Journal of Immunological Methods 318 (2007) 88–94www.elsevier.com/locate/jim

Research paper

The assembly of single domain antibodies intobispecific decavalent molecules

Emily Stone, Tomoko Hirama, Jamshid Tanha, Hong Tong-Sevinc, Shenghua Li,C. Roger MacKenzie, Jianbing Zhang ⁎

Institute for Biological Sciences, National Research Council of Canada, Ottawa, Ontario, Canada K1A 0R6

Received 27 April 2006; received in revised form 1 August 2006; accepted 2 October 2006Available online 17 November 2006

Abstract

Bispecific antibodies present unique opportunities in terms of new applications for engineered antibodies. However, designingideal bispecific antibodies remains a challenge. Here we describe a novel bispecific antibody model in which five single domainantibodies (sdAbs) are fused via a linker sequence to the N-terminus of the verotoxin B (VTB) subunit, a pentamerization domain,and five sdAbs are fused via a linker sequence to the VTB C-terminus. Fifteen such decavalent bispecific molecules, termeddecabodies, were constructed and characterized for the purpose of identifying an optimal decabody design. One of the fifteenmolecules existed in a non-aggregated decavalent form. In conjunction with the isolation of sdAbs with the desired specificitiesfrom non-immune phage display libraries, the decabody strategy provides a means of generating high avidity bispecific antibodyreagents, with good physical properties, relatively quickly.Crown Copyright © 2006 Published by Elsevier B.V. All rights reserved.

Keywords: Bispecific antibody; Decabody; Single domain antibody; Phage display

1. Introduction

Bispecific antibodies (bsAbs) bind to two distinctantigens due to their dual specificity. In tumor therapy,bispecific antibodies specific for tumor cells and effectorcells can mediate the localization of effector cells such ascytotoxic lymphocytes (De Jonge et al., 1998), naturalkiller cells (Ferrini et al., 1991), neutrophils (Heijnenet al., 1997) and monocytes/macrophages (Somasun-daram et al., 1996) at tumor sites. Similarly, bispecific

Abbreviations: bsAb, bispecific antibody; gan, gene accessionnumber; Gb3, globotriaosylceramide; KD, dissociation constant; sdAb,single domain antibody; VTB, verotoxin B subunit.⁎ Corresponding author. Tel.: +1 613 998 3373; fax: +1 613 952 9092.E-mail address: jianbing.zhang@nrc.ca (J. Zhang).

0022-1759/$ - see front matter. Crown Copyright © 2006 Published by Elsdoi:10.1016/j.jim.2006.10.006

antibodies could potentially deliver drugs (Ford et al.,2001), toxins (Bonardi et al., 1993), radionuclides (Gestinet al., 2001) or oligonucleotides (Mirochnik et al., 2004)as well. Monospecific antibodies are presently used ascancer therapeutics but are only effective in instanceswhere antigen binding itself has a therapeutic effect.

Early attempts at bsAb generation involved chemicalcross-linking of antibodies (Nisonoff and Rivers, 1961)and chemical engineering still remains an importantmethod for bsAb production (Graziano and Guptill,2004). However, bsAb generation by this approach hasobvious limitations such as relative process complexity,possible antibody denaturation by the chemical reactionandbatch-to-batch product variation (Cao andLam, 2003).

Overall, biological generation of bsAbs is preferableto chemical methods. Hybridoma cell lines can be fused

evier B.V. All rights reserved.

89E. Stone et al. / Journal of Immunological Methods 318 (2007) 88–94

(Cotton and Milstein, 1973) to generate triomas andquadromas, which secrete bsAbs. The major problemsassociated with this procedure are the instability ofquadromas and triomas and the time-consuming natureof the process. Recent developments in phage displayand antibody engineering technologies have greatlyfacilitated the generation of bsAbs. These techniqueshave enabled the development of bispecific diabodies(Arndt et al., 1999) and di-diabodies (Lu et al., 2003),which are non-covalent complexes comprised of two orfour single chain antibody (scFv) molecules.

