demonstration of a strategy for product purification by high-gradient magnetic fishing: recovery of...

Demonstration of a Strategy for Product Purification byHigh-Gradient Magnetic Fishing: Recovery of SuperoxideDismutase from Unconditioned Whey

Andrea Meyer,†,‡,| Dennis B. Hansen,‡,⊥ Claudia S. G. Gomes,‡ Timothy J. Hobley,‡Owen R. T. Thomas,*,‡,§ and Matthias Franzreb*,†

Institute for Technical Chemistry, Water- and Geotechnology Division, Forschungszentrum Karlsruhe,Hermann v. Helmholtz Platz 1, 76344 Eggenstein-Leopoldshafen, Germany; Center for Microbial Biotechnology,BioCentrum-DTU, Technical University of Denmark, Building 223, Søltofts Plads, DK-2800, Kgs. Lyngby,Denmark; and Department of Chemical Engineering, The University of Birmingham, Edgbaston,Birmingham B15 2TT, U.K.

A systematic approach for the design of a bioproduct recovery process employingmagnetic supports and the technique of high-gradient magnetic fishing (HGMF) isdescribed. The approach is illustrated for the separation of superoxide dismutase(SOD), an antioxidant protein present in low concentrations (ca. 0.15-0.6 mg L-1) inwhey. The first part of the process design consisted of ligand screening in which metalchelate supports charged with copper(II) ions were found to be the most suitable. Thesecond stage involved systematic and sequential optimization of conditions for thefollowing steps: product adsorption, support washing, and product elution. Next, thecapacity of a novel high-gradient magnetic separator (designed for biotechnologicalapplications) for trapping and holding magnetic supports was determined. Finally,all of the above elements were assembled to deliver a HGMF process for the isolationof SOD from crude sweet whey, which consisted of (i) binding SOD using Cu2+-chargedmagnetic metal chelator particles in a batch reactor with whey; (ii) recovery of the“SOD-loaded” supports by high-gradient magnetic separation (HGMS); (iii) washingout loosely bound and entrained proteins and solids; (iv) elution of the target protein;and (v) recovery of the eluted supports from the HGMF rig. Efficient recovery of SODwas demonstrated at ∼50-fold increased scale (cf. magnetic rack studies) in threeseparate HGMF experiments, and in the best of these (run 3) an SOD yield of >85%and purification factor of ∼21 were obtained.

1. IntroductionIn our laboratories we have recently introduced a new

robust isolation technique for macromolecules (1-7)termed High-Gradient Magnetic Fishing (HGMF). In atypical HGMF process, the target species is first boundonto functionalized magnetic supports in a simple batchadsorption reactor and the product-loaded adsorbents aresubsequently separated from the nonmagnetic compo-nents using high-gradient magnetic separation (HGMS)technology, adapted from the mineral processing andwastewater industries. The use of a simple and readilyscaleable batch-binding step for product adsorption makesthis technique particularly interesting for the treatmentof high volume feedstocks, such as whey. Previously wehave focused on the construction and characterization of

magnetic supports followed by a simple demonstrationof their use in an HGMF process (3, 5, 6). To date,however, it has not been demonstrated how an HGMFprocess can be developed systematically for the recoveryof a particular target molecule of interest.

The cheese industry produces excessive amounts ofwhey as a byproduct, e.g., 1.18 × 1011 kg of whey wasproduced worldwide in 1997, rising by 3% each year (8,9), and many scientists in collaboration with dairycompanies are looking for new ways to make use of thisexcess. In this context interest in recovering high-valuemilk proteins from this waste material for use in food,cell culture, and for pharmaceutical purposes has grownconstantly in recent years (9). One interesting targetmolecule in whey is the enzyme superoxide dismutase(SOD), which is present in low concentrations of the orderof 0.15-0.6 mg L-1 (10-13). SOD is present in almostall tissues of eukaryotes and is responsible in vivo forcatalyzing a vital reaction, namely, the dismutation ofhighly reactive superoxide anion radicals. A techniquefor the isolation of SOD from sweet whey is of interest,given that the antioxidant properties of the enzyme maymake it suitable for treating inflammatory diseases (14).Although Holbrook and Hicks (15) described a procedurefor the isolation of SOD from milk, the purpose of theirstudy was one of characterization rather than efficient

* To whom correspondence should be addressed. O.R.T.T.: Tel.+44 121 414 5278. Fax +44 121 414 5377. Email: [email protected]. M.F.: Tel. +49 7247 82 3595. Fax +49 7247 86 6660.Email: [email protected].

† Forschungszentrum Karlsruhe.‡ Technical University of Denmark.§ University of Birmingham.| Current address: Procter & Gamble Service GmbH, Sulz-

bacher Strasse 40, 65823 Schwalbach am Taunus.⊥ Current address: Alpharma ApS, Dalslandsgade 11, DK-2300,

Copenhagen, Denmark.

244 Biotechnol. Prog. 2005, 21, 244−254

10.1021/bp049656c CCC: $30.25 © 2005 American Chemical Society and American Institute of Chemical EngineersPublished on Web 12/29/2004

exploitation of the technique for preparative purposes.To date, most work on preparative recovery of SOD hasused blood as the source material, primarily owing to theseveral-hundred-fold higher concentration of the enzymein this feedstock (10, 13, 16).

In this paper we demonstrate how an HGMF processcan be designed when faced with the task of recoveringa “new” target molecule from a “new” feedstock. Simplescreening studies conducted with a clarified version ofthe intended feed and a selection of different candidateligands supported on conventional commercially availablechromatographic matrices were first used to identify asuitable ligand for binding of the target molecule. In thenext phases magnetic supports functionalized with thechosen ligand were manufactured, and their binding anddesorption performance were subsequently characterized/optimized in small-scale studies, using the unconditioned(i.e., crude and unclarified) feedstock. Finally, usingoptimized conditions acquired from the small-scale ex-periments, we demonstrate the piecing together of anHGMF process. The particular example we have chosenfor illustrating the design of an HGMF process is therecovery of SOD from crude sweet whey.

2. Materials and Methods2.1. Materials. The “430” stainless steel wire matrix

(KnitMesh type 9029) employed in HGMS and HGMFexperiments and the unconditioned crude sweet wheyused throughout this study were received as gifts fromColin Barnes (KnitMesh, South Croydon, Surrey, U.K.)and Waagner Nielsen (Royal Veterinary and AgriculturalUniversity, Copenhagen, Denmark), respectively. Thebicinchoninic acid (BCA) protein assay kit was purchasedfrom Pierce Chemicals (Rockford, IL), TMB-ONE ready-to-use substrate was supplied by Kem-En-Tec A/S (Copen-hagen, Denmark), and superoxide dismutase assay kit(cat. no. 574600) was obtained from Calbiochem-Nova-biochem (San Diego, CA). Bovine serum albumin ( BSA,Fraction V powder, A9647), superoxide dismutase (frombovine erthyrocytes, S2515), lactoperoxidase (from bovinemilk, L2005), lactoferrin (from bovine milk, L9507),immunoglobulin (from bovine milk, I5506), R-lactalbumin(from bovine milk, L5385), â-lactoglobulin (from bovinemilk, L0130), lysozyme (from human milk, L6394), andcatalase (from bovine liver, C9322) were all acquired fromthe Sigma Chemical Company (St. Louis, MO), as werethe chemicals (aminopropyl)triethoxysilane, glacial aceticacid, glutaraldehyde (50% photographic grade), sodiumborohydride, allyl glycidyl ether, N-bromosuccinimide,and iminodiacetic acid (disodium salt). The salts iron(II)chloride hexahydrate and iron(III) chloride tetrahydratewere supplied by J. T. Baker (Deventer, The Nether-lands). Methanol (GPR grade) was purchased from BDHLaboratory Supplies (Poole, Dorset, U.K.). All otherchemicals were obtained from either Merck (Darmstadt,Germany) or the Sigma Chemical Company.

