Development of Specific Adsorbents for Human TumorNecrosis Factor- r: Influence of Antibody Immobilization on

Performance and Biocompatibility

Viktoria Weber,*,† Ingrid Linsberger,† Marion Ettenauer,† Fritz Loth,‡ Matti Hoyhtya,§ andDieter Falkenhagen†

Christian Doppler Laboratory for Specific Adsorption Technologies in Medicine, Center for BiomedicalTechnology, Danube University Krems, Krems, Austria, Fraunhofer Institute for Applied Polymer Research,

Golm, Germany, and VTT Biotechnology, Espoo, Finland

Received October 25, 2004

To develop adsorbents for the specific removal of tumor necrosis factor-R (TNF) in extracorporeal bloodpurification, cellulose microparticles were functionalized either with a monoclonal anti-TNF antibody (mAb)or with recombinant human antibody fragments (Fab). The TNF binding capacity of the adsorbents wasdetermined with in vitro batch experiments using spiked human plasma (spike: 1200 pg TNF/mL; 1 mgparticles in 250µL plasma). Random immobilization of the full-sized monoclonal antibody to periodate-activated cellulose yielded particles with excellent adsorption capacity (258.1( 48.6 pg TNF per mg adsorbentwet weight). No leaching of antibody was detectable, and the adsorbents retained their activity for at least12 months at 4°C. We found that the conditions used during immobilization of the antibody (pH, nature ofthe reducing agent) profoundly influenced the biocompatibility of the resulting adsorbents, especially withrespect to activation of the complement system. Particles obtained by random immobilization of themonovalent Fab fragments on periodate-activated cellulose using the same conditions as for immobilizationof the mAb exhibited only low adsorption capacity (44( 7 pg/mg adsorbent wet weight). Oriented couplingof the Fab fragments on chelate-epoxy cellulose via a C-terminal histidine tag, however, increased theadsorption capacity to 178.3( 8.6 pg TNF/mg adsorbent wet weight. Thus, in the case of small, monovalentligands, the orientation on the carrier is critical to retain full binding activity.

Introduction

Immobilization of antibodies or antibody fragments onsolid surfaces has been studied extensively and is widelyapplied in modern life sciences, e.g., for affinity chroma-tography, biosensors, immunoassays, as well as for thedevelopment of protein arrays.1-4 Early immobilizationtechniques which were based on the direct adsorption ofantibodies on plastic or glass often resulted in decreasedantigen-binding activity and in leaching of the antibody. Toovercome these limitations, numerous strategies for covalentimmobilization of antibodies have been developed.5-7 Mostly,antibodies are covalently bound to activated supports throughexposed amino groups (i.e., lysine side chains) on theirsurface. As an average antibody molecule contains 60-80lysine residues, this coupling method results in randomantibody orientation. In addition, multipoint attachment ofthe antibody molecules may lead to decreased activity dueto conformational changes or to obstruction of the antigenbinding site.8,9 Therefore, site-directed conjugation schemesusing cross-linking reagents that specifically react withresidues opposite to the antigen binding site have been

developed. However, the oriented immobilization of antibodyfragments with various linker systems remains contradictorywith reports on both enhanced and reduced antigen-bindingactivity after the coupling procedure.10-11

A popular strategy for specifically orienting antibodies onsurfaces involves the use of monovalent antibody fragments(Fab) which can either be obtained by enzymatic fragmenta-tion of immunoglobulin G or isolated from antibody librariesby phage display technology.12 Fab contain thiol groups distalto the antigen-binding site which can be utilized for orientedimmobilization on appropriately activated support matrixes.Binding is accomplished through the formation of disulfidebridges between the antibody fragments and the activatedsupport. In addition, recombinant antibody technology allowsfor the preparation of antibody fragments containing tags(e.g., histidine or biotin) at the C-terminus of the molecule,i.e., opposite to the antigen-binding site.

Most of the adsorbents that are currently clinically appliedfor blood purification are selective, i.e., they bind to a rangeof molecules. Interaction of the target molecules and theadsorption matrix is frequently based on physicochemicalproperties, such as charge or hydrophobicity. Although itmay be desirable for some applications to remove a groupof molecules with common characteristics from the plasma(e.g., hydrophobic substances in liver failure, or total IgGin autoimmune diseases with unknown specificity of the

* To whom correspondence should be addressed. Tel:++43 2732 8932632. Fax:++43 2732 893 4600. E-mail: viktoria.weber@donau-uni.ac.at.