Oligomerization domains such as leucine zipper fosand jun components (de Kruif and Logtenberg, 1996),the p53 dimerization domain (Rheinnecker et al., 1996),barnase–barstar (Deyev et al., 2003) and knobs intoholes structures (Ridgway et al., 1996) have also beenused to construct bsAbs. Recently a novel bispecific Ig-like molecule based on the fusion of a single domainantibody to an IgG was reported (Shen et al., 2006).

In designing bsAbs many factors need to be consid-ered. Purity, stability, avidity, size of the bsAbs, flexibilityof the antigen binding sites and immunogenicity of thegenerated molecules need all to be addressed. Despitedecades of work on bsAbs, few, if any, of the known bsAbmodels are satisfactory in all of these aspects.

Parathyroid hormone (PTH) and its related peptides(PTHrP) are used to treat osteoporosis (Whitfield, 2006).High affinity and high specific antibodies are required toestimate the blood PTH level to guide the treatment. AnsdAb, PTH50 (Zhang et al., 2004), was isolated against aPTHrP, PTH2, from a naïve llama sdAb library. Due toits naïve nature, this sdAb only has a moderate affinity.Pentameric antibodies, termed pentabodies, have beengenerated by fusing single domain antibodies (sdAbs) tothe Escherichia coli verotoxin B subunit (VTB) whichself-assembles into a homopentamer. These pentabodieshave binding avidities which are 103 to 104 fold higherthan their monomeric counterparts (Zhang et al., 2004).

In this study we presented a model of decavalentsdAb, or decabody, by fusing an sdAb to the N-terminusand another sdAb to the C-terminus of VTB.

2. Materials and methods

2.1. DNA and E. coli strains

DNA encoding D17E/W34A, G62T/W34A andF30A/G62T/W34A mutants of verotoxin B subunit(Soltyk et al., 2002) was kindly provided by Dr. J.L.Brunton, University of Toronto. E. coli TG1 waspurchased from New England Biolabs (Mississauga,Canada). Immobilized metal affinity chromatography

column was obtained from GE Healthcare (Quebec,Canada).

2.2. Construction of cloning vectors pVT1, pVT2, pVT3,pVT4 and pVT5

To facilitate the fusion of VTB to other polypeptides,including sdAbs, four vectors, pVT1 (gan: AJ619718),pVT2 (gan: AJ619719), pVT3 (gan: AJ619720) andpVT4 (gan: AJ619721) were constructed in which thewild type, the D17E/W34A, the F30A/G62T and theF30A/W34A/G62T versions of VTB genes (Soltyket al., 2002) were flanked by dual restriction sites at eachend. These mutations are located at the Gb3 receptorbinding sites of VTB. The wild type and mutants ofVTB bind to Gb3 with different functional affinities,ranging from an apparent KD of 3×10−9 M for wildtype to no detectable binding for F30A/W34A/G62T(Soltyk et al., 2002). Employing these mutants forpentabody construction will generate molecules withdifferent kidney cell-binding activities. A fifth vectorpVT5, lacking the DNA encoding the c-Myc detectiontag of pVT4, was also constructed.

These vectors have been used in the laboratory togenerate pentabodies with different affinities to humankidney cells as part of an sdAb pentamerizationtechnique (Zhang et al., 2004). In the present studyonly pVT2 is further employed.

2.3. Construction of decavalent sdAbs

Standard molecular cloning procedures were used tocreate the clones described in this work. Briefly, sdAbgenes were amplified with PCR and flanked with DNAencoding linker sequences and restriction sites BbsI/ApaI (for N-terminal fusion) or BspEI/BamHI (for C-terminal fusion). These amplified DNA was digestedwith the corresponding restriction enzymes and ligatedinto the vector pVT2 digested with the same enzymes.

