module. Membranes were installed with a slowflow rate (�1MV/min) in the bottom-to-top or reversedirection to minimize the possibility of air entering themembrane devices. Equilibration of the devices wasaccomplished at flow rates (4–20MV/min) in top-to-bottom or forward direction for 20 membrane volumes oruntil the conductivity and UV absorbance at 280 nm werestable. After the membranes were equilibrated, the buffer

was spiked with a 1:100 dilution of MVM virusstock. Spiking was performed immediately prior toloading to minimize any buffer effects on the virus.MVM challenge titers for all runs were between 108

and 109 virus particles/mL.Virus-spiked load volumes were normalized to 500mL

per mL of membrane with the Sartobind Q, Mustang Q,and ChromaSorb having volumes of 1, 0.35, and 0.08mL,respectively. During the load step, effluent was collected in10 equal volume fractions. Virus concentration wasdetermined for all fractions. After virus loading, themembranes were washed with at least 20 membranevolumes of running buffer.

Membrane Characterization

Membranes were imaged using field-emission scanningelectron microscopy (FESEM) (Model JSM-6500F,Waltham, MA) and analyzed using X-ray photoelectronspectroscopy (XPS) (2000 Physical Electronics 5800 ultra-high vacuum XPS-Auger spectrometer; Chanhassen, MN).The equilibrium ion-exchange capacities of the membraneadsorbers were determined by titration. Membrane sampleswere prepared by soaking them overnight in excess 0.5NNaOH, washing with distilled water until the pH was that ofthe distilled water, and soaking overnight in 2M NaCl. Nextthe solution was titrated with 0.01N HCl. The ion-exchangecapacity is defined as the number of OH� ions per surfacearea. Membrane geometries as specified by the manufactureswere: Sartobind Q; cylindrically wound 15-layer membraneabout 36 cm2 surface area; Mustang Q: 1.8 cm diameter stackof 10 membranes surface area about 24 cm2; ChromaSorb;1.5 cm diameter stack of 8 membranes, surface area about14 cm2.

Results and Discussion

Figure 1 gives the equilibrium ion-exchange bindingcapacities for the membrane adsorbers. All measurementwere conducted in triplicate using membranes fromdifferent lots. Average results are given. ChromaSorbshowed a fourfold higher static binding capacity thanSartobind Q and a 12-fold higher capacity than Mustang Q.While Sartobind Q and Mustang Q contain quaternaryamine based anion-exchange ligands, ChromaSorb containspolyallylamine (primary amine based ligands) (Woo et al.,2011).

Figure 2 gives FESEM images for the three membraneadsorbers. All three membrane adsorbers possess a very openstructure that gives rise to high permeability and rapidconvective flow through the membrane pores, which appearto range from 0.5 to 3mm. According to the manufacturers,Sartobind Q consists of a base regenerated cellulosemembrane, nominal pore size 3mm; Mustang Q consistsof a polyethersulfone membrane, nominal pore size 0.8mm;


ChromaSorb consists of a polyethylene base membrane,nominal pore size 0.65mm.

High resolution XPS spectra for nitrogen for all threemembranes indicate the presence of primary amine (400 eV)and quaternary amine (402 eV) peaks (Liu et al., 2006). ForSartobind Q (Fig. 3A) only a quaternary amine peak isdetected; while for ChromaSorb (Fig. 3C), only a primaryamine peak is present. Though the manufacturer describesMustang Q as containing quaternary amine based ligands,Figure 3B suggests some primary amines are also present atthe membrane surface. This is probably a result of themonomer species and the linker/coupling chemistry used(Johansson et al., 2003).

Table II gives log virus reduction (LRV) for the anionicbuffer set of experiments. The results are given in order ofincreasing conductivity. LRV is defined as log (total virusloaded)� log (virus in all 10 fractions plus in the wash).Limit of detection based on 1mL samples, and accountingfor dilution in DNase solution, was 1.4� 105 virus particles/mL. Thus, LRV values where the virus titer was below limitof detection are given as greater than (>) limit of detection.Uncertainty values for LRV above the level of detectionrepresent 1 standard deviation.

