Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we...

9
F F luidized-bed (FB) processes have been extensively described for the separation of proteins from crude feedstocks (1). Application of this technology to initial protein capture and separation stages offers significant advantages over packed- bed processes. Fluidized sorbent beds provide a practical option to process very crude biological materials containing particles in suspension such as aggregates, cells, and cell debris. In practice, a fluidized bed is created when beaded sorbents are lifted inside a column by an upward liquid flow of buffers and sample solutions. The process of fluidization creates interstitial spaces between the beads through which particles suspended in the feedstock can easily travel. When the physicochemical environment is properly adjusted, proteins of interest are captured by the beaded supports while undesired soluble and insoluble materials are eliminated. The attractiveness of this type of downstream biochemical process is its ability to combine clarification, selective capture, and concentration into a single unit operation (2). So FB capture processes can often be coupled directly to fermentors for continuous extraction of biological materials produced by the cells in culture (3). FB is particularly valuable for processing large volumes of unclarified feedstocks that contain recombinant proteins expressed at relatively low levels (4). The capture of extracellular proteins such as human serum albumin (5) and alcohol dehydrogenase (6) from yeasts has been reported — as well as monoclonal antibodies (MAbs) and growth factors from mammalian cells (7, 8). Intracellular proteins such as those expressed in Escherichia coli (9–12) have also been separated after cell disruption. Use of FB has been extended to the separation of proteins from nonclarified biologicals in the absence of cells, as was the case in separating antithrombin III from milk after removal of fats by centrifugation and caseins by selective precipitation (13). The Solid Phase: At the heart of FB separation processes is the solid- phase sorbent, which is physically characterized by its mean particle size and distribution as well as its density. Increasing particle size raises the terminal velocity of the bead (Equation 1), allowing fluidization at a relatively higher linear upward flow velocity. However, an increase in particle size also raises the mass transfer resistance, which is detrimental to the sorbent binding capacity. Significantly higher particle densities can raise the terminal velocity of the beads. When these particle physical parameters are fully evaluated, an optimal combination converges PRODUCT FOCUS: ALL BIOLOGICS PROCESS FOCUS: DOWNSTREAM PROCESSING (EARLY PRODUCT CAPTURE) WHO SHOULD READ: PROCESS DEVELOPMENT AND MANUFACTURING KEYWORDS: FLUIDIZED-BED CHROMATOGRAPHY , PROTEIN CAPTURE, ZIRCONIA, SOLID PHASE LEVEL: INTERMEDIATE B IO P ROCESS TECHNICAL PAUL SENYSZYN (WWW.ISTOCKPHOTO.COM) Effective Protein Capture in Fluidized-Bed Mode with Zirconia-Based Beads Patrick Santambien, Nicolas Voute, Anthony Schapman, Vincent Ravault, and Egisto Boschetti 46 BioProcess International OCTOBER 2003

Transcript of Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we...

Page 1: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

FF luidized-bed (FB) processeshave been extensivelydescribed for the separationof proteins from crudefeedstocks (1). Application

of this technology to initial proteincapture and separation stages offerssignificant advantages over packed-bed processes. Fluidized sorbentbeds provide a practical option toprocess very crude biologicalmaterials containing particles insuspension such as aggregates, cells,and cell debris. In practice, afluidized bed is created whenbeaded sorbents are lifted inside acolumn by an upward liquid flow ofbuffers and sample solutions. Theprocess of fluidization createsinterstitial spaces between the beadsthrough which particles suspendedin the feedstock can easily travel.When the physicochemical

environment is properly adjusted,proteins of interest are captured bythe beaded supports while undesiredsoluble and insoluble materials areeliminated. The attractiveness of thistype of downstream biochemicalprocess is its ability to combineclarification, selective capture, andconcentration into a single unitoperation (2). So FB captureprocesses can often be coupleddirectly to fermentors forcontinuous extraction of biologicalmaterials produced by the cells inculture (3).

FB is particularly valuable forprocessing large volumes ofunclarified feedstocks that containrecombinant proteins expressed atrelatively low levels (4). The captureof extracellular proteins such ashuman serum albumin (5) andalcohol dehydrogenase (6) fromyeasts has been reported — as wellas monoclonal antibodies (MAbs)and growth factors from mammaliancells (7, 8). Intracellular proteinssuch as those expressed inEscherichia coli (9–12) have also beenseparated after cell disruption. Useof FB has been extended to theseparation of proteins fromnonclarified biologicals in theabsence of cells, as was the case inseparating antithrombin III frommilk after removal of fats bycentrifugation and caseins byselective precipitation (13).

The Solid Phase: At the heart ofFB separation processes is the solid-phase sorbent, which is physicallycharacterized by its mean particlesize and distribution as well as itsdensity. Increasing particle sizeraises the terminal velocity of thebead (Equation 1), allowingfluidization at a relatively higherlinear upward flow velocity.However, an increase in particle sizealso raises the mass transferresistance, which is detrimental tothe sorbent binding capacity.Significantly higher particle densitiescan raise the terminal velocity of thebeads. When these particle physicalparameters are fully evaluated, anoptimal combination converges

PRODUCT FOCUS: ALL BIOLOGICS

PROCESS FOCUS: DOWNSTREAM

PROCESSING (EARLY PRODUCT CAPTURE)

WHO SHOULD READ: PROCESS

DEVELOPMENT AND MANUFACTURING

KEYWORDS: FLUIDIZED-BED

CHROMATOGRAPHY, PROTEIN CAPTURE,ZIRCONIA, SOLID PHASE

LEVEL: INTERMEDIATE

B I O P R O C E S S TECHNICAL

PAUL SENYSZYN (WWW.ISTOCKPHOTO.COM)

Effective Protein Capture in Fluidized-Bed Mode with Zirconia-Based BeadsPatrick Santambien, Nicolas Voute, Anthony Schapman,Vincent Ravault, and Egisto Boschetti

46 BioProcess International OCTOBER 2003

Page 2: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

toward increasing the particledensity while reducing particlediameter. Small-diameter, densebeads are good because of theirlower intraparticle diffusionresistance, reducing residence timewithout decreasing binding capacityor compromising processing speeds.

