Trispecific Antibody Derivatives Production of Recombinant

14352_27050_full.pdf Á Á Á Á -.
-×
-/÷  -×€"ׁ

;C×O;P[i×
i×;[×yDPH[k×HkHh]FPZHhPw;kP][×.D;x]YF×M]h×kOH×*h]FlDkP][×]M

-HD]ZCP[;[k×

PieHDP{D×;[F×/hPieHDP{D×
[kPC]Ft×

HhPq;kPqHi

À

6IMRLMPHI7GLSSRNERWIXEP

.-QQYRSP  •

LXXTWHSMSVKNMQQYRSP

Downloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

Fab Chains As an Ef®cient Heterodimerization Scaffold for the Production of Recombinant Bispeci®c and Trispeci®c

Antibody Derivatives

1 Reinhilde Schoonjans, An Willems, Steve Schoonooghe, Walter Fiers, Johan Grooten, 2 and

Nico Mertens

Due to their multispeci®city and versatility, bispeci®c Abs (BsAbs) are promising therapeutic tools in tomorrow's medicine.

Especially intermediate-sized BsAbs that combine body retention with tissue penetration are valuable for therapy but necessitate

expression systems that favor heterodimerization of the binding sites for large-scale application. To identify heterodimerization

domains to which single-chain variable fragments (scFv) can be fused, we compared the ef®ciency of heterodimerization of CL and

CH1 constant domains with complete L and Fd chains in mammalian cells. We found that the isolated CL:CH1 domain interaction

was inef®cient for secretion of heterodimers. However, when the complete L and Fd chains were used, secretion of L:Fd het-

erodimers was highly successful. Because these Fab chains contribute a binding moiety, C-terminal fusion of a scFv molecule to

the L and/or Fd chains generated BsAbs or trispeci®c Abs (TsAbs) of intermediate size (75±100 kDa). These disul®de-stabilized

bispeci®c Fab-scFv (ªbibodyº) and trispeci®c Fab-(scFv) 2 (ªtribodyº) heterodimers represent up to 90% of all secreted Ab

fragments in the mammalian expression system and possess fully functional binding moieties. Furthermore, both molecules recruit

and activate T cells in a tumor cell-dependent way, whereby the trispeci®c derivative can exert this activity to two different tumor

cells. Thus we propose the use of the disul®de-stabilized L:Fd heterodimer as an ef®cient platform for production of intermediate-

sized BsAbs and TsAbs in mammalian expression systems.The Journal of Immunology,2000, 165: 7050±7057.

B ispeci®c Abs (BsAbs) 3 are versatile tools in the devel- opment of new experimental therapies of various dis- eases. Typically, one part of the BsAb speci®cally rec- ognizes a target molecule or cell (e.g., a cancer cell), whereas the other part is directed to an enzyme, toxin, virus, or immune effec- tor cell. The production of BsAbs is usually achieved by chemical cross-linking of Fab (1) or by the hybrid hybridoma technique (2). The requirement for extensive postproduction puri®cation steps to isolate the bispeci®c component from other Ig-derived by-products has driven the application of recombinant DNA methodology and Ab engineering techniques for production of BsAbs in bacterial expression systems. Bispeci®c single-chain variable fragment (scFv) 2 heterodimers have been produced by direct genetic fusion (3), or indirectly by fusion to helical heterodimerization domains

(4±6). To avoid the rapid blood and whole-body clearance exhib-ited by these small-sized molecules (7, 8), strategies have been

developed to produce intermediate- and larger-sized recombinant BsAb. Ef®cient heterodimerization of complete heavy chains was accomplished by engineering complementarity into the domain in- terfaces of two CH3 molecules (9). For optimal results, also a disul®de bridge had to be introduced into the molecule to stabilize the heterodimer (10). Also single Ab domains, such as CL (11) or CH3 (12), have been applied for homodimerization, generating bivalent minibodies. These intermediate-sized molecules avoid clearance in the kidney, whereas they still have a more ef®cient tissue penetration than complete Abs, which makes them better suited for in vivo therapeutic application (12). To transform these homodimeric bivalent molecules into heterodimeric BsAbs, spon- taneous interaction between the complementary single domains CL and its natural partner CH1 was applied (13). This approach generated a stabilized heterodimer due to the presence of a natural disul®de bridge and gave rise to 63% of heterodimerization when both domains were coexpressed inEscherichia coli(13). In contrast toE. coli, mammalian cells possess extensive control mechanisms that prevent incorrectly folded proteins to proceed along the secretory pathway (14). Consequently, heterodimeriza- tion of CL and CH1 domains might be further enhanced in mam- malian cells. To verify this possibility, we analyzed the ef®ciency of heterodimerization of CL and CH1 domains coexpressed in mammalian cells. The results show that additional VL:V H contri- bution to CL:CH1-mediated heterodimerization greatly enhances ef®cient heterodimerization and secretion. Fusion of a scFv mol- ecule to one or both of the VL-CL (L) and V H -CH1 (Fd) chains resulted in nearly exclusive secretion of heterodimeric fusion mol- ecules, viz BsAbs or trispeci®c Abs (TsAbs) of intermediate size with fully functional bi- or trispeci®c activities. These results put forward the L:Fd interaction as an ef®cient heterodimerization Molecular Immunology Unit, Department of Molecular Biology, Flanders Interuni- versity Institute for Biotechnology, Ghent University, Ghent, Belgium Received for publication October 15, 1999. Accepted for publication September

26, 2000.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby markedadvertisementin accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 This work was supported by the Inter-University Network for Fundamental Re- search (Belgium) Program. R.S. was recipient of a fellowship from Vlaams Instituut voor de Bevordering van Het Wetenshappelijk Technologisch Onderzoek in de In- dustie (IWT) and Vlaamse Liga tegen Kanker. A.W. is a research assistant with the Fonds Wetenschappelijk Onderzoek-Vlaanderen and S.S. with IWT. 2 Address correspondence and reprint requests to Dr. J. Grooten, Department of Mo- lecular Biology, K.L. Ledeganckstraat 35, B-9000 Ghent, Belgium. E-mail address:

Johan.Grooten@dmb.rug.ac.be

3 Abbreviations used in this paper: BsAb, bispeci®c Ab;hPLAP, human placental alkaline phosphatase; scFv, single-chain variable fragment; TsAb, trispeci®c Ab;

BCL, B cell lymphoma.