† Danube University Krems.‡ Fraunhofer Institute for Applied Polymer Research.§ VTT Biotechnology.

1864 Biomacromolecules 2005,6, 1864-1870

10.1021/bm040074t CCC: $30.25 © 2005 American Chemical SocietyPublished on Web 04/21/2005

autoantibodies), a drawback of these adsorbents is theunspecific adsorption of plasma proteins or anticoagulantsduring the treatment. With increasing knowledge of thepathophysiology of certain diseases, there is a trend towardthe development of specific adsorbents that target individualfactors. A major advantage of this approach is that moleculeswhich do not need to be removed will remain in the plasma,which reduces adverse effects of the therapy.

Our work aims at the development of specific adsorbentsbased on cellulose microspheres which are functionalizedwith antibodies or antibody fragments. These particles shallbe applied in the microspheres-based detoxification system(MDS).13-14 The MDS represents an alternative approach toconventional blood purification systems, in which thepatient’s plasma does not perfuse an adsorption column, butis recirculated into the filtrate compartment of the module(Figure 1). Addition of adsorbent microparticles15 to theplasma circuit allows for the rapid removal of pathogenicsubstances. A major advantage of the system is its highflexibility, as it can be applied for the treatment of variousdiseases depending on the specificity of the adsorbents used.Furthermore, adsorbents of different specificity can becombined according to the patient’s needs. Potential ap-plications of the MDS include the treatment of autoimmunediseases (adsorption of autoantibodies), rheologic diseases(adsorption of fibrinogen), as well as sepsis (adsorption ofinflammatory cytokines, complement factors, and otherpathogenic mediators). We wished to develop a generallyapplicable procedure for the preparation of specific adsor-bents and to evaluate the factors contributing to adsorbentefficiency and biocompatibility. Adsorption of tumor necrosisfactor-R was chosen as a model, since TNF is a prominentmediator of cellular immune response and inflammationwhich has been shown to be a key player in the onset ofseptic shock.16,17 On the other hand, TNF is a pleiotropicfactor that may be needed for the resolution of an infection.From this point of view, extracorporeal blood purificationhas clear benefits over systemic administration of antibodiesblocking TNF, as it allows for a modulation of TNFconcentration in the circulation.

The characteristics of the support matrix used for im-mobilization of specific ligands profoundly influence the

performance of the resulting adsorbents. Desirable matrixcharacteristics are (1) open porous structure with innersurface accessible for target molecules, (2) mechanicallystable spherical beads of narrow size distribution, (3) easyderivatization, and (4) chemical stability. In addition, thebiocompatibility of adsorbents used for blood purificationis critical, as the particles are in direct contact with thepatient’s plasma during the treatment. Biocompatibility ofan adsorbent is dependent on the characteristics of bothsupport matrix and ligands. Cellulose represents an idealsupport matrix, as it combines all of the above-mentionedfeatures. In addition, cellulosic polymers are widely appliedin medicine and generally regarded as biocompatible.However, due to its structural similarity to the polysaccharidechains found on bacterial cell walls, cellulose initiatesactivation of the complement system,18 which evolved as animportant element of the innate immune defense againstmicroorganisms and foreign materials. Through generationof the anaphylotoxins C3a and C5a, complement activationleads to a series of secondary inflammatory effects. There-fore, complement compatibility of the adsorbents developedin this study was evaluated by incubation of the particles inwhole blood and quantification of C3a release. It is dem-onstrated that the chemistry used for functionalization of thesupport matrix profoundly influences the biocompatibilityof the resulting particles.

Materials and Methods

Chemicals.Recombinant human TNF was obtained fromR&D Systems (McKinley Place, Minneapolis). Sodiumborohydride, sodium meta-periodate, ethylene glycol, epi-chlorohydrine, and iminodiacetic acid were purchased fromSigma-Aldrich (St. Louis, Missouri). Fresh frozen humanplasma was obtained from a local plasma donation center.Cellulose microparticles were provided by the FraunhoferInstitute for Applied Polymer Research (Golm, Germany).

Determination of Protein Concentration. Protein con-centration was determined with the Biorad Protein Assay(Biorad Laboratories, Hercules, California) using bovineserum albumin as a standard.