In recent years, several solid-phase sorbents have been proposedand characterized specifically for FBapplications (1, 14–17). Among thefirst were hydrogels densified with asolid material such as quartzdispersed through the polymernetwork (1). The density of suchmaterials was not very high, so beaddiameter had to be relatively large(200 µm average) to producesorbents with adequate fluidizationproperties. To improve on thehydrodynamic performance of solidphase sorbents at higher velocities,beads were loaded with densermineral materials (18, 19) orconstructed using heavy cores ofstainless steel covered by a layer ofagarose (20). The use of suchdensity-enhancing materials allowedbeads of smaller diameter to beproduced, which had the addedbenefit of lowering mass transferresistance. More recently, porousmineral oxide particles of very highdensity have been reported, theirpore volumes filled byfunctionalized hydrogels (17).

Fluidization of solid-phaseadsorbents requires a specific devicecalled a contactor. This is a veryimportant element in the separationprocess because it dictates criticalhydrodynamic process parameters.Even so, only a limited number ofreports describing this aspect of FBtechnology are available (7, 21). AnFB contactor basically consists of achromatographic column in whichliquid flow travels from the bottomto the top. Typically, a constantupward flow is applied during sampleloading and column washing. For theprotein desorption phase, that flowcan be reversed as with an ordinarypacked-bed chromatography column(22). Operationally, alternating theflow direction between the loading/washing and elution phases of aprocess has proven to be relatively

impractical. Over multiple processcycles, difficulties can arise withrespect to sorbent cleaning and beadaggregation as well as bed expansion,resulting in unacceptable increases inequilibration and total cycle time.This operation also requires thepresence of a bottom filter to keepbeads from exiting the column. Thatfilter must have holes small enoughto prevent bead loss but largeenough to allow cells and cell debristo pass through without hindrance.In practice, deposits of such solidmaterial often form on the inlet sideof the bottom frit (23), necessitatingspecific cleaning treatments.

An alternative contactor designhas been described in which a lateralinlet is located close to the bottomof the column to introduce samplematerial (24). A slow agitation ofthe lower part of the column createsa mixing zone above which a stablefluidized bed can be formed. Theadvantage of this design is that nofrits are used at the inlet port,eliminating the problem of cloggingcaused by cells and cell debris.Naturally, with this contactortechnology the entire separationprocess, including elution, takesplace in expanded-bed mode.

Herein we discuss a compositematerial composed of porouszirconium oxide filled with asynthetic hydrogel. FB contactorsused were a 1.0-cm diametercontactor from UpFrontChromatography of Copenhagen,Denmark (www.upfront-dk.com)and a 2.5-cm diameter contactorfrom Amersham Bioscienceof Uppsala, Sweden(www.amershambiosciences.com).

FLUIDIZED-BED PARTICLE DESIGN

When beads that have settled at thebottom of a column are submittedto an upward liquid flow, theybecome suspended in proportion tothe linear liquid velocity of thatflow. The settling velocity of solidparticles determines the terminalfluidization velocity. During steady-state fluidization, particle weightand flow resistance are inequilibrium.

Theoretical values for theterminal velocity (ut) of the beadscan be calculated using the Stokesequation, where �p is the particledensity, �l is the liquid density, dp isthe particle diameter, and � is theliquid viscosity (Equation 1). Fromthis equation, the bed expansion iscalculated with the followingRichardson-Zaki equation (26),where � is the bed voidage and n anempirical exponent (Equation 2).That exponent has been in therange of 3–5 for agarose-based

OCTOBER 2003 BioProcess International 47

FB PROCESS DESIGN

A number of important criteriamust be considered in the designof FB processes. The mostimportant are

• physicochemical properties ofthe solid phase adsorbent

• speed of upward liquid flow

• nature of the feedstock fromwhich the biologic of interest is tobe captured

• column or contactor in whichthe sorbent is housed and theoperation performed.

THE LIQUID PHASE

In addition to the sorbent-relatedparameters, other factors playimportant roles in thehydrodynamics of a fluidized bed.These include the density andviscosity of the liquid phase, thelatter depending highly ontemperature.

EEqquuaattiioonn 11

EEqquuaattiioonn 22

Page 3: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

beads (27). Size, size dispersion,geometrical shape, and density ofthe solid phase sorbent all influencethe dynamic fluidized bedperformance (26). Although particlediameter plays a more importantrole than that of particle density,both can be used to controlfluidization parameters.

Large sorbent particles canaccommodate very high upwardfluid velocities for rapidly treatinglarge volumes of feedstock over alimited time. Because diffusion ofmacromolecules in a predeterminedporous network is time dependent,however, large particles can requirelong residence times to achieve theirmaximal binding capacity. At aconstant flow rate (or constantresidence time), macromoleculeuptake has been shown to dependon particle size (28). With a certainlevel of hindrance, diffusion of largeproteins is significantly moresensitive to linear velocity than thatof small proteins. Therefore, largesorbent particles and high flow ratesare not always compatible with highcapture efficiency.