Copyright © 2000 by The American Association of Immunologists0022-1767/00/$02.00Downloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

scaffold for the generation of multifunctional Ab derivatives in mammalian cells.

Materials and Methods

Cell lines

HEK293T, a human embryonic kidney cell line transfected with

SV40 large T-Ag (SV40T

tsA1609 ) (15), was used for transient eu- karyotic expression. TE2 cells are murine and CD3 1

Th1-type

cells are T cells (16). MO4I4 cells are C3H mouse-derived MO4 ®brosarcoma cells transfected with the human placental alkaline phosphatase (hPLAP) gene (17, 18). BALB/c-derived myeloma

BCL1 expresses surface IgM

l(19).

Plasmids and gene assembly

Restriction enzymes, DNA modifying enzymes, and DNA poly- merase were used as recommended by the manufacturers. DNA ampli®cation was performed with Vent-DNA polymerase (New England Biolabs, Beverly, MA). E6 and 2C11 denote the genes or gene fragments of an anti-hPLAP and an anti-murine TCR-asso- ciated CD3 e-chain (anti-CD3) mAb, respectively. Expression plas- mids were constructed in pCAGGS (20). Cloning of the L chain of

E6 mAb anti-hPLAP (IgG2b/

k) in the vector pSV51E6L has been described previously (21). The E6Fd fragment encodes V H , CH1, and the ®rst ®ve amino acids EPSGP of the upper hinge region. Gene assembly was conducted by introduction of suitable restric- tion sites using modifying PCR primers. All PCR-derived frag- ments were sequence veri®ed after cloning. Fusions with E6CH1 or E6CL include the ªelbowº regions (EMKRAD and SAAKTT from the L and Fd chains, respectively) of Fab chains. The isolated coding sequence for the CL domain was fused to the signal se- quenceoftheE6heavychain.scFv(anti-BCL1)and(G4S)3-scFv(anti- CD3)-(His)6 were ampli®ed from pQE-bssFvB1±2C11 (22) and genetically fused to the C terminus of CL or CH1 via a DVPSGPG or (G4S)3 linker, respectively.

Production of Ab fragments

For transient expression, HEK293T cells were transfected accord- ing to the Ca 3 (PO 4 ) 2 precipitation method (23). Twenty hours be- fore transfection, HEK293T cells were seeded at 4310 6 cells/175 cm 2 . Fourteen micrograms DNA of each expression plasmid was added to the cells for 24 h, after which the cells were covered with supplemented DMEM containing 5 mg/L bovine insulin, 5 mg/L transferrin, and 5 mg/L selenium (ITS) replacing FCS. Medium was harvested every 48 h after transfection. Gel ®ltration was per- formed on an XK 16/88 Superdex 200 column (Amersham Phar- macia Biotech, Piscataway, NJ) calibrated with a commercial pro- tein standard mix (Bio-Rad, Richmond, CA). A sample volume of

1 ml was loaded and the column was developed in 15 mM

NaH 2 PO 4 , 150 mM NaCl, pH 7.5, at 1 ml/min. The concentration of the produced recombinant Ab fragments was determined on Western blot; samples were applied in three different dilutions and compared with a serial dilution of a standard of puri®ed TsAb of known concentration. Western blotting, immunodetection, and densitometry Medium fractions (503concentrated) of transfected cells, corre- sponding to 1 ml supernatant, were diluted with nonreducing SDS sample buffer, boiled for 5 min, fractionated by 10% SDS-PAGE, and blotted to a nitrocellulose membrane. Subsequent functional detection of anti-hPLAP activity was achieved directly by incuba- tion of the membrane with hPLAP (Sigma, St. Louis, MO). Im- munodetection of the proteins on blot was as described previously (21); goat anti-mouse Ig gkserum (Sera-Lab, Crawley Down, Sus-

sex, U.K), anti-E-tag (Amersham Pharmacia Biotech, Rainham,U.K.), anti-goat IgG serum conjugated to alkaline phosphatase,

and anti-mouse IgG1 conjugated to alkaline phosphatase were used. For densitometric measurements, blots containing immuno- reactive signals were scanned and analyzed with ImageMaster

VDS software (Amersham Pharmacia Biotech).

Cellular binding

Flow cytometry was performed by concentrating cell culture su- pernatant containing Fab-scFv BsAb or Fab-(scFv) 2

TsAb and di-

alyzing to PBS, after which it was supplemented with 0.5% BSA and 0.02% azide. Washed cells (2310 5 ) were then resuspended in 100 ml of concentrated and dialysed BsAb, TsAb, E6 mAb, or

B1 mAb (5

mg/ml), after which cells were incubated for 30 min at

4°C. Following three wash procedures, the cells were incubated for

30 min with 1

mg/ml of ¯uorescein-conjugated goat anti-mouse Fab (Organon Teknika, Durham, NC). After a ®nal wash proce- dure, cells were analyzed by FACS. Binding of 145-2C11 mAb was detected directly with an anti-hamster FITC-coupled anti- serum (Sera-Lab). Cellular binding (cellular ELISA) was performed by preincu- bating 10 6 TE2 cells with 1mg Fc-Block (PharMingen, San Diego,