2.4. Protein purification

Fusion proteins described in this work were producedas described (Zhang et al., 2004) with the modificationthat protein extraction from E. coli cells was achievedby cell lysis instead of osmotic shock. Briefly, E. colicells harboring the plasmids (Fig. 1) were grown,harvested and lysed. Following centrifugation of the celllysate, the clear supernatants were loaded onto Hi-Trap™ Chelating Affinity Columns (Amersham Bios-ciences, Piscataway, NJ) and the His5 tagged-proteinswere purified following the manufacturer's instructions.

Fig. 1. Construction of decabodies. (A) A diagram of a decavalentsdAb, or decabody. A decabody is generated when an sdAb is fused tothe N-terminus and another sdAb to the C-terminus of pentamerizationprotein VTB. (B) Schematic representation of the constructs describedin this work.

Fig. 2. Size exclusion chromatography profiles of pentavalent anddecavalent sdAbs. Approximately 0.2 mg of each protein was analyzedby Superdex 200™ chromatography in each instance. The elutionvolumes of standard molecule size markers separated under the sameconditions and their elution volumes are indicated.

90 E. Stone et al. / Journal of Immunological Methods 318 (2007) 88–94

2.5. Size exclusion chromatography

The purity and the oligomeric status of the fusionproteins were assessed by Superdex 200 (AmershamPharmacia) size exclusion chromatography. Separationswere carried out in 10 mM HEPES, pH 7.4, containing150 mM NaCl, 3.4 mM EDTA and 0.05% Tween 20.

2.6. Surface plasmon resonance analysis

Peptide antigen PTH2 (SVSEIQLMHNLGKHLN-SMERVEWLRKLLQVD in which the side chains ofresidues in bold italics are linked by a β-lactam bond) orPTH2–streptavidin conjugate are kindly provided by G.Willick. Analyses were performed on BIACORE 1000or BIACORE 3000 instruments (Biacore Inc., Piscat-away, NJ) in 10 mM HEPES, 150 mM NaCl, 3.4 mMEDTA, 0.005% P20, pH 7.4). PTH2 or PTH2–streptavidin conjugate was immobilized at a concentra-tion of 50 μg/ml on CM4 or CM5 sensor chips (BiacoreInc.) by amine coupling according to the manufacturer'sinstructions. Ethanolamine-blocked surfaces served asreference surfaces.

3. Results

3.1. Fusion of sdAbs to the N-terminus of VTB

We previously reported that fusion of sdAb PTH50 tothe C-terminus of VTB greatly improved the avidity of

91E. Stone et al. / Journal of Immunological Methods 318 (2007) 88–94

the sdAb. The resulting C-pentabody, 1V5, bound to itsantigen, a parathyroid hormone peptide, 103 to 104 foldmore strongly than its monomeric counterpart, PTH50(Zhang et al., 2004). To explore the possibility of con-structing bispecific antibodies by fusing one sdAb to theN-terminus and another to the C-terminus of VTB(Fig. 1A), it was necessary to establish that the fusion ofan sdAb to the N-terminus of VTB also generates astable and functional pentameric antibody. PTH50 andVTB, the two building blocks used in the constructionof the C-pentabody 1V5 (Zhang et al., 2004) (Fig. 1B)

Fig. 3. SPR analysis of antigen binding. (A), injection of 5 nM 1V11 overresonance units (RUs) and at a flow rate of 20 μl/min. (B), (C), (D) and (E), injto PTH2–streptavidin conjugate with a surface density of 500 RUs and a flpeptide PTH2 immobilized on a CM5 chip at a surface density of 150 RUs

were joined in the opposite orientation to that of 1V5,generating clone 1V11 (gan: AY860432, Fig. 1B). 1V11was purified by one-step immobilized metal affinitychromatography and 7.4 mg protein was obtained from1 l of E. coli culture. Purified 1V11 protein wasanalyzed by Sephadex 200™ size exclusion chroma-tography to confirm its pentameric status and purity.Using molecular size standards separated under thesame conditions, the mass of 1V11 was determined to be120 kDa, about five times of the size of its monomericsubunit (22.8 kDa), indicating that 1V11 exists as a

peptide antigen PTH2 immobilized on a CM4 chip at a level of 77ection of 10 μMPTH22, 30 μMPTH61, 10 nM 1V12 and 10 nM 1V14ow rate of 20 μl/min, respectively; (F) injection of 10 nM PJR9 overand at a flow rate of 10 μl/min.