Preparation of Periodate- Activated Cellulose Beads.Aliquots of 20 g of water swollen cellulose microparticleswere mixed with 100 mL of 0.1 M sodium meta-periodatesolution. Following gentle shaking of the mixture for 2 h at40 °C in the dark, ethylene glycol (20 mL) was added andshaking was continued for 1 h. Particles were separated fromthe supernatant by centrifugation (4000 g) and thoroughlywashed with water. The content of dialdehyde groups(corresponding to the degree of activation) was determinedby treatment of the particles with sodium hydroxide solutionand titration of sodium hydroxide consumption. Typicaldegrees of activation were in the range of 300-360 µmolper g dry matter.

Preparation of Chelate-Epoxy Cellulose Beads.Chelate-epoxy cellulose beads were prepared according to a modifiedmethod of Pessela et al.19 Aliquots of 10 g of water swollencellulose microspheres (cellulose content 3 g) were mixedwith 25 mL of 2 M sodium hydroxide solution. Epichloro-

Figure 1. Schematic drawing of the MDS. The primary circuitcontaining the patient’s whole blood is separated from the secondarycircuit by a plasma filter. In the secondary circuit, plasma isrecirculated and kept in suspension together with the adsorbentparticles. The purified plasma is filtered back to the primary circuit,and whole blood is reinfused into the patient.

Adsorbents for Human Tumor Necrosis Factor-R Biomacromolecules, Vol. 6, No. 4, 2005 1865

hydrine (12.5 mL) was added and the mixture was shakenfor 30 min at 50°C. Finally, the beads were thoroughlywashed with distilled water. The content of epoxy groups(degree of activation) was determined by treatment withsodium thiosulfate and titration of the formed hydroxyl ionswith hydrochloric acid. Typical preparations contained about470 µmol epoxy groups per gram of dry matter. Subse-quently, the activated cellulose microspheres were incubatedin a solution of 0.2 M iminodiacetic acid (IDA) in 2 Msodium carbonate for 2 h at 25 °C. After washing withdistilled water, samples contained approximately 250µmolepoxy groups per gram of dry matter. Aliquots of 6 g ofthis sample were resuspended in 40 mL of 50 mM phosphatebuffer, pH 6.0, containing 100 mg of CoCl2, incubated for 2h at 25°C, and thoroughly washed with distilled water. Thecontent of free epoxy groups was about 170µmol/g drymatter. The activated microspheres were immediately usedfor antibody immobilization, as a loss of epoxy groups occurson storage.

Antibodies. A chimeric monoclonal antibody (mAB)against TNF (Remicade, Infliximab) was obtained from thelocal pharmacy. Infliximab consists of the variable regionsof a mouse anti-TNF mAb linked to human IgG1 withκlight chains.20 It binds to the TNF homotrimer with highaffinity (Ka ) 1010 M-1) and is currently approved for thetreatment of Crohn’s disease and rheumatoid arthritis.21-23

Recombinant human Fab fragments against TNF wereproduced at VTT Biotechnology using phage display tech-nology. The phage library was constructed from the lym-phocytes of 50 volunteer blood donors obtained from theFinnish Red Cross. The library is in single chain format andkappa (2.1× 107 clones) and lambda (7.4× 107 clones)light chains are kept in separate libraries both having thesame heavy chain variable repertoire. The libraries wereselected through five panning rounds against recombinantTNF (Chemicon, Temecula, California) using standardprotocols. Four positive clones against TNF were found, threefrom the kappa library and one from the lambda library. Oneclone, KC10, was selected for further development. It iscomposed of VK4/JK1 and VH1/D3/JH3a sequences ac-cording to the MRC Centre for Protein Engineering database.The affinity constant of KC10 is 4× 108 M-1 as determinedwith BIAcore analysis.

The single chain fragments were transformed into Fabfragments by cloning human kappa and human IgG1 CH1constant regions with overlapping PCR. Two separate affinitytags were introduced, a hexahistidine tag into the C-terminusof the heavy chain and a biotin tag (Avidity, Denver,Colorado) into the C-terminus of the light chain, for orientedcoupling of the fragments. The fragments were produced inhigh-density fed-batch cultures (4-5 liters) in a Bio-FlowIV bioreactor (New Brunswick, Edison, New Jersey) usingE.coli host strain RV308 for the histidine and AVB100 forthe biotin tagged product. Fragments were purified withProtein G fast flow (Amersham Biosciences, Freiburg,Germany) affinity chromatography.