2.2. Ligand Screening. Screening of a suitable ligandto recover SOD from whey was performed on a Gradifracsystem (Amersham Biosciences, Uppsala, Sweden) using1-mL HiTrap chromatography columns (Amersham Bio-sciences, Sweden) packed with two types of media,namely, Heparin Sepharose HP and Chelating SepharoseHP. Prior to use the latter immobilized metal affinitychromatography (IMAC) columns were charged witheither Cu2+, Ni2+, or Zn2+ ions using 0.1 M solutions ofcopper sulfate, nickel chloride, or zinc chloride, respec-tively. Before loading, all columns were equilibrated withan appropriate buffer, i.e., 10 mM sodium phosphate, pH7 for heparin columns and 10 mM potassium phosphate,

1 M NaCl, pH 6.4 for IMAC (17). Thirty column volumes(CVs) of clarified whey (prepared by filtering through a0.45-µm polyethersulfone membrane (PALL Cooperation,Ann Arbor, MI)) were then applied to columns at asuperficial linear flow velocity of 1.58 m h-1. Followingloading, columns were washed with 10 CV of the ap-propriate equilibration buffer and thereafter eluted at0.79 m h-1 with 10 CV of the same buffer supplementedwith either 1 M NaCl (for heparin) or 1 M NH4Cl (forIMAC). All collected fractions were analyzed for proteincontent and SOD activity as described under AnalyticalMethods.

2.3. Preparation of Cu2+- Charged Magnetic MetalChelator Particles. The detailed methods for thepreparation of Cu2+-charged “type II” magnetic metalchelator particles used in this work have been presentedelsewhere (3, 6). These magnetic support materialsemploy the tridentate chelating agent, iminodiacetic acid(IDA), which is covalently attached to the coated particlesurfaces via a 7-atom hydrophilic spacer arm. Immedi-ately prior to use the IDA-linked supports were chargedwith Cu2+ ions exactly as described by Heebøll-Nielsenet al. (6), i.e., by washing them three times in a 0.1 Mcopper sulfate solution, followed by three washes in 10mM potassium phosphate, 1 M NaCl, pH 6.4. Fourseparate batches of Cu2+-IDA magnetic adsorbents werefabricated: one using ∼13 g IDA per g of magneticsupports in the coupling reaction, designated LS for “lowsubstitution”, and three “high substitution” batches (HS1,HS2, and HS3) prepared using a 4-fold higher IDA toparticle ratio during coupling of the chelating agent. Tominimize possible inconsistencies arising from batch-to-batch differences, adsorbent particles from just a singlebatch were used in each experiment. Throughout allstages of adsorbent preparation, a ∼0.7 T neodymium-iron-boron permanent magnet block (Danfysik A/S,Jyllinge, Denmark) was used for recovery of magneticsupport particles from suspension.

2.4. BSA Binding Tests. Batches of magnetic adsor-bent particles were subjected to a simple quality controltest, which involved characterization of their BSA ad-sorption properties. Seven-milligram portions of magneticchelator supports (charged with Cu2+ or uncharged) were(i) magnetically retrieved from storage buffer (20 mMsodium phosphate, pH 6.8 containing 1 M NaCl), (ii)equilibrated thoroughly by resuspension in a bindingbuffer composed of 10 mM potassium phosphate, pH 6.4,supplemented with 1.0 M NaCl, (iii) reseparated, andthen (iv) incubated at room temperature (22 °C) with1-mL aliquots of BSA solutions (of varying concentrationand made up in binding buffer) on an IKA VXR-S17vibrating shaker (IKA Labortechnik, Staufen, Germany)operated at ∼800 rpm. After 600 s of incubation thesupports were separated and the supernatants wereassayed for residual protein content (see AnalyticalMethods). The amounts of adsorbed BSA were calculatedby difference, and adsorption isotherms were subse-quently plotted and fitted to the Langmuir (18) and bi-Langmuir models. In the simple monomodal Langmuirexpression (eq 1), the equilibrium concentrations of theadsorbed and bulk-phase protein are represented by Q*and C*, respectively. Binding is fully described by twoterms, a dissociation constant, Kd, and Qmax, the maxi-mum capacity for the adsorbed protein. In the bi-Langmuir model (eq 2), two types of binding sites areassumed, namely, “tight” and “weak”, and these aredescribed by the dissociation constants KdA and KdB,respectively. The corresponding maximum protein bind-ing capacity at each of these sites are given by QmaxA and

Biotechnol. Prog., 2005, Vol. 21, No. 1 245

QmaxB, and the sum of the two capacities equals the totalcapacity, Qmax.

Collected data were fitted to these models using theø2 minimization procedure of Microcal Origin softwareversion 4.1.

2.5. SOD Batch Adsorption/Desorption Studies.Small-scale SOD batch binding and elution studies werecarried out in 2-mL eppendorf-style tubes using whey anddefined amounts of the aforementioned Cu2+-IDA linkedmagnetic adsorbent. As a starting point for methoddevelopment the basic buffer employed in these experi-ments was the same as that used in the chromatographicscreening study described earlier, i.e., 10 mM potassiumphosphate, pH 6.4. This was supplemented with up to 1M NaCl for equilibration/binding and washing operationsor with various concentrations of ammonium chloride(0.2-1 M NH4Cl) for elution. Between each step, thesupports were separated magnetically, using side-pullmagnetic racks (Perseptive Biosystems, Framingham,MA), to allow the supernatant to be removed before thenew buffer was added. The supports were equilibratedin binding buffer prior to resuspending in whey andincubating for 600 s at ∼800 rpm on a vibrating shaker.After the batch binding step, the supernatants werecollected, and the supports were subsequently washedtwice with 1 mL of the appropriate washing buffer for30 s. Finally, bound proteins were desorbed in either oneor two successive rounds of elution, by incubating thewashed “product-loaded” supports with elution buffer for600 s at 800 rpm on a vibrating shaker, before magneti-cally retrieving the particles, and removing the super-natant(s). All of the above operations were carried outat a temperature of ∼22 °C.

2.6. High-Gradient Magnetic Fishing SystemSetup. A schematic illustration of the HGMF systememployed in this work is shown in Figure 1. A distinctionof the present study, cf. our previous work on HGMF, isthat it is the first to feature a magnet system designedspecifically for use in bioprocessing (19). A photographof this laboratory-type permanent magnet separator(Steinert HGF-10, Steinert Elektromagnetbau GmbH,Koln, Germany) is shown in Figure 1, and a schematicillustration of how it works is presented in Figure 2. Themagnetic flux density in the 1.5 cm air-gap between thepoles was determined with the aid of a gaussmeter(Lakeshore model 410, Westerville, OH) fitted with atransverse probe, as 0.56 and 0.03 T in the “on” and “off”positions, respectively. A filter canister (4.4 mL, 56 mmlong × 10 mm i.d.) was filled with a rolled mat of wovenmesh composed of “430” stainless steel fibers (∼110 µmthickness) to occupy ∼11% of the working volume (i.e.,to give a voidage of 0.89 and void volume of 3.9 mL)before positioning vertically between the pole shoes. Theprocess setup consisted of (i) a stirred batch adsorptionreactor; (ii) the magnet and the filter canister mentionedabove; (iii) two peristaltic pumps (Masterflex L/S Easy-Load model 7518-00, Cole Parmer Instruments Co.,Vernon Hills, IL, MA); and (iv) a model 2212 Heliracfraction collector (LKB Bromma, Bromma, Sweden). Thedifferent flow directions used for particle loading, wash-ing, product elution, and particle recovery were controlled

by manual three-way (2 mm i.d.) Teflon valves (Kebo LabA/S, Albertslund, Denmark). During loading of themagnetic filter with the field on, the particle/feedstocksuspension was fed through valves 1-3 via pump 1. Forwashing and protein elution, a recycle-loop was createdby changing valves 2 and 3 so that after filling the loopvia valve 4, the liquid could be pumped in a circle withpump 2 after switching the field off. The field was thenswitched back on, and after changing valve 3, the washedoff or eluted material was collected. Finally, switchingoff the field while pumping allowed the supports to berecovered from the HGMF rig.