CA) for 30 min. Two micrograms mAb E6 or 2

mg concentrated and dialyzed Fab-(scFv) 2

TsAb were added to 2310

5 washed

TE2 cells in 100

ml PBS, supplemented with 0.5% BSA and 0.02% azide, and were incubated for 30 min at 4°C. After three wash procedures with supplemented PBS, the cells were incubated for

30 min with 5 U/mlhPLAP(Sigma). After a ®nal wash procedure,

cells were resuspended in 600 ml 10% DEAE withp-nitrophenyl phosphate. Two hundred microliters of each sample was trans- ferred in triplicate to a multiwell plate, after which A 405
was measured. Absorbing-out of Ab with cells was performed by incubating three times with 6310 5 target cells (MO4I4, BCL1, or TE2) or irrelevant cells (SP2/0). The nonbound fraction was analyzed on Western blot and revealed with anti-mouse IgG serum.

Surface plasmon resonance

Af®nity analysis was performed using a BIAcore 2000 (BIAcore, Uppsala, Sweden). Fab were prepared by papain digestion and subsequent protein A chromatography (Pierce, Rockford, IL). His- tagged TsAb was puri®ed from serum with immobilized metal af®nity chromatography using cobalt as a ligand. The TsAb as well as hPLAP (Sigma) were further puri®ed on phenyl Sepharose and

Q-Sepharose columns (Amersham Pharmacia Biotech).

T cell proliferation assay

Splenocytes from syngenic C3H/He mice and BALB/c mice were used for MO4I4 ®brosarcoma cells and BCL1 lymphoma cells, respectively. All mice were purchased from Charles River Breed- ing Laboratories (Sulzfeld, Germany). MO4I4 and BCL1 tumor cells were pretreated with 50 mg/ml mitomycin C at 37°C in the dark for 12 h or 90 min, respectively. After removal of mitomycin

C, 5310

4 treated cells were cocultured with 1310 5 splenocytes in a round-bottom well in the presence of 0.5 mg/ml of bispeci®c

Fab-scFv or trispeci®c Fab-(scFv)

2 molecule. After 48 h, the cul- ture was pulsed with 0.5 mCi of [ 3

H]thymidine (1 mCi/ml). After

18 h the cells were disrupted by freeze-thawing; the DNA was

spotted on a ®lter and washed. The incorporated radioactivity was measured by scintillation counting (Top-Count; Packard, Meriden, CT). All experiments were performed in triplicate. 51

Cr release assay

Redirecting cytotoxicity from CTL responses was assayed using a standard 51

Cr release assay with syngenic CTL cells that were7051The Journal of ImmunologyDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

alloreactively primed. Brie¯y, 4310 6 splenic responder cells (C3H/HeOUico) were mixed with 4310 6 splenic stimulator cells (C57BL/6) treated with 50 mg/ml mitomycin C for 60 min at 37°C in the dark. The mixed cell population was cocultured in 2-ml cultures in supplemented RPMI 1640 in the presence of 30 U/ml of murine IL-2. These cultures were incubated at 37°C in 7% CO 2 in humidi®ed air for 5 days.

MO4I4 cells were incubated with 150

mCi Na 51
CrO 4 (Amer- sham Pharmacia Biotech) for 90 min at 37°C and washed carefully to minimize spontaneous release. Effector cells from the mixed lymphocyte culture were harvested and washed; 2.5310 5 cells were plated in triplicate in 96-well U-bottom plates containing 53 10 3 tumor cells (E:T ratio 50:1) and Fab-scFv BsAb (1mg/ml) in a total volume of 200 ml. After a 4-h incubation at 37°C, 30mlof the culture supernatant was transferred to a Lumaplate (Packard), air dried, and counted. The percentage of speci®c lysis was cal- culated as 1003(experimental release)2(spontaneous release)/ (maximum release)2(spontaneous release). Maximum release was the value obtained from target cells incubated with 2% SDS. Spontaneous release never exceeded 14% of the maximum release.

Results

CL:CH1 heterodimerization is inef®cient in mammalian cells unless enlarged with variable domains To assess the eukaryotic secretion of homo- and heterodimers from individual domains of Ab L and Fd chains, HEK293T cells were transiently (co)transfected with pCAGGS expression vectors con- taining as an insert the isolated CL or CH1 domain. These domains are derived from mouse Ab E6 (IgG2b, k) (24) speci®c for hPLAP (25). However, no heterodimeric product could be detected, even not if for the purpose of more sensitive detection the CH1 domain was modi®ed with an E-tag.

To assess whether the presence of either the V

H or the VL do- mains is required for progression of these Ab derivatives through the endoplasmic reticulum, the CL and CH1 domains were coex- pressed with their corresponding extended counterparts, namely, the complete Fd chain and the native L chain, respectively (i.e., CL:VHCH1 and VLCL:CH1). Also here, no secreted het- erodimers, either CL:Fd or L:CH1, could be detected. Only L monomers and L:L homodimers were demonstrated in culture ¯u- ids of L gene-(co)transfected HEK293T cells. However, coexpres- sion of CL and CH1, both enlarged with their corresponding vari- able domains (in fact representing L and Fd chains) generated ef®cient expression of L:Fd heterodimers (Fab). The Fd chain on its own was never detectable, neither as a monomer nor as a ho- modimer. Thus the Fd chain can only be secreted in the form of a heterodimer with the L chain, whereas the L chain preferentially forms heterodimers with the Fd chain upon coexpression. These results identify the L and Fd chains as a minimal con®guration for obtaining ef®cient heterodimerization and secretion of CL:CH1- containing Ab derivatives in mammalian cells. Fab-constituting L and Fd chains mediate ef®cient heterodimerization of scFv molecules Because enlarging the CL and CH1 domains with VL and V H signi®cantly increases the ef®ciency of heterodimerization and se- cretion, we wanted to establish whether the corresponding L and Fd chains could be used as a heterodimerization scaffold to pro- duce BsAbs and even TsAbs. Because the L:Fd heterodimer itself constitutes a functional binding site, a C-terminal elongation of the L and Fd chain with a peptide linker and scFv molecule(s) would generate BsAb or TsAb with broad action radius. To create a