92 E. Stone et al. / Journal of Immunological Methods 318 (2007) 88–94

pentamer (Fig. 2). The size exclusion chromatographyprofile indicated that the 1V11 preparation was freeof aggregates and degradation products. This type ofmolecule is designated an N-pentabody to distinguishfrom fusion of sdAb to the C-terminus of VTB, or C-pentabody.

The binding of 1V11 to antigen was examined bysurface plasmon resonance. The antigen bindingprofiles ofPTH50 (Zhang et al., 2004), 1V5 (Zhang et al., 2004) and1V11 (Fig. 3A) confirmed that the VTB fusion proteinsbound immobilized peptide much more effectively thanthe monomeric sdAb. PTH50 binds to immobilizedpeptide with a KD of 4×10−6 M (Zhang et al., 2004).The binding of 1V11 to immobilized peptide was analyzedat low concentrations in order to maximize the bindingvalency and assess the avidity gain conferred by sdAbpentavalency. Although the dissociation was multiphasic,a portion of the dissociation data was fitted to a 1:1interaction model to obtain an approximation of the off-rate. Under these conditions of antigen surplus, an aviditygain of 3–4 orders of magnitude was observed. Theseresults indicate that fusion of sdAb to the N-terminus ofVTB, like fusion to the C-terminus, greatly improves theavidity of the sdAb.

3.2. Construction of 1V13, a model for decavalent,bispecific sdAbs

Besides PTH50, PTH22 (gan: AF447916) andPTH61 (gan: AF447920) were also obtained from theisolation of sdAbs against PTH2 (Zhang et al., 2004).PTH61 and PTH22 bind to the same PTH peptideantigen as PTH50 but only effectively when biotinylatedpeptide is in complex with streptavidin (Fig. 3B and C).These two sdAbs were chosen as model sdAbs toconstruct the first decavalent, bispecific sdAb.

Genes encoding PTH22 and PTH61were amplified byPCR and inserted into the BspEI/BamHI and BbsI/ApaIsites of pVT2, respectively. This generated clonepJR1V13 which encodes 1V13 (gan: AY860434) inwhich PTH61 is fused to the N-terminus and PTH22 tothe C-terminus of D17E/W34A. Two control clones,pJR1V12 (gan: AY860433) and pJR1V14 (gan:AY860435), were also constructed by fusing PTH22 orPTH61 to only one end of VTB (D17E/W34A) (Fig. 1B).

Nine, 1 and 6 mg of fusion proteins 1V12, 1V13 and1V14 were purified from 1 l cultures of E. coli cellsharboring pJR1V12, pJR1V13 and pJR1V14. Theirpurity and pentameric status were examined by sizeexclusion chromatography on a Sephadex™ 200column. As indicated in Fig. 2, 1V12 and 1V14 formpure pentamer with no aggregation and degradation.

The major 1V13 peak (Fig. 2) indicates that 1V13 formspentamer. However, the homogeneity of the 1V13pentamer is less than satisfactory. One aggregation peak(at an elution volume of 4 ml, corresponding to amolecular mass of approximately 2000 kDa) and otherundesired peaks were observed (Fig. 2).

The N-pentabody 1V12 and the C-pentabody 1V14showed significantly stronger binding to PTH2–strepta-vidin (Fig. 3D and E) than their monomeric counterpartPTH61 (Fig. 3C) and PTH22 (Fig. 3B). This clearlyindicated that pentamerization of sdAbs is an easy way toimprove the avidity of sdAbs with moderate affinities.