Prior to carrying out HGMF recovery of SOD thesupport trapping characteristics of the magnetic filterwere examined in breakthrough studies, which wereconducted using magnetic support particles (amine-terminated intermediates and finished adsorbents) sus-pended in various liquors (i.e., different buffers or whey)at a final concentration of 7 g L-1. With the field switchedon these suspensions were fed to the magnetized filterat a linear flow rate of 24 m h-1. The breakthrough ofparticles in the filter effluent was monitored by gravi-metric measurement of the particle mass in collectedsamples, as detailed in Analytical Methods.

Q* ) QmaxC*

Kd + C*(1)

Q* ) QmaxAC*

KdA + C*+ QmaxB

C*KdB + C*

(2)

Figure 1. Photograph of the laboratory type (SteinertHGF-10) switch on-off permanent magnet separator (top)and schematic representation of the high-gradient magneticfishing system (bottom). Key: BA, batch adsorption reactor;EB, elution buffer; FC, fraction collector; MF, magnetic filter;RL, recycle loop, P1 and P2, pumps, V1-V5, valves; WB, washbuffer.

246 Biotechnol. Prog., 2005, Vol. 21, No. 1

The extent of in-system mixing during transition fromone buffer to another in an HGMF experiment was deter-mined in the following way. A UV1 detector (AmershamBiosciences, Uppsala, Sweden) was inserted into theHGMF system immediately after valve V3 (Figure 1). Thefilter canister was filled to 10% breakthrough capacitywith amine-terminated magnetic particles suspended inbuffer 1 (20 mM Tris-HCl buffer, pH 8) as describedabove. After loading, the system was washed with thefield switched on. Then, with the field still on buffer 2,i.e., buffer 1 supplemented with 0.5 M NaCl and 10% (v/v) acetone, was applied to the canister until the acetoneconcentration measured by the UV monitor reached 1%(v/v), i.e., 10% breakthrough. At this point, pumping wasstopped, and valves V3 and V4 were closed to isolate therecycle loop (Figure 1), before switching the field off andchasing the particles out of the filter and around the loopin the reverse direction, at 92 m h-1 for 600 s. Supportswere subsequently recaptured within the filter by switch-ing the field back on, and the washings were sent out ofthe HGMF rig to the fraction collector. The recycle loopwas then refilled with the same volume of buffer 2 asthat determined in the previous cycle, and a second runwas performed in a manner identical to that just de-scribed for the first. The conductivities of the first andsecond buffer pools collected from the HGMF recycle loopwere measured off-line using a Meterlab model CDM210conductivity meter (Radiometer, Brønshøj, Copenhagen),and the NaCl concentration in each was determined byreference to a standard curve constructed using 0-0.5M NaCl suspended in 20 mM Tris-HCl pH 8 containing10% acetone (v/v). Comparison of the NaCl concentrationin the first and second buffer pools emerging from the

HGMF system employed in this study, with the bufferinitially applied, showed 15% and 3% drops, respectively.

Between experiments the filter canister was discon-nected from the HGMF system and cleaned by flushingin both directions with water at high flow rate for a periodof at least 600 s.

2.7. Recovery of SOD from Whey using the HGMFProcess. Equilibrated Cu2+-charged magnetic metalchelator supports were resuspended in sufficient wheyto give a particle concentration of 7 g L-1 and thereaftermixed at room temperature with an overhead stirrer for600 s. Subsequently, the particle/whey suspension waspumped upward through the magnetic filter canister ata linear flow rate of 24 m h-1, while the magnetic fieldwas switched on. Shortly before particle breakthroughwas expected (determined in the preceding filter char-acterization experiments) pumping was stopped. Therecycle loop was then filled with washing buffer (10 mMpotassium phosphate, 30 mM NaCl, pH 6.4) and afterturning the field off, the suspension was pumped throughthe recycle loop under reversed flow at a velocity of 92m h-1 for 120 s to wash out entrained and/or looselyadsorbed materials. The particles were subsequentlyrecaptured, by turning on the magnetic field, and the“dirty” washing buffer was pumped out of the HGMF rig.A second cycle of washing was performed before com-mencement of SOD elution. Bound SOD was desorbedfrom the particles in two cycles of 600 s each, in a fashionanalogous to the washing procedures just described, byfilling the recycle loop with elution buffer (100 mMpotassium phosphate, 0.8 or 1.2 M NH4Cl, pH 6.4), andrapidly circulating the particles around the closed systemloop.

2.8. Analytical Methods. Particle concentration wasdetermined by dry weight measurements. Samples con-taining magnetic particles were applied to (predried andweighed) 0.45-µm filters (PALL Corporation, Ann Arbor,MI) and filtered under vacuum while being flushed withlarge amounts of water to release entrained salts orsolids. Afterward the filters were dried in a microwaveoven at 300 W for 600 s. After cooling in a desiccator thefilters were weighed, and the particle concentration inthe samples was calculated from the difference in weightof the filter before and after sample application.

Soluble protein (expressed in BSA equivalents), SOD,and lactoperoxidase (LPO) contents in samples weredetermined using, respectively, the bicinchoninic acid(BCA) protein assay, superoxide dismutase assay kit, andTMB-ONE ready-to-use substrate. All three assays werescaled for use in a Cobas Mira spectrophotometric robot(Roche Diagnostics, Switzerland).

The basis for the commercial SOD assay employedhere, and first described by Nebot et al. (20), is that SODaccelerates the alkaline autoxidation of 5,6,6a,11b-tet-rahydro-3,9,10-trihydroxybenzo[c]fluorene (BXT-01050)to a chromophore possessing an apparent εmax at 525 nm.The ratio of the autoxidation rates measured in thepresence and absence of SOD gives a measure of SODactivity, and 1 unit of SOD activity is defined as thatwhich doubles the autoxidation rate of the control blank(20). Although interference arising from the presence ofbiological mercaptans in samples is eliminated by theincorporation of the mercaptan scavenger 1-methyl-2-vinylpyridinium trifluoromethanesulfonate in the assay,the presence of certain proteins can also cause significantartifacts (20 and references therein). Accordingly, priorto employing the SOD assay on samples of crude wheyand fractions collected during SOD purification, we testedit on the following purified proteins: â-lactoglobulin (3

Figure 2. Schematic top view of the switch on-off permanentmagnet system and the separation unit. (a) Field on position.(b) Field off position. Key: 1, blocks of permanent magnetmaterial; 2, circular shaped subyoke; 3, iron yoke; 4, pole shoe;5, separation unit housing; 6, separation matrix.


g L-1), immunoglobulin (1 g L-1), R-lactalbumin (0.7 gL-1), serum albumin (0.3 g L-1), lactoferrin (0.02-0.35 gL-1), LPO (0.01-0.03 g L-1), and lysozyme (<0.001 g L-1),with each being used at concentrations reflecting theirreported contents in bovine whey (21, 22). Of these, onlyLPO was observed to interfere with the assay; itspresence in samples resulted in a concentration-depend-ent “false positive” SOD activity. For the determinationof “true” SOD activity in samples (i.e., due to SOD alone),the correction of the apparent SOD activity for falsepositive contributions due to the presence of LPO provednecessary and was carried out as follows: (i) The preciseLPO content in a given sample was determined with theLPO assay (see below). (ii) The false positive contributionof that amount of LPO to the sample’s apparent SODactivity was then calculated by reference to a standardcurve of apparent SOD activity vs LPO content. (iii) Thetrue SOD activity was subsequently obtained simply bysubtracting the estimated contribution due to LPO fromthe SOD measurements on the sample. For crude bovinewhey correction of the apparent SOD activity (0.857 (0.026 U mL-1, n ) 7) for false positive interference fromendogenous LPO gave a mean value for the true SODactivity of 0.784 ( 0.023 U mL-1 (n ) 7), i.e., ∼91% ofthe measured SOD activity was judged to be true andjust 8% as arising from LPO. A similar value for the trueSOD activity in whey was obtained using a differentapproach, i.e., adding sufficient catalase to knock out thefalse positive interference caused by LPO (20). Whencatalase was added to crude whey at 50, 100, and 150 UmL-1 of sample, the apparent SOD activity dropped to92.9 ( 0.5% (n ) 3) of its initial value.