Diffusion may also be affected byinadequate sorbent pore size. Withrestricted diffusion conditionsand/or limited residence times,proteins are adsorbed mostly to theexternal part of the adsorbentbeads, which limits load quantities.

The effects on adsorption of bedexpansion, axial dispersion, viscosity,and backmixing were investigatedby Chang et al. (29). In solid–liquidFB situations, film andintraparticulate mass transfer effectsare difficult to evaluate separatelyfrom hydrodynamic effects. Wrightet al. (30) contributed to ourunderstanding of those phenomenainvolving adsorbents of differentcharacteristics. It appeared that evenwhen axial dispersion was highbecause of increased speed andexpansion factors, mass transferlimitations were still the dominantfactor in reducing proteinbreakthrough. However, masstransfer is significantly affected byspecific adsorbent properties.Intraparticle diffusion is significantly

improved by decreasing particlediameters (31). Equally, adsorbentswith large pores provide goodefficiencies even at high flow rates,lower retention of biomass, andeasier elimination of feedstockelements that can adsorb to thesurface of the solid phase (32).

During sorbent fluidization, ithas been shown that a particleclassification process takes placebased on the size dispersion ofsorbent beads (33, 34). Their degreeof protein saturation varies from thebottom to the top of the contactoras a function of bead particle size.In fact, comparing the adsorptionperformance of the same beadedadsorbent at different diameters inFB mode showed a higher bindingcapacity for the smallest particlesbecause of their shorter paths withinthe matrix network (35).

Kinetic and thermodynamicphenomena related to proteinadsorption in FB mode are verysimilar to what happens withpacked-bed chromatography.Differences are in axial mixing andparticle density, dictating thevelocity at which the adsorption canbe operated without elutriationphenomena.

Sorbent density can be increasedby incorporating heavy solidparticles such as quartz (1), glass(23), or metal alloys and oxides (18)within the hydrogel network.Because of differences in intrinsicdensity from bead to bead and theconsequent effects on particledispersion, this approach may notyield extremely homogenousmaterial.

Through past observations andmodeling, it is known that for

48 BioProcess International OCTOBER 2003

FFiigguurree 11:: (A) Dynamic binding capacity (DBC) at 10% breakthrough (BT) of lactoperoxidaseon CM HyperZ sorbent as a function of pH and NaCl concentration; (B) binding capacityvariation as a function of pH at physiological ionic strength; 1.1-cm ID � 5 cm lengthcolumn with 5 mL sorbent at a linear flow rate of 300 cm/hr

FFiigguurree 22:: Dynamic binding capacity (DBC) at 10% breakthrough (BT) of human polyclonalIgGs on CM HyperZ in either packed-bed (A) or in fluidized-bed mode (B), as a function ofpH and NaCl concentration; 1.1-cm ID � 5 cm length packed-bed column with 5 mLsorbent at a linear flow rate of 300 cm/hr; 1.0-cm ID � 5 cm length fluidized-bed columnwith 7.8 mL sorbent at a linear flow rate of 350 cm/hr

Page 4: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

50 BioProcess International OCTOBER 2003

particles ranging about 30–100 µmin diameter, densities of around2.5–3.5 will yield expansion factorsof about two, with upward velocitiesof a few hundred cm per hour.Described separations herein wereobtained with zirconium-oxide–based beads designed for proteinadsorption (19). A generalcharacterization of the resultingmaterial was previously described(17, 36, 37).

USING ZIRCONIA-BASED BEADS

The zirconia-based adsorbentdiscussed here consists of porouszirconia beads in which the porevolume has been loaded with ahydrogel carrying carboxyl groups.This material was supplied byCiphergen Biosystems of Fremont,CA (www.ciphergen.com), underthe HyperZ trade name.Characterization of this porousbeaded material was recentlypublished (17). Most of the beads(80%) have diameters ranging from40 to 100 µm, with an averageparticle size of 70 µm. Intraparticleporosity is about 53% of total beadvolume. Particle density is about 3.2 g/mL, and ionic charge densityis about 0.15 mequiv/mL. Bycomparison, a quartz-agarosecomposite material supplied underthe Streamline trade name byAmersham Bioscience of Uppsala,Sweden (www.amershambiosciences.com) has a particle size distributionof 100–300 µm and an averagedensity of 1.2 g/mL.

Ionic Strength: Figure 1 shows thetypical protein adsorptionperformance, for which thelactoperoxidase binding capacity at arelatively high flow rate wasmeasured under dynamicconditions. The optimal pH forprotein binding on CM HyperZsorbent depends on sodium chloride(NaCl) concentration at least overthe range studied, from pH 6.0 at 4 mS/cm to pH 4.5 at 23 mS/cm.The ionic strength zone weinvestigated covered thephysiological value of around 15 mS/cm. For milklactoperoxidase, the bindingcapacity at pH 4.5 in the presenceof 0.15 M NaCl was about 70–80 mg/mL of resin. Under thesame conditions, dynamic bindingcapacity for quartz-loaded agarosebeads was about 20 mg/mL of resin(data not shown). At pH >5, anincrease of NaCl concentrationprogressively reduces the bindingcapacity as expected for classicalionic interactions. Protein bindingat nearly physiological ionicstrengths happens because of thehigh charge density of the hydrogelwithin the zirconia pores.