Random Immobilization of Infliximab and Fab onPeriodate-Activated Cellulose Beads.Periodate-activatedcellulose particles (30 mg wet weight) were washed 3 times

with phosphate buffered saline (PBS; pH 7.4). Particles wereseparated from the supernatant by centrifugation (3 min, 4000g). After a final wash with 0.1 M sodium borate (pH 9.5),the particles were suspended in 6 mL of 0.1 M sodium borate(pH 9.5), and 100µL of a solution containing 10 mg/mLInfliximab in PBS was added. The mixture was brieflyvortexed and incubated overnight (16-20 h) at 37°C withgentle shaking. Beads were separated from the supernatantby centrifugation and washed with PBS. After completeremoval of the supernatant, 100µL of sodium borohydrideor sodium cyanoborohydride, respectively, (2 mg/mL in PBS)was added to reduce Schiff bases. After incubation for 60min at room temperature, particles were removed bycentrifugation and washed 3 times with PBS. The adsorbentswere suspended in PBS and kept at 4°C until furtheranalysis. Coupling efficiency was calculated from the amountof protein (i.e., antibody) in the supernatant before and aftercoupling. Typically, between 30 and 35% of the full-sizedantibody used in an experiment were immobilized on thebeads. Fab fragments were randomly immobilized on pe-riodate-activated cellulose exactly as described for Infliximabusing 500, 1000, and 2500µg of Fab per 30 mg (wet weight)of activated cellulose. Coupling efficiencies for Fab are givenin Table 2.

Leaching of Full-Sized Monoclonal Antibody.Adsorbentsuspensions were centrifuged (4000 g, 3 min) and superna-tants were assayed for leached antibody by human anti-κ

light chain ELISA (Bethyl Laboratories, Montgomery,Texas).

Oriented Immobilization of Histidine-Tagged Fab onCo-Chelate Epoxy Cellulose Beads.Immobilization of Fabon Co-chelate epoxy beads was performed according toPessela et al.19 (Scheme 1). Aliquots of the particles (30 mg

Table 1. TNF Adsorption Capacities (pg TNF/mg Adsorbent WetWeight) of Microparticles Obtained by Random Immobilization ofthe Full-Sized Monoclonal Antibody on Periodate-ActivatedCellulose at Different Values of pH (Mean of 3 Experiments ( SD)

pHTNF adsorption

(pg/mg adsorbent wet weight) ( SD

9.5 258.1 ( 497.4 210 ( 325.5 98 ( 26

Table 2. Coupling Efficiencies and Amounts of TNF Bound forImmobilization of Fab to Periodate-Activated Cellulose (PAC) andChelate-Epoxy Cellulose (CEC)a

matrix

amount ofFab used

for coupling to30 mg beads

(µg)

couplingefficiencyb

(%)

Fab immobilizedon the particles

(µg per mg)

TNF bound toimmobilized

Fab(pg per µg)

PAC 500 100 17.5 ( 0.9 1.7 ( 0.1PAC 1000 57 20 ( 2.4 1.6 ( 0.2PAC 2500 23 21 ( 3.1 1.9 ( 0.2CEC 500 100 17.5 ( 0.7 5.9 ( 0.6CEC 1000 80 28 ( 2.2 5.7 ( 0.4CEC 2500 35 29 ( 1.8 6.1 ( 0.5

a Data are given as means of 3 experiments ( SEM. b Calculated fromthe amount of protein (i.e., antibody) in the supernatant before and aftercoupling.

1866 Biomacromolecules, Vol. 6, No. 4, 2005 Weber et al.

wet weight) were incubated overnight with 500, 1000, and2500µg of histidine-tagged Fab in 30 mM phosphate buffer(pH 7.0). Beads were separated from the supernatant bycentrifugation (4000 g, 3min) and treated with 5 mL of 1Mglycine (pH 8.5) for 4 h atroom temperature to block residualepoxy groups. Adsorbents were washed and stored in 30 mMphosphate buffer (pH 7.0) at 4°C.