In the assay for LPO activity, LPO catalyzes theoxidation of 3,3′,5,5′-tetramethyl benzidine. The forma-tion of the deep blue product was followed at 37 °C bymonitoring the change in absorbance at 600 nm. Incontrast to the false positive activity shown by LPO inthe SOD assay, pure SOD registered no apparent LPOactivity, even when used at a concentration as high as0.52 g L-1.

In this work we have employed “yield factors” (frac-tional yield × purification factor) recommended by Hearleet al. (23) to aid selection of the best combinations of SODpurity and recovery.

3. Results and Discussion

3.1. Ligand Selection. Both heparin affinity (24, 25)and immobilized metal affinity principles (26) have beenemployed previously for the chromatographic purificationof SODs from various sources, but a comparative evalu-ation of these two techniques for SOD isolation from wheyhas not been reported until now. The results of chro-matographic screening studies performed on HiTrapimmobilized metal affinity chromatography (IMAC) andheparin columns (Table 1) confirm that the best yield,purification, and yield factors were obtained by IMAC

on Cu2+-IDA Sepharose HP, with the second best per-formance being obtained by the Heparin Sepharose HPmatrix. Conversely, the performance of both Ni2+- and,to an even greater extent, Zn2+-charged IMAC columnsfor which no binding was detected (results not shown)was very poor and did not warrant further study. In viewof the superior binding performance of the Cu2+-IDAsupport, combined with the ready availability, low cost,and robustness of the synthetic metal chelating agent,iminodiacetic acid (IDA), the latter was chosen in prefer-ence to heparin as the ligand for construction of magneticadsorbents with selectivity for SOD.

3.2. BSA Binding Tests. Following their manufactureand prior to use in studies with unconditioned whey, thevarious batches of Cu2+-IDA magnetic adsorbent particleswere first subjected to a performance/quality control test,which involved determining adsorption isotherms withbovine serum albumin (BSA), a protein known for itsability to bind to Cu2+-IDA supports (27). Figure 3 showsthe adsorption isotherms obtained with the variousbatches of magnetic chelator particles used in this study.Strong and high capacity binding of BSA was observedwith the Cu2+-charged supports. This adsorption behaviorwas clearly mediated by the presence of chelated Cu2+

ions, as only very low levels of BSA binding were notedfor uncharged supports (Figure 3a). The Cu2+-chargedBSA binding data were fitted to Langmuir (eq 1) and bi-Langmuir (eq 2) models, and resulting parameters ob-tained for the curves presented in Figure 3 are shown inTable 2. The gross similarity of the three Cu2+-chargedhigh sub batches (HS1, HS2, and HS3) was confirmed bythe observation that their BSA binding data collapsedalong common Langmuir and bi-Langmuir fit curves(Figure 3a). Evidently the LS adsorbent possessed some-what reduced sorption performance for BSA comparedto its HS counterparts. Although at first glance thesimple Langmuir isotherm yielded apparently satisfac-tory fits (Figure 3a, broken lines), Scatchard (28) analysis(Figure 3b) gave strongly curved plots consistent withat least two different types of binding affinities. Notsurprisingly, therefore, remodeling the data to the bi-Langmuir isotherm (Figure 3a and b, solid lines) resultedin superior fits. The Langmuir fits grossly underesti-mated the tightness of BSA binding (initial slopes of the

Table 1. Summary of Data from Recovery of SOD fromClarified Sweet Whey on Heparin Sepharose HP andM2+-Charged Chelating Sepharose HP HiTrap ScreeningColumns

purification step%

yield

specificactivity

(U mg-1)purification

factoryield

factora

clarified whey 100 0.075 1.0 1.0Heparin Sepharose HP 67.3 1.30 17.4 11.7Cu2+-Chelating Sepharose HP 70.5 2.72 36.4 25.1Ni2+-Chelating Sepharose HP 2.8 0.92 12.3 0.34

a Yield factor ) fractional yield × purification factor (ref 23).

Figure 3. (a) Equilibrium adsorption of BSA on magneticchelator particles. (b) Scatchard plots showing heterogeneousadsorption of BSA to Cu2+-charged magnetic chelators. Key: (b)Cu2+-charged LS, (4) HS1, (0) HS2, and (3) HS3 magneticchelators; Uncharged (O) LS and (9) HS2. Broken and solid linesthrough the data represent the fit of Langmuir (eq 1) and bi-Langmuir (eq 2) isotherms, respectively, with the parameterscited in Table 2.


isotherms), and differences between the LS and HSadsorbents were totally obscured (Table 2, initial slopesof ∼0.2 L g-1 for both Cu2+-charged supports). By contrastthe bi-Langmuir model not only revealed that BSAbinding of both adsorbent types was much tighter, butalso that the Cu2+-HS material was roughly twice aspotent as the Cu2+-LS variant (Table 2, initial slopes )15.2 vs 7.9 L g-1).

3.3. Characterization/Optimization of SubstepsThat Make Up HGMF. Having selected a suitableligand, manufactured magnetic adsorbents featuring it,and confirmed the suitability of the latter materials forprotein recovery by immobilized metal affinity adsorp-tion, the next tasks en route to construction of an HGMF-based SOD extraction process were systematic charac-terization and optimization of each of the substeps (i.e.,product adsorption, adsorbent collection, adsorbent wash-ing, and product desorption). With the exception of theadsorbent collection study (3.3.3), all of this work wasperformed at small scale, i.e., using eppendorf vials andpermanent magnet racks. The equilibration, wash, andelution conditions previously employed to good effect inthe recovery of SOD from clarified whey by chromatog-raphy on Cu2+-Chelating Sepharose HP (see Table 1)were initially applied in these experiments, but weresystematically altered as optimal conditions for each ofthe steps in HGMF (i.e., binding, washing, and elution)were identified.

3.3.1. SOD Adsorption. Numerous other whey pro-teins present at concentrations much higher (21, 29) thanthat of SOD, e.g., serum albumin (27), immunoglobulins(30, 31), lactoferrin (30, 32, 33), LPO (31, 34), lysozyme(33, 35), R-lactalbumin (7, 36), and â-lactoglobulin (7, 35)are also known for their ability to bind to immobilizedmetal affinity adsorbents. Not surprisingly, therefore,when suspended in whey, the Cu2+-charged magneticadsorbents bind much more of these species than SOD.SDS-PAGE analysis (results not shown) confirmed thatlactoferrin, LPO, and immunoglobulin were substantiallyenriched in eluates from Cu2+-charged supports afterhaving been contacted with crude whey. R-Lactalbuminand â-lactoglobulin had also been bound in significantamounts, albeit less specifically, but BSA had not beenadsorbed. Given SOD’s very low concentration in whey(0.15-0.6 mg L-1) and that it runs in exactly the sameposition as the most abundant whey protein species,â-lactoglobulin (present at 3 g L-1), under both reducingand nonreducing conditions, visual detection of the targetenzyme in Coomassie blue and silver stained SDS-polyacrylamide gels was not possible. Attempts to detectSOD’s presence in crude samples using specific zymo-graphic (a two-stage technique that involves proteinseparation by electrophoresis, followed by in situ assayof enzyme activity) methods (37) developed for and testedon systems rich in endogenous SODs (e.g., bovine blood,E. coli extracts) were likewise compromised by the verylow levels of SOD in all whey-derived samples.