The capability for capturingproteins at physiological ionicstrengths is of great interest forpreparative applications. Iteliminates the need to dilutefeedstocks to ionic strength levelscompatible with traditional cationexchangers. Dilution of thefeedstock generates two majorissues: raising the volume to processand lowering the proteinconcentration, thus reducingcapture efficiency. So a peculiarproperty of the zirconia sorbentresolves a major problem of proteincapturing often associated with highdilution of target proteins fromexpression media.

The capture of antibodies atphysiological ionic strength was alsostudied, with results suggesting thatantibodies produced in vitro can becaptured directly from cell culturesupernatants without priorconductivity treatment (38).Polyclonal antibody experimentsusing the zirconia-based sorbent in

FB mode demonstrated a similarbehavior. As shown in Figure 2 atpH 4.5, the highest bindingcapacity was obtained when theconcentration of NaCl was 0.15 M.With the pH increased to 5.5, themaximum binding capacity movedtoward a lower ionic strength.Optimal productivity thus may beachieved by adjusting the feedstockpH to 4.5 and loading directly(without dilution) rather thanreducing the pH to only 5.5 anddiluting the feedstock several-foldbefore loading. Additionally, thesedata suggest that high bindingcapacities can be achieved using thismaterial, probably requiring athorough evaluation of the effects ofpH and conductivity.

Hydrodynamics: Unlike withpacked-bed chromatography,increasing the upward flow velocityin FB mode not only reducesresidence time but also increasesbed expansion. Residence time isthe dynamic contact time forproteins and sorbents, whichchanges linearly with flow rate.Figure 3 shows experimentsperformed using a 10-mm diameterFB contactor. As expected, bindingcapacity diminishes when flow rateincreases — primarily because of thereduction in residence time. Withthe zirconia-based sorbent, bindingcapacity decreased 50% (from 60 to30 mg/mL) when residence timewas reduced from 7.5 to 5 minutes.

Under similar hydrodynamicconditions, the quartz-loadedagarose cation exchanger behaved

FFiigguurree 33:: Dynamic binding capacity (DBC)at 10% breakthrough (BT) of humanpolyclonal IgG (hIgG at 5 mg/mL) on CMHyperZ and SP Streamline sorbents; 1.0-cmID � 10 cm length column with 7.8 mLsorbent and 50 mM acetate containing 150mM NaCl at pH 4.7 and a linear flow rateof 50–550 cm/hr.

FFiigguurree 44:: Bed expansion properties ofzirconia-based cation exchanger as afunction of cell density; 10-cm ID columnwith 7.8 mL sorbent corresponding to a 10-cm settled bed, with hybridoma celldensities as shown

Page 5: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

52 BioProcess International OCTOBER 2003

similarly; however, it demonstrated alower binding capacity because ofthe significantly larger diameter ofthe solid-phase adsorbent (100–300 µm compared with40–100 µm). The binding capacityof about 18 mg/mL at 7.5 minutesresidence time decreased between 2 and 4 mg/mL at 5 minutes.

PROTEIN CAPTURE FROM

UNCLARIFIED FEEDSTOCKS

A typical FB application is theselective capture of recombinantproteins from crude fermentationbroths, where cells and cell debrisare present along with the targetprotein. Interaction of cells and celldebris with the surface of the solidphase material should be prevented— or at least limited to acceptablelevels. This is particularly critical forintracellular proteins, for which celllysis is required beforechromatographic processing.Preparation of cells before lysing(and the lysing procedure itself) cansignificantly decrease the presence oftroublesome contaminants in theresulting feedstocks for loading.

One approach is to use virtual

ultrasonic filters at the bioreactoroutlet to reduce the number of cellsintroduced into the FB contactor(39). This reduces cell adhesion tothe diffusion plate (or frit) at thebottom of the contactor. Usingcontactors without diffusion platesprevents cell deposits; however, suchtechnology does not allow elutionof the adsorbed biologicals inpacked-bed mode. Another problemto be resolved is the phenomenonof cells and cell debris sticking ontothe external surface of the beads.Introducing a viscosity-enhancingadditive (such as glucose) to thewashing buffers has been describedas a possible solution (40).

Another issue to deal with is themodification of bed expansion thathappens upon sequential use ofdifferent liquids (e.g., equilibrationbuffer, crude feedstock, washingbuffers, and elution and regenerationsolutions). For example, injectingunfiltered feedstocks typically leadsto an increase of the bed expansion,primarily because cells are presentbut also to a lesser extent because ofthe intrinsic viscosity of the solution.

ANTIBODY CAPTURE FROM

HYBRIDOMA CELL CULTURE

Antibody separation fromhybridoma or CHO culture (and thesupernatants of other mammaliancell lines) is one of the mostimportant applications of FBprocesses (8). Mammalian cellsgenerally have anionic characteristicsthat make them amenable for usewith cation exchangers (7, 40, 41),involving minimal cell interaction.Because antibodies can be adsorbedonto anionic solid phases, conditionsare ideal for the FB process.

A well-known limitation ofclassical cation exchangers is theirneed for diluted feedstock to reachthe required ionic strengths forbinding. This limitation iscircumvented by the zirconiasorbent, significantly reducing theloading time and enhancing captureefficiency. For proper ionization ofboth proteins and sorbents, thefeedstock pH was lowered to4.5–4.7 and directly loaded. In theinvestigated cell culture broth, celldensity was 12.5 � 106 cells/mLand antibody concentration was 56 µg/mL.