3.3. Improvement of the physical properties of 1V13

1V13 consists of eight blocking units: PTH61, linker 1,VTB (D17E/W34A), linker 2, PTH22, linker 3, the c-Mycdetection tag and the 5×histidine purification tag (Fig. 1B).While sdAbs and theVTB subunit are known to have goodstability properties, the linkers between these decabodybuilding blocks may significantly affect the physicalproperties of the decabody molecule. Accordingly,different linker sequences were investigated in attempt toobtain a decabody sequence that gives a homogeneouspentameric molecule.

Four linkers, Linker SHORT (GGGGS), LinkerLONG (GGGGSGGGGS), Linker LLAMA (AHH-SEDPSSK) (van der-Linden et al., 2000) and LinkerHUMAN (KRVAPELLGGPS) (Guddat et al., 1993) werechosen to replace linkers L1 and L2 of 1V13 (Fig. 1B).These linkers were chosen for their general applicabilityand the flexibility they confer on the protein domains theyconnect. Insertion of each of the above linker sequences ateither or both of the 1V13 L1 and L2 positions gives 16possible decabody constructs. In practice 15 of theseconstructs were generated.

Small cultures (3 ml) of the 15 clones were analyzedby Western blotting for the production of the proteins(data not shown). Four clones with relatively highproduction levels, JR9 (gan: AY860436), JR10 (gan:AY869437), JR13 (gan: AY860438), and JR14 (gan:AY860439), were chosen for protein purification from200 ml cultures. Approximately 1 mg of pure proteinwas obtained from 200 ml cultures of each of the fourclones, corresponding to a yield of about 5 mg/l of E.coli culture.

The purified proteins JR9, JR10, JR13 and JR14 wereanalyzed by size exclusion chromatography. Significantimprovements in the physical properties of all fourproteins were observed, relative to 1V13 (Fig. 2). Theaggregation peaks (at an elution volume of 4 ml) weresignificantly lower and the decabody peaks were more

93E. Stone et al. / Journal of Immunological Methods 318 (2007) 88–94

homogeneous than that of 1V13. In the instance of JR9,no obvious aggregation or degradation was observed.

Surface plasmon resonance analysis of JR9 showedthat it had strong binding to unconjugated PTH2 (Fig. 3F).This is different from 1V12 and 1V14 which bind onlyvery weakly to directly immobilized peptide (data notshown). In addition the interaction was characterized by arelatively slow dissociation.

4. Discussion

We demonstrated that it is feasible to make noveldecavalent bispecific antibody reagents based on sdAbsand VTB, joined by suitable linker sequences, in spite ofthe complexity of such molecules. Optimization of linkersequences produced a molecule, JR9, which was homo-geneous, did not aggregate and expressed well in E. coli.Based on a previously described pentabody model (Zhanget al., 2004) and results presented here, it is assumed thatpJR9 has free access to ten antigen molecules.

A unique feature of the bsAb molecule describedhere is its high valency. Five sdAbs are displayed oneach side of the doughnut-shaped VTB. Pentavalentdisplay transforms low affinity sdAbs into high aviditymolecules that bind effectively surface antigens presentat sufficient density for high avidity binding.

The pentavalency of this bsAb model will beadvantageous in tumor targeting. Most tumor associatedantigens are also present in normal tissues but in lowerdensity. Targeting tumor tissues with high valency mol-ecules in which the monovalent antigen bindingcomponent has low affinitywill provide selective bindingof the reagents to tumor sites in instances where antigendensities on normal tissues are below the threshold levelrequired for multivalent binding.