model molecule, a scFv speci®c for the murine myeloma BCL1 Id(22) was fused to the C terminus of the L chain with a six-amino-

acid linker, whereas the anti-murine CD3 escFv (22) was fused to the C terminus of the Fd chain with a (Gly 4 Ser) 3 linker. Coexpression of the fusion genes (L-scFv and Fd-scFv) with the native complementary chain (L or Fd) is then expected to yield L:Fd-scFv and L-scFv:Fd heterodimers as schematically repre- sented in Fig. 1,AandB. Western blot analysis of the culture supernatant of HEK cells, cotransfected as described above, showed secretion of the expected heterodimers (Fig. 1,DandE). The secreted Ab products consist predominantly of the L-scFv:Fd and L:Fd-scFv heterodimers, along with minor bands representing

L-scFv or native L chain monomers and homodimers.

Similarly, coexpression of both enlarged L-scFv and Fd-scFv chains in HEK293T cells yielded a dominant L-scFv:Fd-scFv het- erodimer. This was concluded based on the observed and expected

100 kDa of the dominant protein band (Fig. 1,CandF). Staining

of the blotted proteins with hPLAP to reveal the speci®city of the Fab portion (which is only formed in the heterodimer) con®rmed the position of BsAbs and TsAb (Fig. 1,D±F). The presence of a stabilizing disul®de bridge in Fab-scFv and Fab-(scFv) 2 het- erodimers was con®rmed by SDS-PAGE under reducing condi- tions (results not shown). The relative amount of the secreted products was estimated by densitometric scanning of immunoreactive signals on a Western blot developed with anti-murine IgG gkserum at different sample dilutions. Up to 90% of the secreted Ig-derived proteins was in the correct heterodimeric format when either Fd-scFv( a-CD3) was heterodimerized with the native L chain or L-scFv( a-BCL1) was heterodimerized with the Fd chain or with the L-scFv fusion prod- uct. Sometimes, an excess of L chain-derived proteins was ob- served, which were either in a monomeric form or appeared as a disul®de-stabilized dimer. However, Fd chain derivatives always appeared as a heterodimer. Expression levels of the heterodimeric products were estimated by calibrated immunostaining to be 3 mg/ ml/24 h for both the Fab-scFv and the Fab-(scFv) 2 molecules. FIGURE 1.Heterodimerization of scFv molecules by L and Fd chains. Schematic representation of a BsAb created by extension of the C terminus of the L chain (A) or the Fd chain (B), or of a TsAb by extension of both the L and Fd chains with a scFv (C). Secretion of L:Fd heterodimers by cotransfected HEK293T cells was assayed by Western blotting and immu- nodetection with anti-murine IgG gkanti-serum or with hPLAP.D,L chain elongated with scFv.E, Fd chain elongated with scFv.F, elongation of both L and Fd chains. Darker ellipses, CL-containing fusion molecules; brighter ellipses, CH1-containing fusion molecules. Arrowheads show re- spective positions on the blot. M, Molecular mass markers (kDa).

7052 Fab-scFv BISPECIFIC AND TRISPECIFIC ANTIBODIESDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

We conclude that C-terminal extension of the L and Fd chains with the various scFv does not hamper heterodimerization or se- cretion of the L:Fd heterodimer. Thus, the heterodimerization scaf- fold constitutes an instrument for ef®cient generation of disul®de- stabilized BsAbs and TsAbs of intermediate size (75 and 100 kDa, respectively) in mammalian cells. Binding characteristics of Fab-scFv and Fab-(scFv) 2 molecules To verify the functionality of the individual binding moieties of the anti-tumor cell/anti-T cell Fab-scFv and Fab-(scFv) 2 molecules, their binding characteristics on Ag-positive cells were determined

by ¯ow cytometry and compared with the binding characteristicsof the mAbs from which the respective binding moieties were

derived. As shown in Fig. 2, the (anti-hPLAP3anti-CD3) Fab- scFv molecule recognizes bothhPLAP-transfected MO4I4 and CD3 1 TE2 cells. Also the (anti-hPLAP3anti-CD33anti-BCL1)

Fab-(scFv)

2 derivative combines the binding characteristics of its donor mAbs and recognizes its Ags on MO4I4, TE2, and BCL1 1 B cells. Thus both types of Ab derivatives retained the function- ality and speci®city of the constituting Fab and scFv components. To verify whether the different binding moieties of the bispeci®c

Fab-scFv and trispeci®c Fab-(scFv)

2 are capable of concurrent binding to their respective Ags, we immobilized the BsAb and TsAb by binding to one of its Ags, the CD3 ligand on T cells, and detected this binding by the subsequent retention of the hPLAP Ag (Fig. 3A). In the TsAb, two more axes of bispeci®c binding can be determined, i.e., anti-hPLAP3anti-BCL1 and anti-CD33anti- BCL1. These concurrent bindings were assayed by detection of the