Adsorption of TNF. Binding of TNF to the differentadsorbents was evaluated in batch-experiments using spikedhuman plasma (1200 pg TNF /ml). Adsorbents were mixedwith freshly spiked human plasma (1 mg of particles in 250µL of plasma) and incubated in an end-over-end shaker for60 min. Samples from the supernatant were taken after 5,30, and 60 min of incubation, and TNF was quantified byELISA (Biosource, Nivelles, Belgium).

Activation of the Complement System.Blood was drawnfrom healthy volunteers and anticoagulated with heparin (3IU/mL). Adsorbent particles were incubated in freshly drawnblood at 37°C using 100 mg of adsorbent and 1900µL ofblood. Samples were taken after 5 and 60 min of incubation,respectively, and the reaction was stopped by addition of 80µL of ice-cold sodium-EDTA (30% w/v in 0.9% NaCl).Particles were removed by centrifugation, and C3adesArgas a marker of complement activation was quantified in thesupernatant by ELISA (Progen, Heidelberg, Germany).

Results and Discussion

Cellulose Microparticles as Starting Material for Ac-tivation. Figure 2 shows an electron micrograph of thecellulose particles used as starting material both for oriented

and random immobilization of the antibodies. The particlesare highly porous and in a size range of 1-10 µm. Particlesize distribution remained unchanged upon recirculation ofa suspension (50% v/v of particles in 0.9% saline) at a flowrate of 4L/min for 30 h, proving that the particles aremechanically stable.

Immobilization of Full-Sized Antibody on PeriodateActivated Cellulose. Adsorption capacities of particlesobtained by random immobilization of the full-sized mono-clonal antibody at different values of pH are summarized inTable 1. Adsorption capacity was highest for particlesobtained by coupling at pH 9.5 according to the protocoldescribed in Materials and Methods. However, high pHduring the coupling reaction led to particles which strongly

Scheme 1: Oriented Immobilization of Histidine-Tagged Fab Fragments on Chelate-Epoxy Cellulosea

a Cellulose microparticles are activated with epichlorohydrine and incubated with iminodiacetic acid to introduce IDA into the support. IDA-epoxy celluloseis then incubated in the presence of cobalt salts to obtain chelate-epoxy cellulose. This support is used to immobilize the histidine-tagged Fab fragmentsin an oriented manner.

Figure 2. Electron micrograph of cellulose microparticles used ascarriers for antibodies and antibody fragments, respectively.

Adsorbents for Human Tumor Necrosis Factor-R Biomacromolecules, Vol. 6, No. 4, 2005 1867

activated the complement system (see below). Therefore,further immobilization reactions were performed at pH 7.4.To assess their stability, adsorbent suspensions were storedin PBS (pH 7.4, 0.3% sodium azide), and samples wereanalyzed for leaching of antibody and TNF binding capacityafter 3, 6, and 12 months of storage, respectively. No lossof activity was observed over a period of at least 12 months.Leaching of antibody from the particles was tested byanalyzing supernatants of particle suspensions for humanκ-light chains. Antibody concentration in the supernatantremained below the detection limit of 40 ng/mL in allsamples.

Stability of Chelate-Cellulose Microparticles.To assessthe stability of the chelate-cellulose microparticles, samplesfrom consecutive activation steps (before and after reactionwith IDA and after addition of Co2+) were stored at 4°Cand at 23°C, respectively, and their epoxy group contentwas analyzed over time. Starting from 470µmol/g, there wasa slow but constant decrease in epoxy group content whichreached a plateau at 390 and 310µmol/g at 4° and 23°C,respectively, after 300 h of storage. The starting level ofepoxy groups in the IDA treated sample was 250µmol/g, asabout half of the epoxy groups had already been consumedfor binding of IDA. Again, the epoxy group contentdecreased over time to 200µmol/g after 300 h at 4°C.Addition of Co2+ to the IDA treated sample resulted in afurther significant decrease in epoxy groups from 250 to 150µmol/g indicating that Co2+ might exert a catalytic effecton the hydrolysis of epoxy groups. Therefore, immobilizationof protein ligands should be carried out immediately afteractivation of the cellulose particles to avoid excessive lossof epoxy groups. After immobilization of the Fab fragments,the resulting adsorbents fully retained their activity for atleast 6 months.

Adsorption of TNF: Random versus Oriented Im-mobilization of Full-Sized Antibody and Fab Fragments.Random immobilization of the full-sized antibody to perio-date-activated cellulose beads yielded particles with highTNF adsorption capacity (258( 49 pg TNF/mg adsorbentwet weight) as well as high adsorption rate (Figure 3).