Immunogenicity is one of the major considerations indeveloping therapeutic proteins including antibodies. Infact, overcoming a human anti-mouse antibody responsewas a major hurdle in the development of monoclonalantibodies as therapeutic drugs. Immunogenicity alsoneeds to be addressed when designing bsAbs. Thedecabodies described here contain two protein domains,sdAbs and VTB. It has been demonstrated that sdAbsfrom camelids are non-immunogenic in mice (Cortez-Retamozo et al., 2004). Therefore, the use of camelidsdAbs should not pose a problem in terms of immuno-genicity. With the recent development of fully humansdAbs (Jespers et al., 2004; To et al., 2005), decabodiescould be constructed with human sdAbs.

It is known that VTB is immunogeneric and evenprotective when immunized with adjuvants (Marcatoet al., 2001). Yet it was found difficult to raise neu-

tralizing antibodies against VTB when lipopolysaccha-ride was not included (Marcato et al., 2005). Theimmunogenicity of decabodies which contains VTBremains to be evaluated.

Acknowledgements

We thank Dr. G. Willick for providing the antigenPTH2 and PTH2–streptavidin, Sara Hahn for technicalassistance and Dr. Andrea Bell for proofreading themanuscript.

References

Arndt, M.A., Krauss, J., Kipriyanov, S.M., Pfreundschuh, M., et al.,1999. A bispecific diabody that mediates natural killer cellcytotoxicity against xenotransplantated human Hodgkin's tumors.Blood 94, 2562.

Bonardi, M.A., French, R.R., Amlot, P., Gromo, G., et al., 1993.Delivery of saporin to human B-cell lymphoma using bispecificantibody: targeting via CD22 but not CD19, CD37, or immuno-globulin results in efficient killing. Cancer Res. 53, 3015.

Cao, Y., Lam, L., 2003. Bispecific antibody conjugates in therapeutics.Adv. Drug Deliv. Rev. 55, 171.

Cortez-Retamozo, V., Backmann, N., Senter, P.D., Wernery, U., et al.,2004. Efficient cancer therapy with a nanobody-based conjugate.Cancer Res. 64, 2853.

Cotton, R.G., Milstein, C., 1973. Letter: fusion of two immunoglob-ulin-producing myeloma cells. Nature 244, 42.

De Jonge, J., Heirman, C., de Veerman, M., Van Meirvenne, S., et al.,1998. In vivo retargeting of T cell effector function by recombinantbispecific single chain Fv (anti-CD3×anti-idiotype) induces long-term survival in the murine BCL1 lymphoma model. J. Immunol.161, 1454.

de Kruif, J., Logtenberg, T., 1996. Leucine zipper dimerized bivalentand bispecific scFv antibodies from a semi-synthetic antibodyphage display library. J. Biol. Chem. 271, 7630.

Deyev, S.M., Waibel, R., Lebedenko, E.N., Schubiger, A.P., et al.,2003. Design of multivalent complexes using the barnase⁎barstarmodule. Nat. Biotechnol. 21, 1486.

Ferrini, S., Prigione, I., Miotti, S., Ciccone, E., et al., 1991. Bispecificmonoclonal antibodies directed to CD16 and to a tumor-associatedantigen induce target-cell lysis by resting NK cells and by a subsetof NK clones. Int. J. Cancer 48, 227.

Ford, C.H., Osborne, P.A., Rego, B.G., Mathew, A., 2001. Bispecificantibody targeting of doxorubicin to carcinoembryonic antigen-expressing colon cancer cell lines in vitro and in vivo. Int. J.Cancer 92, 851.

Gestin, J.F., Loussouarn, A., Bardies, M., Gautherot, E., et al., 2001.Two-step targeting of xenografted colon carcinoma using abispecific antibody and 188Re-labeled bivalent hapten: biodis-tribution and dosimetry studies. J. Nucl. Med. 42, 146.

Graziano, R.F., Guptill, P., 2004. Chemical production of bispecificantibodies. Methods Mol. Biol. 283, 71.

Guddat, L.W., Herron, J.N., Edmundson, A.B., 1993. Three-dimensional structure of a human immunoglobulin with a hingedeletion. Proc. Natl. Acad. Sci. U. S. A 90, 4271.