As pore diffusional resistance is minimized in membraneadsorbers, a constant dynamic capacity over a large range offlow rates is frequently observed (Specht et al., 2004; Hanet al., 2005). Though experiments were conducted at flowrates between 4 and 20MV/min, Table II indicates thatfor conductivities up to 14.4mS/cm, all three adsorbersdisplayed virus clearance greater than limit of detection.Since the pI of MVM is 5–6 (Anouja et al., 1997), MVM isnegatively charged at the pH values tested here.

Consequently, given the limited data below the limit ofdetection, it not possible to determine the effect of flow rateon MVM clearance.

At conductivities above 14.4mS/cm, Mustang Q givesvery low LRV. Sartobind Q displays LRV above the limit ofdetection for conductivities up to 21.4mS/cm and pHabove 7.5. Salts in solution shield charges between thevirus particles and quaternary ammonium ligands, leadingto the lower observed capacities at higher conductivities.Figure 1 indicates that Mustang Q has a lower staticcapacity compared to Sartobind Q, which could explain thecompromised virus clearance at 200mM NaCl, regardlessof pH.

ChromaSorb displays virus clearance in excess of the limitof detection at all conductivities tested. As indicated inFigure 1, the static binding capacity of ChromaSorb is muchhigher, which could partially explain the higher capacity athigher salt concentrations. Since quaternary amines aremuch stronger anion exchangers than primary amines, theyare more effective at low ionic strength. The presence of alkylgroups attached directly to the nitrogen atom prevents anysecondary hydrogen bonding interactions in these systems.In contrast, the hydrogen atoms of primary amines can leadto secondary, hydrogen bonding interactions that result inhigher capacities for primary amine based ligands at highsalt concentrations where Coulombic interactions arediminished (Johansson et al., 2003). Consequently, thecapacity of the ChromaSorb remains high at all conductivi-ties tested. Figure 3B indicates that the Mustang Q alsodisplays a primary amine peak. However, the much lowerstatic capacity of the Mustang Q may explain the lowcapacity at higher conductivities.

Table III gives virus clearance data for the cationic bufferset of experiments arranged in order of increasingconductivity. As was the case for the anionic buffer set ofexperiments, the effect of flow rate on virus clearance cannotbe determined. The difference between the anionic andcationic buffer set of experiments is the presence ofphosphate and acetate ions, as well as an expanded pHrange. The binding capacities of Sartobind Q and MustangQ are affected strongly by conductivity, in agreement withthe results in Table II. Mustang Q, unlike Sartobind Q,shows a large decrease in capacity at pH 6.0 andconductivities of 6.0 or 22.6mS/cm. This result is mostlikely due to the lower static binding capacity for MustangQ. As was the case in Table II, uncertainty values for LRVabove the level of detection represent 1 standard deviation.While the accuracy of the standard deviation is limited giventhat only two runs were conducted for each condition, theuncertainties do indicate if differences in LRV are likely to besignificant.

Table III indicates that the presence of phosphate ions hasa detrimental effect on the ChromaSorb binding capacity.Phosphoric acid has three pKa values: 2.3, 7.2, and 12.1.Consequently, at pH 6.0 most of the phosphate will bepresent as H2PO

�4 , while at pH 7.5 and 9.0, most will be

present as HPO2�4 . When the phosphate is present as HPO2�

4

Sartobind Q Mustang Q ChromaSorb0

1

2

3

4

5

6

7Io

n-E

xcha

nge

Cap

acity

(µ

eq /

cm2)

Figure 1. Static binding capacity Sartobind Q, Mustang Q and ChromaSorb.

Error bars represent one standard deviation, n¼ 3.



it has the greatest reduction on the static binding capacity ofthe ChromaSorb. Since phosphate is doubly charged and canform strong hydrogen bonding interactions with the Hatoms attached to the primary amine it will preferentiallybind to the primary amine. Further the presence of a doublenegatively charged phosphate will tend to repel negativelycharged MVM particles. Consequently a significant decreasein MVM binding capacity is observed. At pH 6.0, the singlycharged dihydrogen phosphate ion has a much lower effecton capacity. In fact, at a conductivity of 6.0mS/cm, the LRVfor the ChromaSorb is above the limit of detection. At aconductivity of 25.4mS/cm, pH 6.0 and 50mM phosphate,ChromaSorb shows a slight decrease in MVM bindingcapacity, probably due to combination of charge screeningand competitive binding with dihydrogen phosphate ion.

