WO2003012069A2 - Compositions and methods for generating chimeric heteromultimers - Google Patents

Abstract

The present invention provides a technique for specific assembly of monomericpolypeptides to form a heteromultimer. This technique is particularly useful for generating a genetically diverse repertoire of heteromultimers such as antigen-binding units. The invention also provides both non-single-chain and single-chain antigen-binding units that are assembled by the technique described herein. The present invention also provides recombinant polynucleotides, vectors, host cells, and kits for producing the subjectantigen-binding units. Further provided by the invention are methods of using the subject antigen-binding units.

Description

COMPOSITIONS AND METHODS FOR GENERATING CHIMERIC

HETEROMULTIMERS

CROSS REFERENCE TO RELATED APPLICATION This application is a continuation-in-part of U.S. Utility Patent Application 09/921,144, filed August 1, 2001, pending, which is hereby incorporated herein by reference in its entirety.

TECHNICAL FIELD This invention is in the field of immunology. Specifically, the invention relates to the generation of chimeric heteromultimers such as non-single-chain antigen-binding units using unique heterodimerization sequences. This invention also relates to the generation of single-chain antigen-binding units stabilized by the subject heterodimerization sequences. The compositions and methods embodied in the present invention are particularly useful for identifying antigen-binding units that are of major diagnostic and/or therapeutic potential.

BACKGROUND OF THE INVENTION Antibodies or immunoglobulins are molecules that recognize and bind to specific cognate antigens. Because of their exclusive specificities, antibodies, particularly monoclonal antibodies, have been widely used in the diagnosis and treatment of a variety of human diseases. The basic immunoglobulin (Ig) in vertebrate systems is composed of two identical light ("L") chain polypeptides (approximately 23 kDa), and two identical heavy

("H") chain polypeptides (approximately 53 to 70 kDa). The four chains are joined by disulfide bonds in a "Y" configuration. At the base of the Y, the two H chains are bound by covalent disulfide linkages. The L and H chains are organized in a series of domains.

The L chain has two domains, corresponding to the C region ("CL") and the other to the

V region ("VL"). The H chain has four domains, one corresponding to the V region

("VH") and three domains (CHI, CH2 and CH3) in the C region. The antibody contains two arms (each arm being a Fab fragment), each of which has a VL and a VH region associated with each other. It is this pair of V regions (VL and VH) that differ, from one antibody to another (due to amino acid sequence variations), and which together are responsible for recognizing the antigen and providing an antigen-binding site. More specifically, each V region is made up from three complementarity determining regions (CDR) separated by four framework regions (FR). The CDR's are the most variable part of the variable regions, and they perform the critical antigen binding function. The CDR regions are derived from many potential germ line sequences via a complex process involving recombination, mutation and selection.

Research in recent years has demonstrated that the function of a binding antigen can be performed by fragments of a whole antibody. Exemplary antigen binding fragments are (i) the Fab fragment consisting of the VL, VH, CL and CHI domains; (ii) the Fd fragment consisting of the VH and CHI domains; (iii) the dAb fragment (Ward, E. S. et al, Nature 341, 544-546 (1989) which consists of a VH domain; (iv) isolated CDR regions; and (v) F(ab')₂ fragments, a bivalent fragment comprising two Fab fragments linked by a disulfide bridge at the hinge region; and (vi) the Fv fragment consisting of the VL and VH domains of a single arm of an antibody. The Fv fragment is the smallest functional unit required for high affinity binding of antigen.

One major challenge in the antibody field has been to reconstitute a vast diverse repertoire of immunoglobulins that mimics the immunoglobulin pool in the human immune system. Such a repertoire generally has a complexity ranging from 10⁸ to 10¹³ distinct immunoglobulins. The generation of such a repertoire would greatly facilitate the identification and production of immunoglobulins capable of interacting specifically with therapeutic targets. However, the design and production of such a repertoire has traditionally been hampered by the lack of a stabilizing means for assembly of the minimal functional unit, namely the Fv fragment. It is a well-known problem in the art that the VH and VL regions, when expressed alone, have very low interaction energy (Glockshuber et al. (1990) Biochemistry 29(6): 1362-1367). The two components dissociate at low protein concentrations and are too unstable for many applications at physiological body temperature. It is also a long-recognized technical obstacle that large proteins, such as whole antibodies (albeit extremely stable), do not express at an appreciable level in the host cell, thus rendering the construction of a highly diverse antibody repertoire very difficult.