The presence of cells influencedthe expansion factor of the resin at agiven linear flow rate. As Figure 4illustrates, a bed expansion factor oftwo was obtained at about100 cm/hr in the presence of cellswithout dilution (12.5 � 106

cells/mL). However, when the cellconcentration was reduced fourtimes, the flow rate required toachieve that bed expansion factorof two increased to about200–230 cm/hr. So a trade-off wasidentified between volume andspeed. Antibody loading wasperformed without dilution at anupward flow of 75–100 cm/hr,whereas washing and elution stepswere performed at 300–350 cm/hrto maintain a constant bedexpansion factor. Antibody elutionwas achieved by increasing the NaClto 1M while maintaining the pHconstant.

Analysis of collected fractionsshowed that the eluate containedmost of the antibodies at a purity ofabout 70% (Figure 5). Flowthrough

FFiigguurree 55:: SDS-PAGE analysis in reducing conditions of collected fractions from monoclonalantibody separation using CM HyperZ in fluidized-bed mode with unclarified feedstock (HC = heavy chains, and LC = light chains). Lanes 1 and 4 are molecular markers; lane 2 isthe crude sample; lane 3 is the elution.

Page 6: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

54 BioProcess International OCTOBER 2003

proteins were not visible onelectrophoresis because they werevery dilute. When compared to datafrom a classical packed-bed column(trials performed in a preliminarystudy), the expanded-bed processshowed a similar recovery, althoughthe overall purity of the collectedantibody was slightly lower: 10–15%less (Table 1). The FB processshowed no aggregation of particlesor channeling during operation. Avery low amount of cell adhesionwas observed, although it was notpossible to quantify.

SEPARATION OF PROTEINS

FROM YEAST SUSPENSION

In a preliminary study, theexpansion of CM HyperZ sorbentwas measured using suspensionscontaining yeast cells only. We usedSaccharomyces cerevisiae in theseexperiments. As expected, at aconstant upward flow rate the bedexpansion factor increased as afunction of increasing biomass load(Figure 6). For example, at a linearvelocity of 291 cm/hr theexpansion factor was about two inthe absence of biomass, whereas itwas about 2.5 in the presence of10% (w/v) biomass. This was foundto be similar for quartz-loadedagarose beads (results not shown).

We performed an experimentusing S. cerevisiae suspension at 10%(w/v) biomass in a buffer ofphysiological ionic strength and pH 4.5 containing 1 mg/mLpolyclonal antibodies. The 10-mmdiameter FB contactor contained 10-mL of settled resin (column heightwas 14 cm). The equilibration bufferwas 50 mM acetate at pH 4.5,

containing 5 mM citrate and 0.15 MNaCl. The washing buffer was thesame but contained 0.4 M glucoseto increase viscosity and aid indetaching possibly adherent cells onthe sorbent. As described in theprevious section, elution wasperformed by increasing NaCl to1M while maintaining a constantpH. The linear velocity was keptconstant at 300 cm/hr. From threeseparate experiments, antibodyrecovery values averaged 83%. Puritymeasured by SDS-PAGE was higherthan 90%, and contamination byyeast cells was at about 0.23%compared with the number of cellsloaded.

Several examples of proteincapture from yeast fermentationhave been reported in the literature.Cell densities studied varied greatlyfrom strain to strain, but the highestreported was with Pichia pastorisfermentation (42, 43). The strain iswell adapted to large-scaleapplications because of very highcell density and expression levelsthat can be reached with a relativelysimple and inexpensive inductiontechnique. Large-scale separations ofangiogenesis inhibitors (44) andenzymes (45) have been reported,with decisive advantage of proteincapturing using FB in place oftraditionally packed chromatographybeds. Recently, the separation ofendostatin from recombinant P. pastoris in FB mode was reportedusing quartz-loaded agarose beads(46) or zirconia-based sorbents (42).

LYSOZYME FROM E. COLI LYSATE

E. coli lysates are well knownexamples of feedstock from whichexpressed proteins have beencaptured and separated. Theselysates contain large amounts of celldebris that can be very difficult toseparate out using classicalseparation methods such as filtrationand centrifugation. Even whenfiltered, E. coli lysates are frequentlyturbid and relatively viscous,generating problems when injectedinto packed columns. The use of FBadsorption offers a potentiallyeffective alternative to this difficultapplication. Several examples have

been reported in which fluidizedbeds were used in the recovery ofrecombinant proteins expressed inthe periplasmic compartment (11)or intracellularly (47, 48).

We separated lysozyme from E. coli lysate by small-scale FB andthen scaled up the process.Separation of lysozyme wasperformed at two different biomassconcentrations (2.5% and 11.5%ww/v) resulting in different bedexpansion factors at the same linearvelocity (Table 2). The equilibrationbuffer was 50 mM acetate at pH 4.5, containing 5 mM citrateand 50 mM NaCl. E. coli lysate(280 mL) was loaded into afluidized bed contactor of 10-mmdiameter containing 7.8 mL ofzirconia-based cation-exchangesorbent at 300 cm/hr. Thefeedstock contained 1 mg/mLlysozyme. The column was thenwashed with the same buffer untilnonadsorbed proteins and celldebris were eliminated, and thelysozyme was desorbed by raisingthe concentration of NaCl to 1 Min the same buffer. Whilemaintaining the expansion factor ata constant rate of about two forboth sorbents, we performedexperiments with 2.5% (ww/v)biomass requiring a linear upward

FFiigguurree 66:: Bed expansion properties of CMHyperZ as a function of yeast biomassconcentration at a constant flow rate. 2.5-cm ID column with 20-cm bed highand a linear flow rate of 291 cm/hr (initialbed expansion factor of 2.0)

FFiigguurree 77:: SDS-PAGE analysis innonreducing conditions of collectedfractions from lysozyme (Lys) purification influidized-bed mode. Lane 1 corresponds tothe initial crude feedstock; lane 2 is theflowthrough fraction; lanes 3 and 4 areelutions from samples containing 2.5% and11.5% biomass, respectively.