TsAb bound to hPLAP

1

MO4I4 cells or CD3

1

TE2 cells with a

biotinylated BCL1 Ag (which is a mouse IgM molecule) (Fig. 3B). All three approaches demonstrated multiple bispeci®c binding along the three different axes of the TsAb. Fab-scFv fusion molecules mediate tumor cell-dependent activation of T cells The capacity of the (anti-hPLAP3anti-CD3) BsAb and the (anti- BCL13anti-hPLAP3anti-CD3) TsAb to recruit and activate CD3 1 T cells by its anti-T cell and anti-tumor cell reactivity was examined on the basis of a tumor cell-dependent induction of T cell proliferation and T cell-mediated cytotoxicity. Due to its dual anti-tumor cell binding moieties, the TsAb was assayed in two experimental settings, one primed with the hPLAP 1

MO4I4 tumor

cells, and the other primed with the BCL1 tumor cells. As shown in Fig. 4,AandC, induction of both proliferative and cytotoxic T cell activities was apparent only in the presence of tumor cells, T cells, and BsAb or TsAb, but not in the absence of either of these components. Clearly, generation of T cell reactivity was dependent on tumor cell-induced cross-linking of the monovalent a-CD3 moiety of the BsAb or TsAb resulting from the interaction of the anti-tumor cell moiety with its corresponding tumor-associated Ag. The level of T cell reactivity depended on the amount of Ab added; the optimal concentration was determined to be between

0.1 and 1

mg/ml for both the BsAb and the TsAb (Fig. 4,BandD). From these experiments we conclude that no conformational restrictions impair the concurrent cognitive interaction of the two binding moieties of the BsAb, and especially of the three binding moieties of the TsAb, to cross-link two different cells. As a con- sequence, these novel types of recombinant Abs exert activities typical of BsAbs, namely, recruitment and tumor cell-dependent FIGURE 2.Binding characteristics of Fab-scFv and Fab-(scFv) 2 mole- cules reveal dual and triple Ag speci®city. Binding of Fab-scFv, Fab- (scFv) 2 , and parental mAbs was assessed by indirect immuno¯uorescence and ¯ow cytometry analysis (thick lines). Except for mAb 145-2C11, which was detected with FITC-conjugated anti-hamster Ig Ab, FITC-con- jugated anti-mouse Fab was used for detection. Negative controls were stained with FITC-labeled secondary Ab alone (thin lines).

FIGURE 3.Fab-(scFv)

2 leads to bispeci®c binding along its three axes.A, Dual binding of the Fab-scFv BsAb (B) and the Fab-(scFv) 2

TsAb (T) along the (anti-CD33

anti-hPLAP) axis was demonstrated by detection of TE2 cell-bound BsAb or TsAb by cellular ELISA with hPLAP and a colorigenic hPLAP substrate. In negative controls (open columns), no Ab or cells were added or TsAb was replaced with the monospeci®c anti-hPLAP E6 mAb.B, Dual binding activity of the TsAb (anti-BCL13anti- hPLAP3anti-CD3) along the (anti-hPLAP3anti-BCL1) and (anti-CD33anti-BCL1) axes was demonstrated by indirect immuno¯uorescence. Binding of TsAb to hPLAP 1

MO4I4 or CD3

1

TE2 cells was detected using biotinylated

BCL1 IgM mAb as secondary Ab and FITC-labeled avidin (®lled curves). Negative controls were stained with second- ary Ab and FITC-labeled avidin (open curves).

7053The Journal of ImmunologyDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

activation of effector T cells. The TsAb can do this with two different Ags. Stability of L:Fd heterodimerized bi- and tribodies To be useful in therapy settings, recombinant Abs need to be suf- ®ciently stable, and the recombinant model should not impair the functionality of the Ab. To check the TsAb for its possible ten- dency to form dimers or higher aggregates, we puri®ed a His- tagged TsAb on a metal chelating column and subsequently ap- plied the concentrated material on a size exclusion column (Fig.

5A). More than 95% eluted as a monomer at 100 kDa. A very small

fraction of the TsAb was found as a dimer, but the material eluting in the void volume did not contain any TsAb (as determined by Western blotting and immunodetection, data not shown). To determine the fraction of puri®ed material that is active, the TsAb was absorbed out by immunoprecipitation with target cells. The unabsorbed fraction was detected by Western blotting and quanti®ed by comparison to a series of 2-fold dilutions of the un- absorbed sample (Fig. 5B). Capture of the TsAb by MO4I4 or BCL1 cells resulted in a 100-fold reduction of the amount of un-

bound TsAb. Similar results were obtained when the correspond-ing anti-hPLAP and anti-BCL1 mAb were immunoprecipitated.

Compared with SP2/0 cells (used as a negative control), incubation with TE2 cells (CD3 1 ) did not result in signi®cant capture of the TsAb. However, the anti-CD3 mAb 145.2C11 was not captured effectively by the TE2 cells, likely as a consequence of the low CD3 expression levels on these T cells. In conclusion, this exper- iment shows that the large majority of the L-scFv:Fd-scFv het- erodimers are present in a correct conformation. We also determined the stability of the BsAb and the TsAb after incubation at 37°C in PBS or in freshly prepared mouse serum (Fig. 5C). The stability was measured as the remaining bispeci®c activity in the Ab samples that were incubated for different time intervals. Bispeci®c activity was assayed with a T cell proliferation assay. Regression analysis of the data obtained predicts half lives of the BsAb and TsAb of;30 h when incubated in serum, as compared with 80 h when incubated in PBS (R 2 .0.9). These data indicate that the bibody and tribody format is compatible with therapeutic use in vivo. Because the effect of two C-terminal fusions on the af®nity of the Fab has not been documented yet, we compared the binding

FIGURE 4.Fab-scFv and Fab-(scFv)