More recently, three approaches have been developed to generate stable VL and VH complexes. However, each of these techniques bears a number of intrinsic limitations and none of them circumvents the aforementioned technical hurdles completely. The first approach uses a peptide linker to connect the VL and VH as a single-chain ("scFv") (Huston et al. (1988) Proc. Natl Acad. Sci. U.S.A 85:5879-5883). While the resulting scFv exhibits substantial antigen-binding activity, not all antibodies can be made as single chains and still retain high binding affinity (Huston et al. (1988) Proc. Natl. Acad. Sci. U.S.A. 85:5879-5883; Stemmer et al. (1993) Biotechniques 14(2): 256-265). In part, this is due to the interference of linker sequences with the antigen binding sites. The second approach involves inserting a pair of cysteine residues in the VL and VH regions to generate a disulfide-bond stabilized Fv ("dsFv") (Brinkmann et al. (1993) Proc. Natl Acad. Sci. U.S.A. 90(16): 7538-7542). The incorporated disulfide linkage, however, is unstable under reducing conditions in many host cells. For instance, in cytosol of E. Coli, the inter-molecular disulfide bond is often insufficient to stabilize the VL and VH complex. Moreover, this method typically requires 3 -dimensional structural information of the V regions to ensure that the cysteine pair is inserted in a proper place without disruption the binding activity. Because the 3 -dimensional information of vast majority of the existing antibodies is unknown, this approach has little practical utility, and is particularly unsuited for antibody library construction, especially for constructing antibody repertoires derived from B cells. The third approach for stabilizing the VL and VH regions utilizes the disulfide bonds native to the CHI and CL domains. This method proceeds with grafting a disulfide-bond linked CHI and CL domains to the C-termini of the VL and VH regions in order to reconstitute a Fab fragment. While the resulting Fab fragment is generally more stable and often exhibits higher binding affinity than scFv, Fab is not optimal for high level expression and antibody repertoire construction due to its large size.

Certain dimerization sequences that form coiled-coil structures have also been employed to assemble multivalent antibodies. Specifically, U.S. Patent No. 5,932,448 describes a bispecific F(ab')₂ heterodimer linked by the Fos and Jun leucine zippers. The Fos and Jun leucine zippers are well-characterized sequences known to preferentially form heterodimers. However, they still exhibit significant propensity to form homodimers under physiological buffer conditions and/or at physiological body temperature (O'Shea et al. (1992) Cell 68: 699-708; Vidal et al (1996) Proc. Natl Acad. Sci. U.S.A.). In fact, the JunlJun homodimer is so stable that formation of Fosi 'Jun heterodimer in vitro requires dissociation of the JunlJun homodimer by first heating or reduction with 2-mercaptoethanylamine (see U.S. Patent No. 5910,573 column 7 lines 35-37; U.S. Patent No. 5,932,448, column 16 lines 15-30). When tested in vivo, both Fos and Jun yield detectable amounts of homodimers (see, e.g. column 15, lines 41-43 of U.S. Patent No. 5,932,448; and Vidal et al. (1996) Proc. Natl. Acad. Sci. U.S.A). While the existence of some homodimerization propensity may not be of substantial concern for the production of a single antibody species, such propensity is problematic for antibody repertoire construction, where high efficiency of heterodimerization between VL and VH regions is required.

Aside from Fos and Jun leucine zippers, U.S Patent No. 5,824,483 by Houston et al. describes the construction of a combinatorial library of coiled-coil dimerization peptides. Houston et al. proposes that the library is useful for identifying a polypeptide that is capable of interacting specifically with a selected macromolecule ligand such as antibodies (see last paragraph bridging pages 8 and 9). Apparently, Houston et al. concerns the selection of "antigen peptides" that bind to targeted antibodies, rather than the construction and selection of target antibodies. Focusing on an entirely different purpose, Houston et al. does not describe or even suggest the use of coiled-coil sequences to generate stable antigen-binding units. Thus, there remains a considerable need for improved compositions and methods to generate stable antigen-binding units and repertoires thereof to effect identification of therapeutic antigen-binding units. An ideal antigen-binding unit would be more stable than an Fv fragment, but would preferably be smaller than a Fab fragment to allow large- scale production and efficient display. Such antigen-binding unit would also serve as a building block for constructing multivalent and/or multispecific antibodies. The present invention satisfies these needs and provides related advantages as well.