The strong base quaternary amine ligands of Sartobind Qand Mustang Q are much less affected by the presence of thedoubly charged hydrogen phosphate ion at pH 9.0. Themajor contributor to the decrease in MVM binding is due to

charge shielding at higher conductivities. The presence ofacetate ions has no effect on the binding capacity of theChromaSorb device. In the case of SartobindQ andMustangQ it is again the conductivity of the solution that affectsMVM capacity. Like dihydrogen phosphate, acetate is singlycharged. Consequently the capacity of the ChromaSorb isunaffected by the presence of 50mM acetate.

Interestingly for Sartobind Q andMustang Q, at pH 9.0 inthe presence of 200mMNaCl the addition of acetate insteadof phosphate leads to a greater decrease in LRV. Thesedifferences between the behavior of the primary amine andquaternary amine are most likely due to the fact that thequaternary amine acts as a pure ion exchange ligand. Theprimary amine on the other hand, is affected by acid baseinteractions. In particular, here the interactions will beaffected by the acidity of the conjugate base of the primaryamine and the basicity of the acetate and phosphate speciespresent. Improved membrane adsorber performance may beachieved by diafiltering the feed in order to exchange the

Figure 2. FE-SEM images of membranes. Figures A, D, and G (1,000�) and B, E, and H (10,000�) give images of the Sartobind Q, Mustang Q and ChromaSorb membranes.

Figures C, F, and I (1,000�) give cross sectional images of the Sartobind Q, Mustang Q, and ChromaSorb membranes, respectively.


feed buffer and optimize the ionic strength. However thiswill add to the purification cost and reduce productrecovery.

Breakthrough curves for all three membrane adsorbers aregiven in Figure 4. For Sartobind Q and Mustang Q the feedconsisted of 200mM NaCl, pH 7.5, and a flow rate of 4 MV/min. Since the ChromaSorb did not show virus break-through at this condition, breakthrough behavior at

100mM NaCl, 25mM phosphate, pH 7.5, and flow rate12MV/min are presented. Breakthrough occurs for all threemembrane adsorbers after 50MV of feed have passedthrough the membrane. However, the concentration in theflow through is low and even after 500MV is still less than15% of the feed concentration. Similar behavior has beenobserved by others for adsorption of virus (Wickramasingheet al., 2006), proteins (Yang and Etzel, 2003), and plasmidDNA (Endres et al., 2003). The slow approach to the feedconcentration in the flow through has been explained usingthe ‘‘car parking’’ model (Talbot et al., 2000). Since MVMparticles are large and attach randomly to ligands, geometricblockage, steric hindrance and charge repulsion will slow therate of binding as surface coverage increases. This effect isgreatest when the solute species is large relative to thespacing between the binding sites. After breakthroughoccurs, the flow through concentration remains low relativeto the feed concentration.

Comparing Figures 4 and 1 indicates that the flowthrough concentration for the Mustang Q is the lowestthough it has the lowest ion-exchange binding capacity.Since MVM particles are much larger than OH� ions, thecapacity of the three membrane adsorbers for MVM couldbe very different if a large number of ligands are unavailablefor MVM binding due to steric hindrance and stericexclusion.

The three membrane adsorbers tested here containdifferent base membranes with different nominal poresizes. Thus, in addition to ligand chemistry, optimization ofthe membrane pore structure and three dimensionalarrangement of the ion-exchange ligands are importantfactors to consider for maximizing virus capture. Stackingmultiple membranes (in a radial flow configuration) willmodify the ‘‘effective’’ pore size through the membrane.Further, the membrane pore surface should be covered withan open, three-dimensional layer of ion-exchange ligands.Models that describe attachment of virus particles to themembrane surface indicate that the spacing of the ion-exchange groups relative to the size of the adsorbed species isimportant. Finally, development of primary amine based

Figure 3. XPS N1s high-resolution region scans from top-to-bottom: Sartobind

Q, Mustang Q, and ChromaSorb. Counts are plotted against binding energy.