2

lead to cell-cell cross-linking and tumor cell-dependent activation of effector T cells.A, Mitomycin-treated MO4I4

or BCL1 cells were incubated with C3H or BALB/c syngeneic spleen cells, respectively, (E:T ratio 2:1) and 0.5

mg/ml BsAb (B) or TsAb (T) (®lled

columns). To prove speci®city, selected components were omitted from the reaction (open columns).B, A titration curve starting from 1

mg/ml of the BsAb

(E) and the TsAb in a reaction with MO4I4 tumor cells (l), or of the TsAb with BCL1 tumor cells (OE) is shown.C,

51

Cr-labeled MO4I4 cells or BCL1

cells were cocultured for 4 h with C3H or BALB/c syngeneic splenocytes, respectively, (E:T ratio 50:1) in the presence of BsAb (B) or TsAb (T) (®lled

columns). To prove speci®city, selected components were omitted from the reaction (open columns).D, Titration curve starting from 1

mg/ml of BsAb

(anti-hPLAP3anti-CD3) (E) or TsAb assayed for its anti-hPLAP3anti-CD3 axis (l) or its anti-BCL13anti-CD3 axis (OE) of T cell cytotoxicity with

either the labeled MO4I4 or BCL1 tumor cells as a target.

7054 Fab-scFv BISPECIFIC AND TRISPECIFIC ANTIBODIESDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

af®nity of the Fab moiety in the TsAb with the original Fab pre- pared from the E6 mAb. According to the measured surface plasma resonance parameters listed in Table I, the af®nity of the TsAb is very comparable to the af®nity measured for the native Fab. This indicates that the C-terminal fusions do not induce con- formational changes that effect the Ag-binding site. We conclude that the bi- and tribody molecules created have a low tendency to aggregate and can be considered as stable proteins, suitable for therapeutic use.

Discussion

For therapy, large BsAbs exhibit a better pharmacokinetic behavior as

compared with Ab derivatives of low m.w. However, this advantageis compromised by the strongly reduced tissue penetration of large

BsAbs, a feature in which small size Ab derivatives excel. Interme- diate size BsAbs, larger than Fab and smaller than whole Abs, com- bine the high tissue penetration of small molecules with the slow whole-body clearance of large molecules, and hence may have a bet- ter format for therapeutic applications. The drawback of these recom- binant BsAbs is their requirement for ef®cient dimerization, essential for optimal expression and ease of puri®cation. In this respect, het- erotypic CL-CH1 interaction was shown to be a promising scaffold for production of bispeci®c minibodies inE. coli(13). We propose to improve the heterodimerization ef®ciency of the CL:CH1 domains by using a mammalian type of expression sys- tem. Mammalian cells are known to exert a more stringent quality

FIGURE 5.Stability of Fab-scFv and Fab-(scFv)

2 bi- and tribody.A, The TsAb (anti-BCL13anti-hPLAP3anti-CD3) was puri®ed from cell culture

supernatant and analyzed on a Coomassie Brillant Blue-stained SDS-PAGE and on a Superdex 200 XK 26/88 (Amersham Pharmacia Biotech). The column

was calibrated with standard proteins of 669, 150, 43, 13, and 0.5 kDa. The material eluting in the void (Vo) mainly contained contaminating proteins and

almost no TsAb. The peaks corresponding to monomeric (M) and dimeric (D) TsAb are indicated.B, Capture of the Tsab with target cells.i, Dilution series

of the TsAb immunodetected after SDS-PAGE with anti-mouse IgG serum shows that 1% of the starting dose can still be detected.ii, The same amount

of TsAb was washed three times with MO4I4 (hPLAP 1 ), BCL1 (BCL1 1 ), TE2 (CD3 1 ), or SP2/0 (irrelevant) cells.iii, The same experiment was performed

with the parental mAbs E6 (anti-hPLAP), B1 (anti-BCL1), and 145.2c11 (anti-CD3).C, In vitro stability of the bi- and tribody: 1

mg/ml of Ab fragment

was incubated at 37°C for up to 26 h in either PBS or freshly prepared mouse serum. The stability was measured as the remaining activity in a T cell

proliferation assay with MO4I4 (hPLAP 1 ) cells, which accounts for the presence of two speci®cities (anti-hPLAP and anti-CD3).

7055The Journal of ImmunologyDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

control on proteins proceeding along the secretory pathway. In particular, they contain the endoplasmic chaperone BiP/GRP78 that mainly binds to the CH1 domain of the IgH (26) and much weaker to the L chain (27). In agreement with this strict control, we never observed secreted CH1 fragments or Fd chains by them- selves in the culture ¯uids of transfected cells, whereas L chains alone were ef®ciently secreted. Normally, the interaction of L chains with IgHs displaces the associated BiP and thus frees the Abs for secretion. However, coexpression of CL with CH1 or Fd did not lead to secretion of a CL:CH1 or CL:Fd heterodimer. Sim- ilar, cotransfection of L chains, instead of CL, did not result in secretion of L:CH1 heterodimers. Enlargement of CH1 with V H , resulting in formation of a Fd chain, was necessary to obtain se- cretion of L:Fd heterodimers in cotransfected cells. These het- erodimers represented over 90% of the secreted products, re¯ect- ing tight control of their secretion. These results indicate that interaction of CL with CH1 is not ef®cient to free the complex for secretion, and an additional VL:V H interaction is necessary to re- lease the CH1 domain from its endoplasmic chaperone, thus lead- ing to ef®cient secretion. A similar dependence on VL:V H inter- actions for BiP displacement was observed in a recent study of Lee and coworkers, showing that the BiP:Fd interaction could not be displaced when the VL or V H domains were mutated to prevent folding (28). Having established coexpression of L and Fd as the minimal scaffold for ef®cient heterodimerization in mammalian cells, we analyzed the applicability of L:Fd heterodimerization for generat- ing intermediate size BsAbs and TsAbs. To avoid possible sterical hindrance when cross-linking two cells, and to allow a better reach for distant Ags, the C terminus of the L and Fd chains was pre- ferred over the N terminus to fuse scFv molecules. The resulting L:Fd-scFv bispeci®c heterodimer typically represented.90% of the secreted Ig pool. This represents a signi®cant improvement over other expression systems, such as CL:CH1-driven het- erodimerization of minibodies inE. coli, where 63% of the total Ig pool was bispeci®c (13). Furthermore, with the L:Fd template, ef®cient heterodimerization was achieved without introduction of antigenic heterodimerization domains or application of extensive postproductional processing. A ®nal advantage, intrinsic to the present model, derives from the formation of a Fab binding moiety only in correctly folded heterodimers. Thus, BsAb or TsAb can easily be separated from irrelevant side products, such as L-scFv: L-scFv homodimers, by a single immunoaf®nity puri®cation step with the appropriate Ag. Also, an af®nity puri®cation directed against the heavy-chain fusion product would have the same result, because the Fd chain is always in the heterodimer format.