SUMMARY OF THE INVENTION A principal aspect of the present invention is the design of a technique for specific assembly of monomeric polypeptides to form a stable heteromultimer. This technique of heteromultimer production facilitates high throughput production of functional heteromultimers and avoids the assembly of undesired homodimers. The method is particularly useful for generating a genetically diverse repertoire of heteromultimers such as antigen-binding units. The technique can readily be adapted to a variety of "genetic package display" technologies that facilitate the selection of antigen-binding units possessing the desired binding specificities. Such genetic package display technologies are detailed in U.S. Patent Nos. 6248516, 5969108, 5885793, 5837500, 5571698, 5223409, 5514548, WO9005144, EP0368684, WO09201047, WO09311236, and WO09708320. The subject antigen-binding unit is assembled and stabilized by the pairwise affinity of a distinct pair of heterodimerization sequences. The sequences are distinct in that at least one member of the heterodimerization pair is essentially incapable of forming homodimers under physiological buffer conditions and/or at physiological body temperatures. In certain embodiments, the stabilized antigen-binding unit not only has a molecular size smaller than a Fab fragment, but also exhibits the required binding specificity and affinity. Moreover, certain non-single-chain antigen-binding units of the present invention bear higher binding affinities than the corresponding conventional single-chain antibodies (scFv). The antigen-binding unit is particularly suited for antibody library construction and display. Certain configurations of the subject antigen- binding unit serve as convenient building units for multivalent and multispecific immunoglobulins. Specifically, the present invention provides a non-single-chain antigen-binding unit comprising: (a) a light (L) chain polypeptide comprising a light (L) chain variable region fused to a first heterodimerization sequence; (b) a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures. Preferably, both of the first and second heterodimerization sequences are essentially incapable of forming homodimers under physiological buffer conditions and at physiological body temperatures. In another aspect, the present invention provides a non-single-chain antigen- binding unit comprising: (a) a light (L) chain polypeptide comprising a light (L) chain variable region fused to a first heterodimerization sequence; (b) a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences which are derived from heterodimeric receptors. In one aspect, the first and second heterodimerization sequences comprising heterodimerization receptor sequences that mediate heterodimerization of the receptors. In yet another aspect, the subject heterodimerization sequences form a coiled-coil dimer. In still another aspect, the L and the H chain polypeptides dimerize via non-covalent pairwise affinity of the two heterodimerization sequences. Preferably, the L or the H chain polypeptide further comprises a flexon that is flanked by the variable region and the heterodimerization sequence. Both the L and H polypeptide sequences may be derived from human L and H chains. To further stabilize the heterodimeric antigen-binding units, cysteine residues can be introduced to provide disulfide bonds between the first and the second heterodimerization sequences. The non-single-chain antigen-binding units may be monovalent or multivalent. They may be monospecific or multispecific. Preferred multispecific Abus are bispecific, trispecific and tetraspecific molecules. In a separate embodiment, the present invention provides a single-chain antigen- binding unit comprising a light (L) chain variable region and a heavy (H) chain variable region connected by a first and a second heterodimerization sequence spanning the distance between the C-terminus of one of the region to the N-terminus of the other region, wherein the two regions form an intra-molecular dimer via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures. Preferably, both of the first and second heterodimerization sequences are essentially incapable of forming homodimers under physiological buffer conditions and at physiological body temperatures. In another aspect, the present invention provides a single-chain antigen-binding unit, wherein the VL and VH regions form an intra-molecular dimer via pairwise affinity of the first and second heterodimerization sequences which are derived from heterodimeric receptors. In one aspect, the first and second heterodimerization sequences comprising heterodimerization receptor sequences that mediate heterodimerization of the receptors. In yet another aspect, first and second heterodimerization sequences form a coiled-coil dimer. In another aspect, the first and second heterodimerization sequences dimerize via non-covalent pairwise affinity. Both the VL and VH regions can be derived from the corresponding sequences in a human L and H chains, respectively. Both the non-single-chain and single-chain antigen-binding units can be conjugated to a chemically functional moiety. Exemplary functional moieties include but are not limited to signal peptides, agents that enhance immunologic reactivity, agents that facilitate coupling to a solid support, vaccine carriers, bioresponse modifiers, toxins, detectable labels, paramagnetic labels, and drugs. Preferred heterodimerization sequences contained in the subj ect