Table II. MVM removal results from the anionic buffer set of

experiments.

Conductivity

(mS/cm)

NaCl

(mM) pH

Flow rate

(MV/min) LRV-S LRV-M LRV-C

6.0 50 9.0 4 >3.58 >3.25 >3.70

6.0 50 9.0 20 >3.47 >3.25 >4.01

7.2 50 7.5 4 >3.57 >3.25 >4.06

7.2 50 7.5 20 >3.54 >3.25 >4.04

14.4 125 8.25 12 >3.55 >3.25 >4.12

21.4 200 9.0 4 >3.52 1.17� 0.00 >3.70

21.4 200 9.0 20 >3.47 1.00� 0.07 >3.95

22.0 200 7.5 4 0.98� 0.05 1.62� 0.01 >3.71

22.0 200 7.5 20 1.65� 0.04 1.28� 0.02 >4.03

>Virus titers were below level of detection. LRV-S, LRV-M, and LRV-Crefer to LRV for Sartobind Q, Mustang Q, and ChromaSorb, respectively.Errors represent one standard deviation with n¼ 2.



ion-exchange ligands that are effective at high conductivitieswill require appropriate substitution of groups capable offorming secondary interactions with MVM particles nearthe cationic sites.

Conclusions

A comprehensive evaluation of virus removal capability ofleading commercial anion-exchange membrane adsorbershas been conducted. The capacity for binding MVM isseverely compromised at higher conductivities for quater-nary amine based membrane adsorbers such as SartobindQ and Mustang Q. The binding capacity of the primaryamine based ChromaSorb is far less sensitive to conductivityof the feed solution. However, it is severely compromised

by the presence of competing hydrogen phosphate ions.Membrane adsorbers like ChromaSorb that containprimary amine based ligands make use of secondaryinteractions to maintain a high capacity at high feedconductivities. Consequently, the presence of ions that arealso capable of forming hydrogen bonds with the primaryamine groups will have a detrimental effect on the capacity.

Maximizing the capacity of anion exchange membraneadsorbers for capture of large contaminants such as virusparticles will require optimization of the membrane porestructure and the three dimensional arrangement of the ion-exchange groups. Maximizing the density of ion-exchangegroups alone will not necessarily maximize capacity, ascapture of large virus particles can lead to steric hindranceand steric exclusion effects that render a number of the ionexchange groups inaccessible.

Funding was provided by the NSF Industry/University Cooperative

Research Center for Membrane Applied Science and Technology at

the University of Colorado, as well as Colorado State University and

Clemson University. We also thank Dr. Mark Bailey for his input and

direction with this project.

References

Ahrer K, Buchbacher A, Iberer G, Jungbauer A. 2006. Thermodynamic

stability and formation of aggregates of human immunoglobulin G

characterised by differential scanning calorimetry and dynamic light

scattering. J Biochem Biophys Methods 66(1–3):73–86.

Anouja F,Wattiez R,Mousset S, Caillet-Fauquet P. 1997. The cytotoxicity of

the parvovirus minute virus of mice nonstructural protein NS1 is

related to changes in the synthesis and phosphorylation cell proteins.

J Virol 71(6):4671–4678.

Boi C. 2007. Review: Membrane adsorbers as purification tools for mono-

clonal antibody purification. J Chromatogr B 848(1):19–27.

Brandt S, Goffe RA, Kessler SB, O’Connor JL, Zale SE. 1988. Membrane-

based affinity technology for commercial scale purifications. Bio-

Technology 6(7):779–782.

Brown A, Bechtel C, Bill J, Liu H, Liu J, McDonald D, Pai S, Radhamohan A,

Renslow R, Thayer B, Yohe S, Dowd C. 2010. Increasing parvovirus

filter throughput of monoclonal antibodies using ion exchange mem-

brane adsorptive pre-filtration. Biotechnol Bioeng 106(4):627–637.