In view of the above in small-scale batch bindingexperiments, we first determined the optimal magnetic

adsorbent concentration for SOD recovery from uncon-ditioned whey. Figure 4 shows that virtually completeremoval of the SOD present in the crude whey feedstockcould be achieved by using the LS adsorbent at aconcentration of close to 12 g L-1, whereas the use of HSparticles allowed the same result to be achieved at lowersupport concentrations (typically ∼7 g L-1). Importantly,despite huge competition for binding sites on the Cu2+-charged LS and HS supports from the other metal chelatebinding proteins present in whey, complete removal ofSOD was attained at the expense of only a ∼15% dropin the concentration of total whey protein (Figure 4).Clearly therefore, preferential adsorption of SOD vs otherwhey proteins had been achieved with both adsorbenttypes. For all further experiments on development of aHGMF process, HS supports were selected in preferenceto the LS materials, and these were employed at theminimum concentration required for quantitative re-moval of SOD from whey, i.e., 7 g L-1. The impact of thischoice on the performance of the resulting HGMF processis highly significant. For example, in HGMF use of theLS adsorbent for the above task would entail >1.7-foldfaster exhaustion of the magnetic filter’s capacity foradsorbent particles and a >40% reduction in the volumeof feedstock that could be treated per cycle.

3.3.2. Switch On-Off Permanent Magnet Separa-tor. Using variable field electromagnets we have pre-viously determined that the magnetic particles de-scribed in this work can be collected by HGMS frombiological suspensions flowing at 100 m h-1 using moder-ate field strengths (e.g., ∼0.4 T) that are easily suppliedby rare earth permanent magnets (1, 2). A distinction ofthe HGMF system featured in this work (section 2.5,Figure 1) is that it is built around a cyclically operatedon-off permanent magnet (19). This separator wasdesigned with bioprocessing in mind, and this study isthe first to report its use for bioproduct purification byHGMF.

The unusual capacity to turn a permanent magnet onand off at will is realized by mounting an arrangementof magnet blocks within the center of a cylindrical ironsubyoke, which can be rotated along its central axis,within a fixed iron yoke (Figure 2). In the on position(Figure 2a) the system resembles a conventional magnetyoke and generates a flux density in the gap between thepole shoes of ∼0.56 T. Rotation of the cylinder stepwisethrough 90° (to the off position; Figure 2b) reduces theflux density at the pole region to ∼0.03 T, the mainreason for this being that the magnetic flux lines areshort-circuited by the iron yoke. The main advantagesof this particular magnetic separator design over systemsemploying electric coils include the absence of a dedicatedpower supply (i.e., no heat generation), low operatingcosts, and easy accessibility to the magnet field region.Moreover, the small size of the lab prototype (∼70 kg)used here makes for an easily transported system,allowing small-scale HGMF trials with real industrialfeedstocks to be performed on-site at interested biotechcompanies.

Table 2. Estimated Parametersa for Adsorption of BSA on Type II Cu2+-Charged Magnetic Metal Chelators

Langmuir model (eq 1) bi-Langmuir model (eq 2)

supporttype

Qmax(µmol g-1)

Kd(µM)

initialslope (L g-1)b

QmaxA(µmol g-1)

KdA(µM)

QmaxB(µmol g-1)

KdB(µM)

Qmax(µmol g-1)c

initialslope (L g-1)d

LS 1.26 6.4 0.197 0.58 0.076 0.95 4.1 1.53 7.86HS 1.56 7.4 0.211 0.33 0.023 1.32 1.5 1.65 15.15

a Figure 3 adsorption isotherm data were fitted using the ø2 minimization procedure of Microcal Origin software version 4.1; b Initialslope ) Qmax/Kd. c Qmax ) QmaxA + QmaxB. d Initial slope ) QmaxA/KdA + QmaxB/KdB.


3.3.3. Adsorbent Collection by HGMS. Having se-lected an appropriate concentration of magnetic adsor-bent particles to employ during batch binding operation,the next step was to examine their collection by HGMS.The volume of particle-liquid suspension processed percycle of the HGMF system is constrained by the supportholding capacity of the magnetic filter. Under idealconditions the filter will capture 100% of the incomingmagnetic support particles until a certain capacity isreached, after which sudden breakthrough will occur andall new particles entering the filter pass through. Inreality, however, some magnetic particles will escape thefilter before its total capacity is attained. The filter shouldtherefore be operated in such a way that the highestpossible support holding capacity is achieved at theexpense of minimal support loss. In much of our previouswork on the characterization of magnetic particle break-through in simple buffered solutions and under variouscombinations of magnetic flux density and fluid velocity,instead of “finished” magnetic adsorbent particles, wehave employed a defined intermediate in their manufac-ture, namely, an amine-terminated magnetic particle ofvery similar size and magnetic properties (1-6). Theavailability of this “model” magnetic material in muchlarger amounts than the finished adsorbents permittedmore extensive trials to be conducted. In this work wehave carried out HGMS breakthrough studies with bothtypes of magnetic supports, i.e., the amine-terminatedmodel material and the Cu2+-charged magnetic chelators,and the results are presented in Figure 5. When sus-pended in 10 mM potassium phosphate, pH 6.4 basedequilibration or elution buffers (containing either 1 MNaCl or 0.8 M NH4Cl, respectively), the particle break-through profiles for fresh Cu2+-charged magnetic chela-tors and both fresh and recycled amine-terminatedsupports were, for practical purposes, identical, withparticle breakthrough occurring shortly after a filtercapacity of 110 g L-1 was reached. That the breakthroughbehavior of the amine-terminated support was unalteredafter being subjected to a second HGMS particle capturecycle is testament to the fact that no adsorbent ag-glomeration occurred during the first and is entirely inkeeping with the near ideal superparamagnetic proper-ties (i.e., absence of magnetic memory) displayed by thesemagnetic support particles (1, 4, 5). In stark contrast tothe above, the breakthrough behavior of amine-termi-nated and Cu2+-charged supports were very different to

one another when the suspending phase was switchedfrom equilibration buffer to unconditioned whey. In thelatter feedstock, the breakthrough of amine-terminatedparticles occurred much earlier (i.e., ∼90 g L-1 vs 110 gL-1 in buffer), whereas that of the Cu2+-charged adsor-bents was noticeably delayed (∼125 g L-1). The prema-ture breakthrough observed with the amine-terminatedsupports suspended in the low ionic strength wheyfeedstock (4.4 mS cm-1 vs 80.3 mS cm-1 for equilibration/wash buffer) most likely reflects some degree of electro-static interaction between the positively charged supportparticles and negatively charged insoluble componentspresent in the feedstock. It is plausible that such aninteraction would lead to entrainment of insoluble wheymaterial within the cake of magnetic supports collectedin the magnetic filter, increasing the space occupiedwithin the filter, which in turn would result in a loss ofthe latter’s magnetic particle holding capacity. An ex-planation for the apparent increase in capacity of themagnetic filter for the Cu2+-charged supports whensuspended in whey versus equilibration buffer is lessobvious. We have previously noted that in addition tochanges in suspending solution properties and chemicalderivatization of the surfaces of magnetic supports, theadsorption of protein to magnetic adsorbent particlesexerts a powerful effect on the zeta potential of the latter(1). For example, following the adsorption of trypsin tobenzamidine-linked magnetic particles at the binding pHof 7.5, the magnitude of the net surface charge droppedfrom -30 to -15 mV. In the present case, it is conceivablethat, following adsorption of soluble proteins from thewhey suspension, the zeta potential of Cu2+-chargedmagnetic adsorbents would be expected to move closerto zero. This in turn should permit closer particle-particle approach and could, at least in part, account forthe observed increase in particle holding capacity of thefilter (Figure 5) compared to the same particles sus-pended in a protein-free equilibration buffer.

3.3.4. SOD Desorption. An important feature of thepresent HGMF system is that once captured within thefilter all subsequent operations involve the use of an in-system recycle loop. Although immediately following

Figure 4. Effect of Cu2+-charged LS and HS type II magneticchelator particle concentration on the removal of SOD fromunconditioned whey. Key: (O, b) LS; (2, 4) HS1; (0, 9) HS2;and (3, 1) HS3 Cu2+-magnetic chelators; SOD activity (opensymbols); soluble protein content (filled symbols).