In contrast, particles obtained by random immobilizationof Fab to periodate-activated cellulose exhibited only weakTNF adsorption. Immobilization of 500, 1000, and 2500µgof Fab per 30 mg of particles resulted in coupling efficienciesof 100, 57, and 23%, respectively, corresponding to 17.5,20, and 21µg of Fab immobilized per mg of particles. Thus,use of higher amounts of Fab (2500µg vs 500µg) did notresult in significantly increased immobilization of Fab (Table2). TNF adsorption was 44( 7 pg/mg adsorbent wet weightafter 60 min for the series using 2500µg Fab/30 mg particles(Figure 3).

For Fab immobilized to chelate-epoxy cellulose using 500,1000, and 2500µg of Fab per 30 mg of particles, couplingefficiencies were 100, 80, and 35%, respectively, corre-sponding to 17.5, 28, and 29µg of Fab immobilized per mgof particles (Table 2). Oriented coupling considerablyimproved TNF binding capacity to 178.3( 8.6 pg TNF/mgadsorbent wet weight. In comparison, 6.1 pg of TNF wereadsorbed perµg of Fab immobilized on chelate-epoxy

cellulose, as compared to 1.9 pg TNF perµg of Fabimmobilized on periodate-activated cellulose.

As shown in a control experiment, unmodified cellulosemicroparticles exhibited only negligible binding of TNF.

Influence of Coupling Conditions to Periodate-Acti-vated Cellulose on Biocompatibility. To measure thecomplement activating activity of the cellulose-based TNFadsorbents, generation of C3adesArg was quantified afterincubation of the adsorbents in whole blood. As shown inFigure 4, complement activation was in the same range forperiodate-activated cellulose as for untreated cellulose. Twoparameters during the antibody immobilization procedurewere found to significantly influence the complementactivating activity of the adsorbents, namely (1) the nature

Figure 3. Adsorption of TNF by specific adsorbents in batchexperiments. Samples were taken after 5, 30, and 60 min ofincubation and TNF was quantified by ELISA. Ctrl, unmodifiedcellulose; Mab-random, monoclonal antibody randomly immobilizedto periodate-activated cellulose (PAC); Fab-random, Fab fragmentsrandomly immobilized to PAC; Fab-oriented, Fab in oriented im-mobilization on chelate-epoxy cellulose. Both for Fab-random andFab-oriented, the results for particles obtained by coupling 2500 µgof Fab per 30 mg of cellulose are shown. Data are given as mean ofthree experiments ( SEM.

Figure 4. Influence of coupling conditions on complement activatingactivity of the adsorbents. Generation of C3adesArg was quantifiedafter incubation of blood with (1) untreated cellulose microparticles,(2) PAC; (3) PAC after immobilization of monoclonal antibody at pH7.4 and reduction with sodium cyanoborohydride; (4) PAC aftertreatment with sodium cyanoborohydride, but without antibody (pH7.4); (5) PAC after immobilization of antibody and reduction withsodium cyanoborohydride at pH 9.5; (6), as (5), but with borohydrideinstead of cyanoborohydride.

1868 Biomacromolecules, Vol. 6, No. 4, 2005 Weber et al.

of the reductant used for stabilization of Schiff bases (sodiumborohydride vs sodium cyanoborohydride) and (2) the pHduring the coupling reaction. Adsorbents obtained by reduc-tion with borohydride exhibited significantly higher comple-ment activation than particles treated with cyanoborohydride.Borohydride reduces aldehyde groups to hydroxyls andconverts Schiff bases to secondary amines, whereas cy-anoborohydride is milder24 and effects Schiff base reductionwithout reducing aldehydes (Scheme 2). We hypothesize thatit is the generation of additional neighboring hydroxyl groupsduring the reduction with borohydride that triggers comple-ment activation. This is supported by the finding thatdeposition of the complement protein C3b on biomaterialsincreases with increasing surface hydroxyl density.25 Regard-ing the influence of the pH during the coupling reaction,complement activation was strongly enhanced in adsorbentsproduced at pH 9.5 as compared to pH 7.4. This may beexplained by the fact that, due to cleavage of the C2-C3-bonds of the glucose rings, the cellulose backbone isweakened after activation with periodate and thereforebecomes susceptible to high pH. For this reason, furthercoupling reactions were performed at pH 7.4. Remarkably,complement activation was attenuated after immobilizationof the antibody (Figure 5, bars 3 vs 4).