Both Fab-scFv and Fab-(scFv)

2 molecules,which we refer to as bibody and tribody, were fully functional in bispeci®c binding and in cross-linking of effector T cells with tumor cells. Furthermore,

tribodies exhibited dual binding along the three axes of the mol-ecule, demonstrating reach and ¯exibility of the different binding

sites. As a result, the same tribody induced T cell reactivity against two types of tumor cells by alternating use of its dual anti-tumor cell binding moieties (combined with its anti-CD3 moiety). The new bi- and tribody model was shown to have a low tendency to aggregate and to be stable in physiological conditions, making it suitable for therapeutic use. Due to its high level of disul®de-stabilized and speci®c het- erodimerization, this molecule is a useful alternative for generating BsAb and TsAb. The tribody model provides an easy way to con- struct monovalent TsAbs, which is most valuable considering the heterogeneity of most tumors. Another application can be found in increasing the avidity for the tumor antigen or for the tumor cell. In summary, the highly ef®cient heterodimerization of L and Fd chains in mammalian cells constitutes an ideal platform for generating fully functional, disul®de-stabilized BsAbs and TsAbs of intermediate size by C-terminal enlargement with one or two scFv molecules. The dependence on mammalian expression renders these novel recombi- nant Ab derivatives suitable for production in mammalian cell facto- ries or transgenic ¯ock if needed as a therapeutic agent.

Acknowledgments

We thank Dr. J. Demolder and Dr. W. Lammerant (Ghent University) for the use of unpublished plasmids. Dr. M. Hall (University of Birmingham, U.K.), Dr. M. De Broe (University of Antwerp, Belgium), and Dr. G. LeClercq (Ghent University) are acknowledged for donating HEK293T cells, MO4I4 cells, and 145.2C11 mAbs, respectively.

References

1. Glennie, M. J., H. M. McBride, A. T. Worth, and G. T. Stevenson. 1987. Prep-

aration and performance of bispeci®c F(ab9 g) 2 antibody containing thioether- linked Fab9 gfragments.J. Immunol. 139:2367.

2. Milstein, C., and A. C. Cuello. 1983. Hybrid hybridomas and their use in im-

munohistochemistry.Nature 305:537.

3. Mack, M., G. RiethmuÈller, and P. Kufer. 1995. A small bispeci®c antibody con-

struct expressed as a functional single-chain molecule with high tumor cell cy- totoxicity.Proc. Natl. Acad. Sci. USA 92:7021.

4. Pack, P., and A. PluÈckthun. 1992. Miniantibodies: use of amphipathic helices to

produce functional, ¯exibly linked dimeric Fv fragments with high avidity in

Escherichia coli. Biochemistry 31:1579.

5. Luo, D., M. Geng, A. A. Noujaim, and R. Madiyalakan. 1997. An engineered

bivalent single-chain antibody fragment that increases antigen binding activity.

J. Biochem. 121:831.

6. MuÈller, K. M., K. M. Arndt, and A. PluÈckthun. 1998. A dimeric bispeci®c min-

iantibody combines two speci®cities with avidity.FEBS Lett. 432:45.

7. Wu, A. M., W. Chen, A. Raubitschek, L. E. Williams, M. Neumaier, R. Fischer,

S. Z. Hu, T. Odom-Maryon, J. Y. Wong, and J. E. Shively. 1996. Tumor local- ization of anti-CEA single-chain Fvs: improved targeting by non-covalent dimers.Immunotechnology 2:21.

8. Behr, T., W. Becker, E. Hannappel, D. M. Goldenberg, and F. Wolf. 1995. Tar-

geting of liver metastases of colorectal cancer with IgG, F(ab9) 2 , and Fab9anti- carcinoembryonic antigen antibodies labeled with 99mTc: the role of metabolism and kinetics. Cancer Res. 55:5777.

9. Zhu, Z., G. Zapata, R. Shalaby, B. Snedecor, H. Chen, and P. Carter. 1996. High

level secretion of a humanized bispeci®c diabody fromEscherichia coli.Bio- technology 14:192.

10. Merchant, A. M., Z. Zhu, J. Q. Yuan, A. Goddard, C. W. Adams, L. G. Presta,

and P. Carter. 1998. An ef®cient route to human bispeci®c IgG.Nat. Biotechnol.

16:677.

11. McGregor, D. P., P. E. Molloy, C. Cunningham, and W. J. Harris. 1994. Spon-

taneous assembly of bivalent single chain antibody fragments inEscherichia coli.

Mol. Immunol. 31:219.