Page 7: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

56 BioProcess International OCTOBER 2003

flow of 260 cm/hr — comparedwith 120 cm/hr for 11.5% (ww/v)biomass. Lysozyme binding capacitywas about 19 mg/mL and 25 mg/mL, with 11% (ww/v)biomass and 2.5% (ww/v) biomass,respectively. By contrast, lysozymerecovery was better when performedat the lowest biomass concentration.

Electrophoresis results (Figure 7)demonstrate that the lysozymepurity increased from about 25% inthe initial cell lysate to about 75% inthe pooled elution fractions. The FBremained stable during operation,with no channeling or aggregationof particles. When the sameexperiments were performed usingquartz-loaded agarose cationexchanger, problems with sorbentelutriation arose in the presence ofbiomass content >2.5%. Also, the

net at the bottom of the contactorcolumn clogged rapidly at highbiomass levels as cell debris wasdeposited during the loading phase.

This study illustrates theimportance of experimentallydetermining the optimal conditionsfor FB operations. Biomassconcentration is a critical parameterand must be considered in each caseto optimize productivity andprevent sorbent fouling and columnclogging.

Scale-Up: Having developed apurification method at small scale,we directly scaled up the processusing a larger column of the sametype. Settled solid phase (1,700 mL)was loaded in a 100-mm diametercolumn. All other separationconditions remained the same as inthe small-scale process. Biomass

concentration was 2.5% (ww/v).Table 3 provides comparativeresults. Binding capacity results werevery similar to the small-scaleexperiment: close to 20 mg oflysozyme per mL of resin. Finalrecovery (88%) and purity (higherthan 80%) were also very similar.Total cycle time did not changebecause the same linear velocity wasused. A recent paper evaluating thescale-up of FB processes in theisolation of proteins from E. colilysates reports that it is often easierto achieve good results fromrelatively large-scale experimentsthan with small laboratory-scaleprocedures (48).

EFFECTIVENESS

An optimized sorbent for FBadsorption of biologicals cansignificantly contribute to theadoption of direct-captureapproaches from very crudefermentation broths. Althoughthermodynamic and kineticproperties of FBs are essentially thesame as for packed beds, effectiveadsorption can be operated underrelatively high upward flow rates,justifying the high density of solid-phase adsorbents. In that respect,the material described hereinrepresents a technologicalimprovement over traditionalmaterial. Separation examplesprovided above suggest that it ispossible to capture proteins directlywithout prior dilution; in fact, theadsorbent operates near to thephysiological ionic strength,constituting a strong advantage overmore classical ion-exchange sorbentsused for similar applications.

In the presence of cells or celldebris, the adsorption properties andfluidization dynamics were bothsatisfactory, exceeding the operatinglimitations of existing sorbentsdesigned for the application.Relatively small bead sizes prevent usewith contactors using large-diameternets; however, a simple solution is touse contactors with lateral inlets andperform all operations including theelution step in FB mode. Proteinelution typically occurs in a largervolume with FB than with packed-

TTaabbllee 22:: Comparative purification of lysozyme from E. coli lysate at two biomass concentrations(2.5% and 11% ww/v) in fluidized-bed mode

CM HyperZ sorbent SP Streamline sorbent11% 2.5% 11% 2.5%

Loading time (28 CV) 180 min 83 min 350 min 228 min

Linear flow for loading withan expansion of factor 2 120 cm/hr 260 cm/hr 65 cm/hr 100 cm/hr

Estimated lysozyme bindingcapacity (mg/mL) 19 25 11 23.5

Estimated recovery (%) 68% 89% 62% 84%

TTaabbllee 11:: Purification of mouse IgG1 from hybridoma cell culture by packed-bed chromatographywith clarified supernatant and by fluidized bed with cell culture supernatant containing cells

PACKED-BED CHROMATOGRAPHY

Volume Absorbance IgGStep Solution (mL) @280 nm µg/mL Purity

Load Supernatant 300 1.21 28 N/A

Flowthrough 320 1.03 0 N/A

Wash Equilibration 45 0.01 0 N/Abuffer

Elute 1 M NaCl 34 0.40 282 85%

FLUIDIZED-BED CAPTURE

Volume Absorbance IgGStep Solution (mL) @280 nm µg/mL Purity

Load Cell culture 400 7.60 56 N/A

Flowthrough 430 6.00 1 N/A

Wash Equilibration 47 0.44 0 N/Abuffer

Elute 1 1 M NaCl 8 2.00 714

Elute 2 1 M NaCl 38 2.00 528 70%

Page 8: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

58 BioProcess International OCTOBER 2003

bed desorption because of therequired elution volume and volumerepresented by the bed expansionfactor. In most cases, the latter ispractically negligible compared withthe initial feedstock volume, and itdoes not significantly influence theconcentration effect of FB capture. Asrequired, elution volumes can bediminished by reducing the expansionfactor during elution phase.

ACKNOWLEDGMENT

The authors kindly express theirgratitude to Jim Spencer from Ciphergenin Fremont, CA, for his continuous andcritical support throughout ourpreparation of this manuscript.

REFERENCES1 Hjorth, R; Kampe, S; Carlsson, M.

Analysis of Some Operating Parameters ofNovel Adsorbents for Recovery of Proteins inExpanded Beds. Bioseparation 1995, 5;217–223.