Conclusions

Covalent immobilization of a full-sized anti-TNF antibodyor of Fab fragments specific for TNF to cellulose micropar-ticles yielded adsorbents that were highly efficient both interms of adsorption capacity and kinetics. The full-sizedantibody exhibited excellent binding capacity after randomimmobilization to periodate activated cellulose, indicatingthat the orientation of the ligand on the support was notcritical for performance of the adsorbent. In contrast, Fabfragments immobilized under identical conditions resultedin very low adsorption capacity. Although this may partlybe attributed to their lower affinity constant for TNF ascompared to the full-sized antibody, we demonstrated that

oriented immobilization of the antibody fragments led to asignificant increase in binding capacity of the resultingadsorbents implying that, for small, monovalent antibodyfragments, the orientation on the carrier is critical. Recom-binant antibody technology offers various possibilities tointroduce tags for oriented immobilization into the antibodyfragments. In our study, a C-terminal histidine tag was usedin combination with iminodiacetic acid-epoxy activation ofthe cellulose microparticles. In a two-step reaction, thehistidine tag interacts with immobilized iminodiacetic acidleading to correct orientation of the Fab on the surface priorto formation of covalent bonds via amino groups of theantibody.

For random immobilization on periodate activated cel-lulose, it was shown that the biocompatibility of the ad-sorbents depends on the characteristics of both supportmatrix and ligand, as well as on the conditions used duringthe immobilization reaction.

In summary, the attachment of antibodies or antibodyfragments to activated cellulose microparticles offers amethod to produce highly efficient and specific adsorbentsfor the removal of a variety of pathogenic factors inextracorporeal blood purification.

Acknowledgment. This work was supported by theChristian Doppler Forschungsgesellschaft.

References and Notes

(1) Peluso P.; Wilson, D. S.; Do, D.; Traan, H.; Venkatasubbaiah, M.;Quincy, D.; Heidecker, B.; Poindexter, K.; Tolani, N.; Phelan, M.;Witte, K.; Jung, L. S.; Wagner, P.; Nock, S. Optimizing antibodyimmobilization strategies for the construction of protein microarrays.Anal. Biochem. 2003, 312, 113-124.

(2) Vankova, R.; Gaudinova´, A.; Sussenbekova´, H.; Dobrev, P.; Strnad,M.; Holık, J.; Lenfeld, J. Comparison of oriented and randomantibody immobilization in immunoaffinity chromatography ofcytokinins.J. Chromatogr. A1998, 811, 77-84.

(3) Lu, B.; Smyth M. R.; O’Kennedy, R. Oriented immobilization ofantibodies and its applications in immunoassays and immunosensors.Analyst1996, 121, 29R-32R.

(4) Reimer, U.; Reineke, U.; Schneider-Mergener, J. Peptide arrays: frommacro to micro.Curr. Opin. Biotechnol. 2002, 13, 315-320.

(5) Danczyk, R.; Krieder, B.; North, A.; Webster, T.; HogenEsch, H.;Rundell, A. Comparison of antibody functionality using differentimmobilization methods.Biotechnol. Bioeng. 2003, 84 (2), 215-223.

(6) Brogan, K. L.; Wolfe, K. N.; Jones, P. A.; Schoenfisch, M. directoriented immobilization of F(ab’) antibody fragments on gold.Anal.Chim. Acta, 2003, 496, 73-80.

(7) Wimalasena, R. L.; Wilson, G. S. Factors affecting the specificactivity of immobilized antibodies and their biologically activefragments.J. Chromatogr. 1991, 572, 85-102.

(8) Subramanian, A.; Velander, W. H. Effect of antibody orientation onimmunosorbent performance.J. Mol. Recognit. 1996, 9, 528-535.

(9) Turkova, J. Oriented immobilization of biologically active proteinsas a tool for revealing protein interactions and function.J. Chro-matogr. B1999, 722, 11-31.

(10) Wilchek, M.; Miropn, T. Oriented versus random protein immobiliza-tion. J. Biochem. Biophys. Methods2003, 55, 67-70.