Figure 5. Influence of support surface chemistry and suspend-ing phase on magnetic support capture by HGMS. Key: freshamine terminated particles suspended in (O) equilibration/washbuffer, (4) elution buffer, and (b) unconditioned whey; recycledamine terminated particles suspended in (3) equilibration/washbuffer; fresh HS3 Cu2+-charged magnetic chelators suspendedin (0) equilibration/wash buffer and (9) unconditioned whey.S0 ) support concentration applied to filter; S ) supportconcentration exiting filter. Equilibration/wash buffer: 10 mMpotassium phosphate, 1 M NaCl, pH 6.4. Elution buffer: 10 mMpotassium phosphate, 0.8 M NH4Cl, pH 6.4.


collection of adsorbents the particle concentration withincurrent filters can be >100 g L-1, the particle concentra-tion during washing and elution cycles drops consider-ably, the extent being determined by the size of therecycle loop relative to that of the filter. In this work thevoid volume of the filter measured 3.8 mL, that of therecycle loop (including valves) is 12.2 mL, and that of thefilter and loop combined is 16 mL. Thus, on loading thefilter to a capacity of 100-125 g L-1 and releasing/recycling the trapped adsorbents into the loop, the meanparticle concentration would be expected to be in therange of 20-30 g L-1. With this in mind, in subsequentsmall-scale experiments conducted with magnetic racks,aiming to identify washing and elution conditions suit-able for use in HGMF, the support concentration duringbinding was maintained at 7 g L-1, but washing andelution were performed at 3- to 4-fold higher adsorbentconcentrations, to simulate conditions likely to exist inthe recycle loop during HGMF.

Conditions for efficient elution of SOD from washedproduct-loaded adsorbents were systematically identifiedin two stages. In the first of these, the effect of NH4Clconcentration on the efficiency of SOD elution from twice-washed supports was examined (Figure 6, black bars),whereas in the second the impact of increasing bufferstrength was studied (Figure 6, white bars). The easewith which SOD was desorbed into the bulk liquid phasewas strongly dependent on the ammonium chlorideconcentration of the latter. For example, efficient recoveryof SOD (>75% in one step) from the supports requiredNH4Cl concentrations of at least 0.8 M. Variation in thepotassium phosphate concentration (10-200 mM potas-sium phosphate, pH 6.4) in the presence of 0.8 M NH4Cl(Figure 6, white bars) exerted a less pronounced effecton SOD desorption. A 10-fold increase in potassiumphosphate concentration from 10 mM initially to 100 mMraised the relative elution efficiency 1.2-fold, but withfurther increase to 200 mM this gain was almost com-pletely lost.

3.3.5. Adsorbent Washing. Following adsorption ofSOD from whey and prior to elution, we noted alarming

levels of product loss (i.e., ∼30-50% of that adsorbed)during adsorbent washing steps, which were performedwith the standard equilibration buffer (Table 3). Loss ofthe enzyme from loaded magnetic particles during wash-ing could, however, be reduced down to very low levels(<0.7% of that adsorbed), simply by reducing the con-centration of added NaCl from 1 M to 30 mM, therebylowering the buffer conductivity from 80.3 mS cm-1 to alevel matching that of the whey feedstock (i.e., 4.4 mScm-1) used in the batch adsorption step. The positiveimpact of this change in wash buffer on SOD purificationperformance (i.e., yield, purification, and yield factor) isclearly observed in Table 3.

3.4. Recovery of SOD from Unconditioned Wheyby HGMF. With suitable conditions for the HGMFrecovery of SOD from crude whey defined, three separateHGMF runs were conducted. Table 4 presents the dataobtained from analysis of the first HGMF run, and acomparison of the three runs is summarized in Table 5.In runs 1 and 3, following the adsorption step SOD-loaded adsorbents were applied to the filter to a capacityof ∼80 g L-1 such that during washing and elution theparticle concentration was ∼21 g L-1, whereas in run 2the filter loading capacity was increased to ∼86 g L-1,giving an adsorbent concentration in the recycle loop ofnearly 24 g L-1. The ammonium chloride concentrationof the elution buffer used for filling the rig’s recycle loopwas 0.8 M in both elution cycles in runs 1 and 2, but wasraised to 1.2 M in run 3. A common pattern was observedin all three runs, which were all conducted with the samebatch of Cu2+-charged adsorbents (HS3). Analysis of theflow through fractions showed that ∼85% of the totalprotein was lost, but that no SOD was detected (Table4). This was not surprising given that at the adsorbentconcentration of 7 g L-1 employed in the binding step thetarget enzyme is completely adsorbed from the crudewhey (Figure 4). In two subsequent washing cycles,performed within the closed recycle loop system usingthe modified wash buffer, further removal of entrainedand/or loosely adsorbed protein (>7% of that initiallypresent) was achieved at the expense of only ∼0.6% lossof SOD. In run 1, 76.5% of the initially present SOD wasrecovered in two elution cycles with a purification factorof 14.1 and yield factor of 10.8. SOD is an exceptionally

Figure 6. Effect of NH4Cl concentration and buffer strengthon the efficiency of SOD desorption from twice-washed Cu2+-magnetic chelators. Following SOD adsorption supports werewashed twice with 10 mM potassium phosphate, 1 M NaCl, pH6.4, and then eluted with 0.2-1.0 M NH4Cl in 10 mM potassiumphosphate, pH 6.4 (black bars) or 10-200 mM potassiumphosphate buffer pH 6.4 in the presence of 0.8 M NH4Cl (whitebars). The amounts of SOD released are expressed as percent-ages of the total bound prior to elution. HS1 and HS3 particleswere used in the “change in [NH4Cl]” and “change in bufferstrength” series, respectively. The particle concentration duringbinding was 7 g L-1, but was raised to 30 g L-1 during washingand elution. Over 30% of adsorbed SOD was lost in two washesof 10 mM potassium phosphate, 1 M NaCl, pH 6.4 prior toelution.

Table 3. Impact of Washing Buffer on Recovery of SODa

from Unconditioned Whey Using Magnetic Cu2+-ChargedHS3 Particlesb

washing buffer

parameter standardc modifiedd

% yield:wash 1 27.6 ∼0.1wash 2 20.1 ∼0.5combined washes 47.7 <0.7elution 1 25.8 34.0elution 2 16.7 19.5combined elutions 42.5 53.5

purification factor:combined washes 6.2 <0.2combined elutions 12.3 12.4

yield factor:ecombined washes 3.0 <1.3 × 10-3

combined elutions 5.2 6.6a The amounts of SOD released are expressed as percentages

of the total adsorbed (∼99.9%) from the whey. b The particleconcentration during binding was 7 g L-1, but was raised to 30 gL-1 during washing and elution. Elution was performed with 100mM potassium phosphate, pH 6.4, + 0.8 M NH4Cl. c 10 mMpotassium phosphate, pH 6.4 + 1 M NaCl (80.3 mS cm-1) d 10mM potassium phosphate, pH 6.4, + 30 mM NaCl (4.4 mS cm-1).e Yield factor ) fractional yield × purification factor (ref 23).


robust enzyme (38), and so the activity unaccounted formost likely reflects that the missing SOD is still boundto the adsorbents and could be recovered in furtherelution cycles.