Burton SC, Harding DRK. 1998. Hydrophobic charge induction chroma-

tography: Salt independent protein adsorption and facile elution with

aqueous buffers. J Chromatogr A 814(1–2):71–81.

Table III. MVM removal results from the cationic buffer set of experiments.

Conductivity (mS/cm) NaCl (mM) pH Flow rate (MV/min) Phosphate (mM) LRV-S LRV-M LRV-C

0.5 0 9.0 4 0 >3.37 >3.36 >3.21

2.3 0 6.0 20 0 >3.23 >3.37 >3.21

6.0 0 6.0 4 50 >3.26 2.81� 0.01 >3.21

7.6 0 9.0 20 50 >3.16 >3.34 1.12� 0.02

14.6 100 7.5 12 25 2.76� 0.05 2.02� 0.07 1.95� 0.13

21.4 200 9.0 20 0 >3.26 1.35� 0.03 >3.16

22.6 200 6.0 4 0 2.53� 0.05 1.34� 0.04 >3.21

23.7 200 9.0 20 50 (acetate) 0.64� 0.07 0.21� 0.04 >3.11

25.4 200 6.0 20 50 1.23� 0.03 1.57� 0.01 2.49� 0.26

26.5 200 9.0 4 50 0.92� 0.02 1.24� 0.06 1.34� 0.01

>Virus titers were below level of detection. LRV-S, LRV-M, and LRV-C refer to LRV for Sartobind Q, Mustang Q, and ChromaSorb, respectively. Errorsrepresent one standard deviation with n¼ 2.

0 100 200 300 400 5000.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

C /

Co

Dimensionless Volume

Figure 4. Normalized effluent virus concentration (C divided by feed concen-

tration C0) versus dimensionless volume (effluent volume/membrane volume). For

Sartobind Q and Mustang Q, the feed buffer consisted of 20mM Tris, 200mM NaCl, pH

7.5, and 4MV/min. For ChromaSorb, the feed buffer consisted of 20mM Tris, 100mM

NaCl, 25mM phosphate, pH 7.5, and 12MV/min run. Symbols are as follows: Mustang Q

(&), Sartobind Q (*), and ChromaSorb (~).


Curling J, Gottschalk U. 2007. Process chromatography: Five decades of

innovation. BioPharm Int 20(10):70–94.

Curtis S, Lee K, Blank GS, Brorson K, Xu Y. 2003. Generic/Matrix evalua-

tion of SV40 clearance by anion exchange chromatography in flow-

through mode. Biotechnol Bioeng 84(2):179–186.

Endres HH, Johnson JAC, Ross CA, Welp JK, Etzel MR. 2003. Evaluation of

ion-exchange membrane and purification of plasmid DNA. Biotchnol

Appl Biochem 37:259–266.

Ghosh R. 2002. Protein separation using membrane chromatography:

Opportunities and challenges, J Chromatogr A 952(1–2):13–27.

Gottschalk U. 2005. Downstream processing of monoclonal antibodies:

From high dilution to high purity. Biopharm Intern 18(6):42–58.

Han B, Specht R, Wickramasinghe SR, Carlson JO. 2005. Binding

Aedes aegypti densonucleosis virus to ion exchange membranes.

J Chromatogr A 1092(1):114–124.

Johansson B-L, BelewM, Eriksson S, Glad G, LindO,Maloisel J-L, Norrman

N. 2003. Preparation and characterization of prototypes for multi-

modal separation media aimed for capture of negatively charged

biomolecules at high salt conditions. J Chromatogr A 1016(1):21–33.

Klee SR, Tyczka J, Ellerbrok H, Franz T, Linke S, Baljer G, Appel B. 2006.

Highly sensitive real-time PCR for specific detection and quantification

of Coxiella burnetii. BMC Microbiol 6:2.

Knudsen HL, Fahrner RL, Xu Y, Norling LA, Blank GS. 2001. Membrane

ion-exchange chromatography for process-scale antibody purification.

J Chromatogr A 907(1–2):145–154.

Langer ES. 2009. Downstream factors that will continue to constrain

manufacturing through 2013. BioProcess J 8(4):24–28.