(11) Reference deleted in revision.(12) Hoogenboom, H. R. Overview of antibody phage-display technology

and its applications.Methods Mol. Biol. 2002, 178, 1-37.(13) Falkenhagen, D.; Schima, H.; Loth, F. Arrangement for removing

substances from liquids, in particular blood. United States Patent, #5,855,782, 1999.

(14) von Appen, K.; Weber, C.; Losert, U.; Schima, H.; Gurland, H. J.;Falkenhagen, D. Microspheres based detoxification system: a newmethod in convective blood purification.Artif. Organs.1996, 20,420-425.

Scheme 2: Reduction of Periodate-Activated Cellulose withBorohydride and Cyanoborohydride, Respectivelya

a Borohydride is specific for the Schiff base structure and does not affectthe original aldehyde groups. By contrast, cyanoborohydride convertsunreacted aldehydes into hydroxyls in addition to Schiff base reduction.

Adsorbents for Human Tumor Necrosis Factor-R Biomacromolecules, Vol. 6, No. 4, 2005 1869

(15) Vollenkle, C.; Weigert, S.; Ilk, N.; Egelseer, E.; Weber, V.; Loth,F.; Falkenhagen, D.; Sleytr, U. B.; Sa´ra, M. Construction of afunctional S-layer fusion protein comprising an immunoglobulinG-binding domain for development of specific adsorbents forextracorporeal blood purification.Appl. EnViron. Microbiol. 2004,1514-1521.

(16) Fiers, W. Tumor necrosis factor. Characterization at the molecular,cellular, and in vivo level.FEBS Lett. 1991, 285, 199-212.

(17) Tracey, K. J.; Cerami, A. Tumor necrosis factor: a pleiotropic cyto-kine and therapeutic target.Annu. ReV. Med. 1994, 45, 491-503.

(18) Johnson R. J. Complement activation during extracorporealtherapy: biochemistry, cell biology and clinical relevance.Nephrol.Dial. Transplant. 1994, 9, 36-45.

(19) Pessela, B. C. C.; Mateo, C.; Carrascosa, A. V.; Vian, A.; Garcı´a J.L.; Rivas, G.; Alfonso, C.; Guisan, J. M.; Ferna´ndez-Lafuente, R.One-Step Purification, Covalent Immobilization, and AdditionalStabilization of a Thermophilic Poly-his-Taggedâ-Galactosidase fromThermussp. Strain T2 by using Novel Heterofunctional Chelate-Epoxy Sepabeads.Biomacromolecules2003, 4, 107-113.

(20) Knight, D. M.; Trinh, H.; Le, J.; Siegel, S.; Shealy, D.; McConough,M.; Scallon, B.; Moore, M. A.; Vilcek, J.; Daddona, P.; Ghrayeb, J.

Construction and initial characterization of a mouse-human chimericanti-TNFantibody.Mol. Immunol. 1993, 30, 1443-1453.

(21) Targan, S. R.; Hanauer, S. B.; van Deventer, S. J.; Mayer, L.; Present,D. H.; Braakman, T.; DeWoody, K. L.; Schaible, T. F.; Rutgeerts,P. J. A short-term study of the chimeric monoclonal antibody cA2to tumor necrosis factor alpha for Crohn’s disease.N. Engl. J. Med.1997, 337 (15), 1029-1035.

(22) Klinkhoff, A. Biological agents for rheumatoid arthritis: targetingboth physical function and structural damage.Drugs2004, 64, 1267-83.

(23) Olsen, N. J.; Stein, C. M. New drugs for rheumatoid arthritis.N.Engl. J. Med.2004, 350 (21), 2167-79.

(24) Peng, L.; Calton, G. J.; Burnett, J. W. Effect of borohydride reductionon antibodies.Appl. Biochem. Biotechnol. 1987, 14, 91-99.

(25) Hirata, I.; Hioki, J.; Toda, M.; Kitazawa, T.; Murakami, Y.; Kitano,E.; Kitamura, H.; Ikada, Y.; Iwata, H. Deposition of complementprotein C3b on mixed self-assembled monolayers carrying surfacehydroxyl and methyl groups studied by surface plasmon resonance.J. Biomed. Mater. Res. 2003, 66A, 669-676.

BM040074T

1870 Biomacromolecules, Vol. 6, No. 4, 2005 Weber et al.