Heijnen, I.A., Rijks, L.J., Schiel, A., Stockmeyer, B., et al., 1997.Generation of HER2/neu-specific cytotoxic neutrophils in vivo:efficient arming of neutrophils by combined administration of

94 E. Stone et al. / Journal of Immunological Methods 318 (2007) 88–94

granulocyte colony-stimulating factor and Fcgamma receptor Ibispecific antibodies. J. Immunol. 159, 5629.

Jespers, L., Schon, O., Famm, K., Winter, G., 2004. Aggregation-resistant domain antibodies selected on phage by heat denatur-ation. Nat. Biotechnol. 22, 1161.

Lu, D., Jimenez, X., Zhang, H., Atkins, A., et al., 2003. Di-diabody: anovel tetravalent bispecific antibody molecule by design. J. Immunol.Methods 279, 219.

Marcato, P., Mulvey, G., Read, R.J., Vander, H.K., et al., 2001.Immunoprophylactic potential of cloned Shiga toxin 2 B subunit.J. Infect. Dis. 183, 435.

Marcato, P., Griener, T.P., Mulvey, G.L., Armstrong, G.D., 2005.Recombinant Shiga toxin B-subunit-keyhole limpet hemocyaninconjugate vaccine protects mice from Shigatoxemia. Infect.Immun. 73, 6523.

Mirochnik, Y., Rubenstein, M., Guinan, P., 2004. Two constructedantibody-derived delivery vehicles for targeting oligodeoxynu-cleotides to prostate tumors expressing prostate specific antigen.Drug Deliv. 11, 161.

Nisonoff, A., Rivers, M.M., 1961. Recombination of a mixture ofunivalent antibody fragments of different specificity. Arch.Biochem. Biophys. 93, 460.

Rheinnecker, M., Hardt, C., Ilag, L.L., Kufer, P., et al., 1996.Multivalent antibody fragments with high functional affinity for atumor-associated carbohydrate antigen. J. Immunol. 157, 2989.

Ridgway, J.B., Presta, L.G., Carter, P., 1996. ‘Knobs-into-holes’engineering of antibody CH3 domains for heavy chain hetero-dimerization. Protein Eng. 9, 617.

Shen, J., Vil, M.D., Jimenez, X., Iacolina, M., et al., 2006. Singlevariable domain-IgG fusion: a novel recombinant approach to Fcdomain-containing bispecific antibodies. J. Biol. Chem. 281,10706.

Soltyk, A.M., MacKenzie, C.R., Wolski, V.M., Hirama, T., et al., 2002.A mutational analysis of the globotriaosylceramide-binding sitesof verotoxin VT1. J. Biol. Chem. 277, 5351.

Somasundaram, C., Arch, R., Matzku, S., Zoller, M., 1996.Development of a bispecific F(ab′)2 conjugate against thecomplement receptor CR3 of macrophages and a variant CD44antigen of rat pancreatic adenocarcinoma for redirecting macro-phage-mediated tumor cytotoxicity. Cancer Immunol. Immun-other. 42, 343.

To, R., Hirama, T., Arbabi-Ghahroudi, M., MacKenzie, R., et al., 2005.Isolation of monomeric human V(H)s by a phage selection. J. Biol.Chem. 280, 41395.

van der-Linden, R., de Geus, B., Stok, W., Bos, W., et al., 2000.Induction of immune responses and molecular cloning of the heavychain antibody repertoire of Lama glama. J. Immunol. Methods240, 185.

Whitfield, J.F., 2006. Osteoporosis-treating parathyroid hormonepeptides: what are they? What do they do? How might they doit? Curr. Opin. Investig. Drugs 7, 349.

Zhang, J., Tanha, J., Hirama, T., Khieu, N.H., et al., 2004.Pentamerization of single-domain antibodies from phage libraries:a novel strategy for the rapid generation of high-avidity antibodyreagents. J. Mol. Biol. 335, 49.