Liu X, Wu N, Wunsch BH, Barsotti RJ, Jr., Stellacci F. 2006. Shape-

controlled growth of micrometer sized gold crystals by a slow reduction

method. Small 2(8–9):1046–1050.

Phillips M, Cormier J, Ferrence J, Dowd C, Kiss R, Lutz H, Carter J. 2005.

Performance of a membrane adsorber for trace impurity removal in

biotechnology manufacturing. J Chromatogr A 1078(1–2):74–82.

Riordan W, Heilmann S, Brorson K, Seshadri K, He Y, Etzel M. 2009a.

Design of salt-tolerant membrane adsorbers for viral clearance. Bio-

technol Bioeng 103(5):920–929.

Riordan WT, Heilmann SM, Brorson K, Seshadri K, Etzel MR. 2009b. Salt

tolerant membrane adsorbers for robust impurity clearance. Biotechnol

Prog 25(6):1695–1702.

Ros C, Burckhardt CJ, Kempf C. 2002. Cytoplasmic trafficking of minute

virus of mice: Low-pH requirement, routing to late endosomes, and

proteasome interaction. J Virol 76(24):12634–12645.

Specht R, Han B, Wickramasinghe SR, Carlson JO, Czermak P, Wolf A, Reif

O-W. 2004. Densonucleosis virus purification by ion exchange mem-

branes. Biotechnol Bioeng 88(4):465–473.

Talbot J, Tarjus G, van Tassel PR, Viot P. 2000. From car parking to protein

adsorption: An overview of sequential adsorption processes. Colloid

Surf A 165(1–3):287–324.

Teeters MA, Root TW, Lightfoot EN. 2002. Performance and scale-up

of adsorptive membrane chromatography. J Chromatogr A 944

(1–2):129–139.

van Reis R, Zydney AL. 2001.Membrane separations in biotechnology. Curr

Opin Biotech 12(2):208–211.

Wickramasinghe SR, Carlson JO, Teske C, Hubbuch J, Ulbricht M. 2006.

Characterizing solute binding to macroporous ion exchangemembrane

adsorbers using confocal laser scanning microscopy. J Membr Sci

281(1–2):609–618.

Wickramasinghe SR, Stump ED, Grzenia DL, Husson SM, Pellegrino J.

2010. Understanding virus filtration membrane performance. J Membr

Sci 365(1–2):160–169.

Woo M, Khan NZ, Royce J, Mehta U, Gagnon B, Ramaswamy S, Soice N,

Morelli M, Cheng K-S. 2011. A novel primary amine-based anion

exchange membrane adsorber. J. Chromatogr A 1218(32):5386–5392.

Yang H, Etzel MR. 2003. Evaluation of three kinetic equations in models of

protein purification using ion-exchange membranes. Ind Eng Chem

Res 42(4):890–896.

Yang T, Malmquist G, Johansson B-L, Maloisel J-L, Cramer S. 2007.

Evaluation of multi-modal high salt binding ion exchange materials.

J Chromatogr A 1157(1–2):171–177.

Yao Y, Lenhoff AM. 2006. Pore size distributions of ion exchangers and

relation to protein binding capacity. J Chromatogr A 1126(1–2):107–

119.

Zhou JX, Tressel T. 2006. Basic concepts in Q membrane chromatography

for large scale antibody production. Biotechnol Prog 22(2):341–349.

Zhou JX, Tressel Gottschalk U, Solamo F, Pastor A, Dermawan S, Hong T,

Reif O, Mora J, Hutchinson F, Murphy M. 2006. New Q membrane

scale-down model for process-scale antibody purification.

J Chromatogr A 1134(1–2):66–73.

Zhou JX, Tressel Yang X, Seewoester T. 2008a. Review: Implementation of

advanced technologies in commercial monoclonal antibody produc-

tion. Biotechnol J 3(9–10):1185–1200.

Zhou JX, Solamo F, Hong T, Shearer M, Tressel T. 2008b. Viral clearance

using disposable systems in monoclonal antibody commercial down-

stream processing. Biotechnol Bioeng 100(3):488–496.



anion exchange membrane adsorbers for flow-through polishing steps: part i. clearance of minute...

Documents