Increasing the particle concentration from 21.4 g L-1

in the first run to 23.7 g L-1 in the second resulted inslightly inferior purification performance; the purificationfactor remained essentially unchanged, but both the yieldand yield factor dropped very slightly to 73.9% and 10.3respectively (Table 5). In runs 1 and 2 the elution bufferemployed for filling the recycle loop contained ammoniumchloride at a concentration of 0.8 M. However, dilutionof the incoming elution buffer with traces of the previoussolution remaining in the recycle loop caused the actualammonium chloride concentration to be 0.68 and 0.78 Min elution cycles 1 and 2, respectively. To compensate forthis drop in eluting power, a third HGMF run wasconducted in which the ammonium chloride concentra-tion of the elution buffer used to fill the recycle loop wasraised to 1.2 M. A significant improvement in purificationwas achieved relative to the two previous runs. More than85% of the initial SOD was recovered in two elution cycleswith an overall purification factor of nearly 21 and yieldfactor of 17.8. Interestingly lab-scale processing of a 1-mLaliquot (i.e., using a magnetic rack) taken from the batchadsorption reaction corresponding to HGMF run 3 re-sulted in poorer overall purification performance (i.e., ayield factor of 14.9 vs 17.8), a lower yield (64.7% vs85.7%), but higher purification factor (23.1 vs 20.7). Apossible explanation for the higher elution efficiencyobserved in HGMF compared to the magnetic racksystem is that HGMF affords more effective compactionand dewatering of the adsorbent particle cake within thefilter than can be achieved under the influence of a lowstrength magnet placed alongside an eppendorf tube.Clearly, further experimentation is required to confirmthis, but for the present, lab-scale experiments wouldappear to give reasonably good predictions of the behaviorof the HGMF system.

3.5. Strategy for HGMF Process Development.The strategy used here for the design of a HGMF processfor SOD capture from whey is a generic approach thatcan be applied to any new type of macromolecule target

and/or feedstock. Figure 7 shows a schematic representa-tion of the approach, which consists of four main stepsprior to arriving at the final process.

In the first step, candidate ligands suitable for bindingthe target must be identified. In the study described herewe have borrowed from chromatography process designin using HiTrap columns. However, rapid small-scalebatch binding studies with commercial chromatographymedia or with commercial or custom-designed magneticsupports would be suitable. A kit of magnetic particlesfunctionalized in different ways would simplify ligandselection, and adaptation to rapid screening methodolo-gies, for example, in 96-well plate format, can be envis-aged.

During step 2 a suitable magnetic support must besourced and characterized. At present custom adsorbentconstruction may be required, as was used in this work,because of the limited selection of cheap, readily availablecommercial types. When this situation changes, steps 1and 2 may be combined. Support binding capacitiesshould preferably be >100 mg g-1 (1) and ideally >200mg g-1 (3) as capacity directly affects the amount ofsupports that must be used. Dissociation constants in thesubmicromolar range (39) are required in view of thebatch adsorption approach employed, and examinationof the binding kinetics is necessary for optimization ofthe batch (or possibly continuous) binding step; typically∼30 s (equivalent to three half-lives) is required forcomplete binding (4). However, optimization of the ad-sorbent concentration in addition to solution properties(pH, ionic strength, salt type, eluent type) during bindingand elution are also worthy of consideration as wasobserved for the SOD-whey system examined in thiswork.

Step 3 requires that the magnetic filter be character-ized with the feedstock and adsorbents to be used in thefinal process, given that filter capacity constrains theamount of supports that can be processed in eachcomplete HGMF cycle and that significant differences inthe behavior of model materials and finished adsorbentsin feedstock and buffers might occur, as was observed inthe present study. Magnetic support capture can beinfluenced by fluid viscosity (but not density), support

Table 4. Summary of Data for the Recovery of SOD from Unconditioned Whey in HGMF Run 1

recovery stepvolume

(mL)SOD(U)

protein(mg)

specific activity(U mg-1)

yield(%)

purificationfactor

yieldfactora

whey 52 41.6 531 0.078 100 1.0 1.0flow through 52 0 450.8wash 1 17 0 29.9wash 2 17 0.26 8.0 0.033 0.6 0.38 0.002elution 1 17 17.8 20.7 0.86 42.8 11.0 4.7elution 2 17 14.0 8.3 1.69 33.7 21.7 7.3combined elutions 34 31.8 29 1.10 76.5 14.1 10.8mass balance (%) 77.1 80.0

a Yield factor ) fractional yield × purification factor (ref 23). b The particle concentration during binding was 7 g L-1 but was raisedto 21.4 g L-1 during washing (10 mM potassium phosphate, pH 6.4, + 30 mM NaCl) and elution (100 mM potassium phosphate, pH 6.4,+ 0.8 M NH4Cl).

Table 5. Comparison of HGMF and Small-Scale Purifications of SOD from Unconditioned Whey Performed withCu2+-Charged HS3 Particles

NH4Cl concentration (M)

runparticle concn

during elution (g L-1)elution

buffer usedduring

elution 1during

elution 2yield(%)

purificationfactor

yieldfactora

HGMF 1 21.4 0.8 0.68 0.78 76.5 14.1 10.8HGMF 2 23.7 0.8 0.68 0.78 73.9 13.9 10.3HGMF 3 21.4 1.2 1.0 1.15 85.7 20.7 17.8magnetic rackb 21.4 1.2 nd nd 64.7 23.1 14.9a Yield factor ) fractional yield × purification factor (ref 23). b Small-scale experiment carried out in parallel with HGMF run 3.


size (variation arising between different batches as wellas possibly from agglomeration) affecting magnetic veloc-ity, feedstock particulates (increasing drag on the filtercake), processing flow rate, magnetic field strength, andtype of filter material (40).

Integration of steps 2 and 3 occurs at stage 4 wherethe process variables already defined are tested, vali-dated against performance in a HGMF process, andwhere necessary further optimized. At this point, modelsof the process could conceivably be developed to predictperformance robustness, and the advantages of continu-ous or semicontinuous processing using magnets withmultiple yokes and/or filters could be weighed, prior toarriving at an optimal final process in step 5.

4. ConclusionsMagnetic particle-based separation technology applied

for bioseparations can be scaled up from laboratorystudies in magnetic racks to an integrated magneticseparation-based process, known as high-gradient mag-netic fishing (HGMF). The ability to transfer the condi-tions found in small-scale studies to a HGMF systemallows a process for product capture from crude feed-stocks to be developed by following a straightforwardstrategy centered around individual optimization of twokey parts: (i) protein adsorption-desorption and (ii)adsorbent capture. Lab-scale optimization of the first keyevent provides a rapid means of screening a variety ofligand systems, buffers, and feedstock formulations. Inthe second part, estimation of the filter capacity forsupport capture requires that the adsorbents to be usedin the final process be characterized using a high-gradient magnetic separator. The switch on-off perma-nent magnet separator used here has proven to besuitable for biotechnological capture processes, and inview of the low operating costs afforded by the use ofpermanent magnets, further examination of the potentialof HGMF at larger scales is warranted. Using thestrategy outlined, SOD could be purified from crude wheyover 20-fold and in high yield (>85%). The fast processingrates and lack of need for feedstream pretreatment makeHGMF an interesting alternative/addition to other sepa-rations such as ultrafiltration and ion exchange chroma-tography used currently for the isolation of milk andwhey proteins.

NotationC* protein concentration in the liquid phase (M)Kd apparent dissociation constant (M)KdA apparent dissociation constant at “tight” sites

(M)

KdB apparent dissociation constant at “weak” sites(M)

Q* protein adsorption capacity at equilibrium (molg-1)

Qmax maximum capacity for adsorbed protein (mol g-1)QmaxA maximum capacity for adsorbed protein at “tight”

sites (mol g-1)QmaxB maximum capacity for adsorbed protein at “weak”

sites (mol g-1)

Acknowledgment

A.M. gratefully acknowledges the receipt of a Ph.D.stipend from the Forschungzentrum Karlsruhe GmbH,and C.S.G.G. thanks the Portuguese Foundation forScience and Technology for financial support (GrantSFRH/BD/1218/2000).

References and Notes(1) Hubbuch, J. J. Development of adsorptive separation sys-

tems for recovery of proteins from crude bioprocess liquors.Ph.D. Thesis, Technical University of Denmark, 2001.

(2) Hubbuch, J. J.; Matthiesen, D. B.; Hobley, T. J.; Thomas,O. R. T. High gradient magnetic separation versus expandedbed adsorption: a first principle comparison. Bioseparation2001, 10, 99-112.

(3) Heebøll-Nielsen, A. High gradient magnetic fishing: Supportfunctionalisation and application for protein recovery fromunclarified bioprocess liquors. Ph.D. Thesis, Technical Uni-versity of Denmark, 2002.