2 Mattiasson, B; Nandakumar, MP.Physicochemical Basis of Expanded-BedAdsorption for Protein Purification. InSeparation Science and Technology, Vol. 2;Academic Press, London, 2000; pp.417–430.

3 Beck, JT; Williamson, B; Tipton B.Direct Coupling of Expanded BedAdsorption with a Downstream PurificationStep. Bioseparation 1999, 8; 201–207.

4 Anspach, FB; et al. Expanded-bedChromatography in Primary ProteinPurification. J. Chromatogr. 1999, 865;129–144.

5 Mullick A; Flickinger, MC. ExpandedBed Adsorption of Human Serum Albuminfrom Very Dense Saccharomyces CerevesiaeSuspensions on Fluoride-Modified Zirconia.Biotechnol. Bioeng. 1999, 65; 282–290.

6 Willoughby, NA; et al. ImmobilisedMetal Ion Affinity ChromatographyPurification of Alcohol Dehydrogenase fromBaker’s Yeast Using an Expanded BedAdsorption System. J. Chromatogr. 1999,840; 195–204.

7 Batt, BV; Yabannavar, VM; Singh, V.Expanded Bed Adsorption Process forProtein Recovery from Whole MammalianCell Culture Broth. Bioseparation 1995, 5;42–53.

8 Blank, GS; et al. Expanded BedAdsorption in the Purification of MonoclonalAntibodies: A Comparison of ProcessAlternatives. Bioseparation 2001, 10; 65–71.

9 Fass, R; et al. Use of High-DensityCultures of Escherichia Coli for High LevelProduction of Recombinant PseudomonasAeruginosa Exotoxin A. Appl. Microbiol.Biotechnol. 1991, 36; 65–59.

10 Barnfield, AK; Hjorth, R;Hammarstrom, A. Pilot Scale Recovery ofRecombinant Annexin V from UnclarifiedEscherichia Coli Homogenate UsingExpanded Bed Adsorption. Biotechnol.Bioeng. 1994, 44; 922–929.

11 Johansson, HJ; Jagersten, J; Shiloach,J. Large Scale Recovery and Purification ofPeriplasmic Recombinant Protein from E.Coli Using Expanded Bed AdsorptionChromatography Followed by New IonExchange Media. J. Biotechnol. 1996, 48;9–14.

12 Noronha, S; Kaufman, J; Shiloach, J.Use of Streamline Chelating for Capture andPurification of Poly-His-TaggedRecombinant Proteins Bioseparation 1999, 8;145–151.

13 Ozyurt, S; Kirdar, B; Ulgen, KO.Recovery of Antithrombin III from Milk byExpanded Bed Chromatography J.Chromatogr. 2002, 944; 203–210.

14 Finette, GM; Mao, QM; Hearn, MT.Optimization Considerations for thePurification of Alpha (1)-Antitrypsin UsingSilica-Based Ion-Exchange Adsorbents inPacked and Expanded Beds. J. Chromatogr.1996, 743; 57–73.

15 Pai, A; Gondkar, S; Lali, A. EnhancedPerformance of Expanded BedChromatography on Rigid SuperporousAdsorbent Matrix. J. Chromatogr. 2000, 867;113–130.

16 Nayak, DP; Ponrathnal, S; Rajan, CR.Macroporous Copolymer Matrix. IV.Expanded Bed Adsorption Application. J.Chromatogr. 2001, 922; 63–76.

17 Voute, N; et al. PerformanceEvaluation of Zirconium Oxide BasedAdsorbents for the Fluidized Bed Capture ofMAb. Int. J. BioChromatogr. 2000, 5; 49–65.

18 Gilchrist, GR; Burns, MT; Lyddiatt,A. Solid Phases for Protein Adsorption inLiquid Fluidized Beds. In Separation forBiotechnology 3; Pyle, DL (Ed.); Royal Societyof Chemistry, London, 1994; p. 186.

19 Coffman, JL; Boschetti; E. EnhancedDiffusion and Related Sorbents forBiopurification. In Bioprocessing andBioseparation, Vol 1; Subramanian, G (Ed.);Wiley-VCH Verlag Gmbh Press: Weinheim,1998; pp. 157–198.

20 Palsson, E; Gustavsson, PE; Larsson,PO. Pellicular Expanded Bed Matrix Suitablefor High Flow Rates. J. Chromatogr. 2000,878; 17–25.

21 Lan, JC; Hamilton, GE; Lyddiatt, A.Physical and Biochemical Characterization ofa Simple Intermediate Between Fluidized andExpanded-Bed Contactors. Bioseparation.1999, 8; 43–51.

22 Shiloach, J; Kennedy, R. ExpandedBed Adsorption Process for Protein Capture.In Separation Science and Technology, Vol. 2;Acad. Press., London, 2000; pp. 431–451.

23 Thommes, J; et al. Isolation ofMonoclonal Antibodies from Cell ContainingHybridoma Broth Using a Protein A CoatedAdsorbent in Expanded Beds. J. Chromatogr.,1996, 752; 111–122.

24 Zafirakos, E; Lihme, A. EBA ’96 FirstIntern. Conf. on Expanded Bed Adsorption,Cambridge UK, 1998, abstract O.12.

25 Fee, CJ; Liten, AD. Buoyancy-Induced Mixing During Wash and ElutionSteps in Expanded Bed Adsorption.Bioseparation 2001, 10; 21–30.