12. Hu, S., L. Shively, A. Raubitschek, M. Sherman, L. E. Williams, J. Y. Wong,

J. E. Shively, and A. M. Wu. 1996. Minibody: a novel engineered anti- carcinoembryonic antigen antibody fragment (single-chain Fv-CH3) which ex- hibits rapid, high-level targeting of xenografts.Cancer Res. 56:3055.

13. Muller, K. M., K. M. Arndt, W. Strittmatter, and A. PluÈckthun. 1998. The ®rst

constant domain (C(H)1 and C(L)) of an antibody used as heterodimerization domain for bispeci®c miniantibodies.FEBS Lett. 422:259.

14. Zhang, J. X., I. Braakman, K. E. Matlack, and A. Helenius. 1997. Quality control

in the secretory pathway: the role of calreticulin, calnexin and BiP in the retention of glycoproteins with C-terminal truncations.Mol. Biol. Cell 8:1943.

15. DuBridge, R. B., P. Tang, H. C. Hsia, P. M. Leong, J. H. Miller, and M. P. Calos.

1987. Analysis of mutation in human cells by using an Epstein-Barr virus shuttle

system.Mol. Cell. Biol. 7:379.

16. Grooten, J., and W. Fiers. 1989. Acquisition by the murine host of responsiveness

toward various neoplastic cell lines, but not toward self, through adoptive transfer of a helper T-lymphocyte clone with antiself speci®city.Cancer Res. 49:3872. Table I.Comparing af®nities of natural Fab9and Fab-moiety in

Fab-(scFv)2 fusion

a k d (S 21
)k a (M 21
S 21
)K D (M)K A (M 21

TsAb 3.2310

24

5.8310

4

5.5310

29

1.8310

8

Fab 3.3310

24

7.6310

4

4.3310

29

2.3310

8 a Apparent af®nities as measured with biologic interaction analysis using SPR on a BIAcore 2000. Fab prepared from the E6 mAb and TsAb were captured with an anti-mouse kAb coupled to a CM5 chip (BIAcore, SE). R max (Ab) was consistently

500650 RU. Puri®ed hPLAP was used as an analyte in concentrations of 200, 100,

50, 25, and 12.5 nM, at a ¯ow rate of 10

ml/min. Glycine-HCl 50 mM (pH 2.0) was used for elution. Thek d ,k a ,K D , andK A values were calculated from the sensogram using BIAeval 2.0 software.

7056 Fab-scFv BISPECIFIC AND TRISPECIFIC ANTIBODIESDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023

17. Smans, K. A., M. F. Hoylaerts, S. Narisawa, J. L. Millan, and M. E. De Broe.

1995. Bispeci®c antibody-mediated lysis of placental and germ cell alkaline

phosphatase targeted solid tumors in immunocompetent mice.Cancer Res. 55: 4383.

18. Hendrix, P. G., S. E. Dauwe, A. Van de Voorde, E. J. Nouwen, M. F. Hoylaerts,

and M. E. De Broe. 1991. Radiolocalisation and imaging of stably HPLAP- transfected MO4 tumours with monoclonal antibodies and fragments.Br. J. Can- cer 64:1060.

19. De Jonge, J., C. Heirman, M. de Veerman, S. Van Meirvenne, M. Moser, O. Leo,

and K. Thielemans. 1998. In vivo retargeting of T cell effector function by re- combinant bispeci®c single chain Fv (anti-CD33anti-idiotype) induces long- term survival in the murine BCL1 lymphoma model.J. Immunol. 161:1454.

20. Niwa, H., K. Yamamura, and J. Miyazaki. 1991. Ef®cient selection for high-

expression transfectants with a novel eukaryotic vector.Gene 108:193.

21. De Sutter, K., V. Feys, A. Van de Voorde, and W. Fiers. 1992. Production of

functionally active murine and murine:human chimeric F(ab9) 2 fragments in

COS-1 cells.Gene 113:223.

22. De Jonge, J., J. Brissinck, C. Heirman, C. Demanet, O. Leo, M. Moser, and

K. Thielemans. 1995. Production and characterization of bispeci®c single-chain

antibody fragments.Mol. Immunol. 32:1405.23. O'Mahoney, J. V., and T. E. Adams. 1994. Optimization of experimental vari-

ables in¯uencing reporter gene expression in hepatoma cells following calcium phosphate transfection.DNA Cell Biol. 13:1227.

24. De Waele, P., V. Feys, A. Van de Voorde, F. Molemans, and W. Fiers. 1988.

Expression in non-lymphoid cells of mouse recombinant immunoglobulin di- rected against the tumour marker human placental alkaline phosphatase.Eur.

J. Biochem. 176:287.

25. Van de Voorde, A., G. De Groote, P. De Waele, M. E. De Broe, D. Pollet,

J. de Boever, D. Vandekerckhove, and W. Fiers. 1985. Screening of sera and tumor extracts of cancer patients using a monoclonal antibody directed against human placental alkaline phosphatase.Eur. J. Cancer Clin. Oncol. 21:65.

26. Kaloff, C. R., and I. G. Haas. 1995. Coordination of immunoglobulin chain fold-

ing and immunoglobulin chain assembly is essential for the formation of func- tional IgG.Immunity 2:629.

27. Hendershot, L. M. 1990. Immunoglobulin heavy chain and binding protein com-

plexes are dissociated in vivo by light chain addition.J. Cell Biol. 111:829.

28. Lee, Y. K., J. W. Brewer, R. Hellman, and L. M. Hendershot. 1999. BiP and

immunoglobulin light chain cooperate to control the folding of heavy chain and ensure the ®delity of immunoglobulin assembly.Mol. Biol. Cell 10:2209.

7057The Journal of ImmunologyDownloaded from http://journals.aai.org/jimmunol/article-pdf/165/12/7050/1128461/7050.pdf by guest on 06 August 2023