(4) Hoffmann, C. Einsatz magnetischer Separationsverfahrenzur biotechnologischen Produktaufbereitung. Ph.D. Thesis,Forschungzentrum Karlsruhe, Karlsruhe, Germany, 2002.

(5) Hubbuch, J. J.; Thomas, O. R. T. High-gradient magneticaffinity separation of trypsin from crude porcine pancreatin.Biotechnol. Bioeng. 2002, 79, 301-313.

(6) Heebøll-Nielsen, A.; Choe, W.-S.; Middelberg, A.; Thomas,O. R. T. Efficient inclusion body processing using chemicalextraction and high gradient magnetic fishing. Biotechnol.Prog. 2003, 19, 887-898.

(7) Meyer, A. Einsatz magnettechnologisher Trennverfahrenzur Aufbereitung von Molkereiproduckten; Ph.D. Thesis,Forschungzentrum Karlsruhe. Karlsruhe, Germany, 2004.

(8) Gerberding, S. J.; Byers, C. H. Preparative ion-exchangechromatography of proteins from dairy whey. J. Chromatogr.A 1998, 808, 141-151.

(9) Hahn, R.; Schulz, P. M.; Schaupp, C.; Jungbauer, A. Bovinewhey fractionation based on cation-exchange chromatogra-phy. J. Chromatogr. A 1998, 795, 277-287.

(10) Hill, R. D. Superoxide dismutase activity in bovine milk.Aust. J. Dairy Technol. 1975, 30, 26-28.

(11) Asada, K. Occurrence of superoxide dismutase in bovinemilk. Agric. Biol. Chem. 1976, 40, 1659-1660.

(12) Korycka-Dahl, M.; Richardson, T.; Hicks, C. L. Superoxidedismutase activity in bovine milk serum. J. Food Prot. 1979,42 (11), 867-871.

(13) Kiyosawa, I.; Matuyama, J.; Nyui, S.; Yoshida, K. Cu-, Zn-and Mn-superoxide dismutase concentration in human co-lostrum and mature milk. Biosci. Biotechnol. Biochem. 1993,57 (4), 676-677.

(14) Greenwald, R. A. Superoxide dismutase and catalase astherapeutic agents for human diseases-A critical review. FreeRadical Biol. Med. 1990, 8, 201-209.

(15) Holbrook, J.; Hicks, C. L. Variation of superoxide dismutaseactivity in bovine milk. J. Dairy Sci. 1978, 61, 1072-1077.

(16) Eremin, A. N.; Metelitsa, D. I. Purification of superoxidedismutase from red blood cells using reversed micelles ofaerosol-OT in heptane. Appl. Biochem. Microbiol. 1996, 32,260-264.

(17) Michalski, W. P.; Prowse, S. J. Cu, Zn superoxide dismu-tase from chicken erythrocytes. Comp. Biochem. Physiol.1991, 100 (2), 371-375.

Figure 7. A general strategy for HGMF process design.


(18) Langmuir, I. The adsorption of gases on plane surfaces ofglass, mica and platinum. J. Am. Chem. Soc. 1918, 40, 1361-1403.

(19) Hoffmann, C.; Franzreb, M.; Holl, W. H. A novel highgradient magnetic separator (HGMS) design for biotechapplications. IEEE Trans. Appl. Supercond. 2002, 12, 963-966.

(20) Nebot, C.; Moutet, M.; Huet, P.; Xu, J.-Z.; Yadan, J.-C.;Chaudiere, J. Spectrophotometric assay of superoxide dis-mutase activity based on the activated autoxidation of atetratcyclic catechol. Anal. Biochem. 1993, 214, 442-451.

(21) Creamer, L. K.; MacGibbon, A. K. H. Some recent advancesin basic chemistry of milk proteins and lipids. Internat. DairyJ. 1996, 6, 639-568.

(22) Vasavada, P. C.; Cousin, M. A. Dairy microbiology andsafety. In Dairy Science and Technology Handbook, 2 ProductManufacturing; Hui, Y. H., Ed.; VCH Publishers Ltd.: Cam-bridge, UK, 1993.

(23) Hearle, D. C.; Aguilera-Soriano, G.; Wiksell, E.; Titchener-Hooker, N. J. Quantifying the fouling effects of a biologicalprocess stream on chromatographic supports. Inst. Chem.Eng. Res. Event 1994, 1, 174-176.

(24) Karlsson, K.; Marklund, S. L. Extracellular superoxidedismutase in the vascular system of mammals. Biochem. J.1988, 255, 223-228.

(25) Tibell, L. A. E.; Sethson, J.; Buevich, A. Characterizationof the heparin-binding domain of human extracellular su-peroxide dismutase. Biochim. Biophys. Acta 1997, 1340, 21-32.

(26) Miyato-Asano, M.; Ito, K.; Ikeda, H.; Sekiguchi, S. Purifica-tion of copper-zinc-superoxide dismutase and catalase fromhuman erythrocytes by copper-chelate affinity chromatogra-phy. J. Chromatogr. A 1986, 370, 501-507.

(27) Andersson, L.; Sulkowski, E.; Porath, J. Immobilised metalion affinity chromatography of serum albumins. Bioseparation1991, 2, 15-22.

(28) Scatchard, G. The attraction of proteins for small moleculesand ions. Ann. N.Y. Acad. Sci. 1949, 51, 660-672.

(29) Kinghorn, N. M.; Paterson, G. R.; Otter, D. E. Quantifica-tion of the major bovine whey proteins using capillary zoneelectrophoresis. J. Chromatogr. A 1996, 723, 371-379.

(30) Al-Mashiki, S. A.; Li-Chan, E.; Nakai, S. Separation ofimmunoglobulins and lactoferrin from cheese whey by chelat-ing chromatography. J. Dairy Sci. 1988, 71, 1747-1755.

(31) Fukumoto, L. R.; Li-Chan, E.; Kwan, L.; Nakai, S. Isolationof immunoglobulins from cheese whey using ultrafiltrationand immobilized metal affinity chromatography. Food Res.Int. 1994, 27, 335-348.

(32) Lonnerdal, B.; Carlsson, J.; Porath, J. Isolation of lacto-ferrin from human milk by metal-chelate affinity chroma-tography. FEBS Lett. 1977, 75, 89-92.

(33) Torres, A. R.; Peterson, E. A.; Evans, W. H.; Mage, M. G.;Wilson, S. Fractionation of granule proteins of granulocytesby copper chelate chromatography. Biochim. Biophys. Acta1979, 385-392.

(34) Sciancalepore, V.; Pizzuto, P.; Stefano, G. de. Purificationof lactoperoxidase by immobilized copper affinity chromatog-raphy. Latte 1994, 19 (11), 1152-1155.

(35) El-Rassi, Z.; Horvath, C. Metal chelate interaction chro-matography of proteins with iminodiacetic acid-bonded sta-tionary phases on silica support. J. Chromatogr. 1986, 359,241-253.

(36) Blomkalns, A. L.; Gomez, M. R. Purification of bovinealpha-lactalbumin by immobilized metal ion affinity chro-matography. Prep. Biochem. Biotechnol. 1997, 27 (4), 219-226.

(37) Misra, H. P.; Fridovich, I. Superoxide dismutase andperoxidase: a positive activity stain applicable to polyacry-lamide electropherograms. Arch. Biochem. Biophys. 1977, 183(2), 511-515.

(38) Fridovich, I. Superoxide dismutases. Adv. Enzymol. 1986,58, 61-97.

(39) Chase, H. A. Affinity separations utilising immobilizedmonoclonal antibodies-A new tool for the biochemical engi-neer. Chem. Eng. Sci. 1984, 39, 1099-1125.

(40) Watson, J. H. P. Magnetic filtration. J. Appl. Phys. 1973,44, 4209-4213.

Accepted for publication September 23, 2004.

BP049656C


demonstration of a strategy for product purification by high-gradient magnetic fishing: recovery of...

Documents