26 Richardson, J; Zaki, W. Sedimentationand Fluidization: Part I. Trans. I. Chem. E.1954, 32; 35–53.

27 Chase, HA; Draeger, NM. AffinityPurification of Proteins Using ExpandedBeds. J. Chromatogr. 1992, 597; 129–45.

28 Kopaciewicz, W; Fulton, S; Lee, SY.Influence of Pore and Particle Size on theFrontal Uptake of Proteins. J. Chromatogr.1987, 409; 111–124.

29 Chang, YK; Chase, HA. Ion ExchangePurification of G6PDH from UnclarifiedYeast Cell Homogenates Using ExpandedBed Adsorption. Biotechnol. Bioeng. 1996,49; 512–526.

30 Wright, PR; Muzzio, FG; Glasser, BJ.Effect of Resin Characteristics on FluidizedBed Adsorption of Proteins. Biotechnol. Prog.1999, 15; 932–940.

31 Li, Q; et al. Interparticle andIntraparticle Mass-Transfer inChromatographic Separations. Bioseparation1995, 5; 189–202.

32 Gondkar, S; et al. Effect of AdsorbentPorosity on Performance of Expanded BedChromatography of Proteins. Biotechnol.Prog. 2001, 17; 522–529.

33 Goossens, WRA. Clarification ofFluidized Particles by Archimedes Number.Powder Technology 1998, 98; 48–53.

TTaabbllee 33:: Scale-up purification of lysozymefrom E. coli lysate at 2.5% (w/v) biomassconcentration (1.0-cm ID � 10 cm laboratoryscale column with 7.8 mL sorbent; 10-cm ID� 20 cm settled bed column with 1,700 mLsorbent)

Lab Scale Pilot Scale

Loading 83 min 44 mintime @260 cm/hr @300 cm/hr

Binding 25 mg/mL 20 mg/mLcapacity

Recovery 89% 88%

Purity 80% 80%

Equilibration wash: 50 mM acetate, 5 mM citrate,50 mM NaCl at ph 4.5Elution: 50 mM acetate, 5 mM citrate, 1 M NaClat pH 4.5Sample: 2.5% biomass E. coli lysate inequilibration buffer containing 1 or 5 mg/mLlysozyme

Page 9: Effective Protein Capture in Fluidized-Bed Mode with ... · place in expanded-bed mode. Herein we discuss a composite material composed of porous zirconium oxide filled with a synthetic

34 Nesbitt, AB; Petersen, FW. A Modelfor the Prediction of Expansion of aFluidized Bed of Poly-Sized Resin. PowderTechnology 1998, 98; 258–264.

35 Karau, A; et al. The Influence ofParticle Size Distribution and OperatingConditions on the Adsorption Performancein Fluidized Beds. Biotechnol. and Bioeng.1997, 55; 54–64.

36 Voute, N; Boschetti, E. Highly DenseBeaded Sorbents Suitable for Fluidized BedApplications. Bioseparation 1999, 8;115–120.

37 Voute, N; et al. Characterization ofVery Dense Mineral Oxide-Gel Compositesfor Fluidized-Bed Adsorption ofBiomolecules. Bioseparation 1999, 8;121–129.

38 Necina, R; Anatschek, K; Jungbauer,A. Capture of Human MonoclonalAntibodies from Cell Culture Supernatant byIon Exchange Media Exhibiting HighCharge Density. Biotechnol. Bioeng. 1998, 60;689–698.

39 De Keyser, JL. High-Cell-DensityPerfused Culture of MAbs. Gen. Eng. News.2002, 22; 52.

40 Ameskamp, N; et al. Pilot ScaleRecovery of Monoclonal Antibodies byExpanded Bed Ion Exchange Adsorption.Bioseparation 1999, 8; 169–188.

41 Lutkemeyer, D; et al. Capture ofProteins from Mammalian Cells in Pilot ScaleUsing Different Streamline Adsorbents.Bioseparation 2001, 10; 57–63.

42 Shiloach, J; et al. Endostatin Capturefrom Pichia Pastoris Culture in a FluidizedBed: From Chip Process Optimization toApplication. J. Chromatogr. 790 (2003) 327-336, submitted for publication.

43 Shepard, SR; Boyd, GA; Schrimsher,JL. Routine Manufacture of RecombinantProteins Using Expanded Bed AdsorptionChromatography. Bioseparation. 2001, 10;51–56.

44 Shepard, SR; et al. Large-ScalePurification of Recombinant HumanAngiostatin. Protein Expr. Purif. 2000, 20;216–227.

45 Murasugi, A; Asami, Y; Mera-Kikuchi,Y. Production of Recombinant Human BileSalt-Stimulated Lipase in Pichia Pastoris.Prot. Expr. Purif. 2001, 23; 282–288.

46 Trinh, L; et al. Recovery of MouseEndostatin Produced by Pichia Pastoris UsingExpanded Bed Adsorption. Bioseparation.2000, 9; 223–230.

47 Brobjer, M. Development and ScaleUp of a Capture Step (Expanded BedChromatography) for a Fusion ProteinExpressed Intracellularly in Escherichia Coli.Bioseparation. 1999, 8; 219–228.

48 Pyo, SH; et al. A Large-ScalePurification of Recombinant Histone H1.5from Escherichia Coli. Protein Expr. Purif.2001, 23; 38–44. ��

Patrick Santambien, PhD, is head of theapplication laboratory, Nicolas Voute isprincipal engineer of the researchlaboratory, Anthony Schapman andVincent Ravault are scientists in theapplication laboratory, andcorresponding author Egisto Boschetti,PhD, is the research director forCiphergen Biosystems (BioprocessDivision), 48 Avenue des Genottes,96800 Cergy Pontoise, France,[email protected].

Circle Reader Service No. 137