antigen-binding units are derived from C-terminal sequences of GABA_B receptor 1 and GABA_B receptor 2, respectively. More preferably, the first heterodimerization sequence is linked to a cysteine residue, said first heterodimerization comprising a GABA_B receptor 1 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 2; and the second heterodimerization sequence is linked to a cysteine residue, said second heterodimerization comprising a GABA_B receptor 2 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4. Alternatively, the first heterodimerization sequence is linked to a cysteine residue, said first heterodimerization comprising a GABA_B receptor 2 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4; and the second heterodimerization sequence is linked to a cysteine residue, said second heterodimerization comprising a GABA_B receptor 1 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 2. The present invention provides a recombinant polynucleotide comprising a coding sequence that encodes the L and/or H polypeptide of a non-single-chain antigen- binding unit. The invention also provides a recombinant polynucleotide comprising a coding sequence that encodes the VL or VH regions of a single-chain antigen-binding unit. Also provided is a vector comprising any one of the recombinant polynucleotides described herein. The vector can be an expression vector, e.g. a phage display vector. Further provided in this invention is a selectable library of expression vectors encoding a repertoire of antigen-binding units, comprising more than one subject vector. Preferably, the selectable library comprises a plurality of phage display vectors. The present invention also provides a host cell comprising the subject recombinant polynucleotides. The recombinant polynucleotide encoding the L chain polypeptide and the polynucleotide encoding the H chain polypeptide may be present in a single vector or in separate vectors. The host cell may be eukaryotic or prokaryotic. In yet another embodiment, the present invention provides a method of producing a non-single-chain antigen-binding unit. The method involves the following steps: (a) expressing in a host cell a first recombinant polynucleotide encoding a light (L) chain polypeptide comprising a light (L) chain variable region fused to a first heterodimerization sequence, and a second recombinant polynucleotide encoding a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures; and optionally (b) isolating the antigen-binding unit expressed in the host cell. The produced antigen-binding unit may also contain heterodimerization sequences that are derived from heterodimeric receptors. Additionally, the non-single- chain antigen-binding expressed in step (a) can be displayed on surface of the host cell. Preferably, the non-single-chain antigen-binding expressed in step (a) is displayed on a phage particle. In still another embodiment, the present invention provides a method of producing a non-single-chain antigen-binding unit, the method comprises the steps of (a) preparing a first recombinant polynucleotide encoding a light (L) chain polypeptide comprising a light (L) chain variable region fused to a first heterodimerization sequence, and a second recombinant polynucleotide encoding a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures; and (b) allowing the first and second polypeptides to dimerize via pairwise affinity of the first and second heterodimerization sequences. The step of dimerization may take place in vitro or in vivo. This invention also includes a method of producing a single-chain antigen- binding unit. The methods involves the steps of (a) expressing in a host cell a polynucleotide comprising a coding sequence that encodes the subject single-chain antigen-binding unit; and optionally (b) isolating the single-chain antigen-binding unit expressed in the host cell. This invention further includes a method of displaying a chimeric heteromultimer comprising at least two polypeptides on a surface of a host cell. This method comprises expressing in the host cell (i) a first recombinant polynucleotide encoding a first polypeptide fused to a first heterodimerization sequence and a surface presenting sequence; (ii) a second recombinant polynucleotide encoding a second polypeptide fused to a second heterodimerization sequence; wherein the first and second polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; wherein at least one of the heterodimerization sequences is incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures. In one aspect, the first and second polynucleotides are expressed by a single phage display vector. In another aspect, the first and second polynucleotides are expressed by separate phage display vectors. The chimeric heteromultimer is preferably a non-single-chain antigen-binding unit of the present invention. The invention also encompasses a method of identifying a non-single-chain antigen-binding unit that is immunoreactive with a desired antigen. The method comprises the steps of: (a) preparing a genetically diverse repertoire of antigen-binding units, wherein the repertoire comprises more than one subject antigen-binding unit; (b) contacting the repertoire of antigen binding units with the desired antigen; and (c) detecting a specific binding between antigen binding units and the antigen, thereby identifying the antigen-binding unit that is immunoreactive with the desired antigen. In one aspect of this embodiment, the repertoire of antigen-binding units is prepared by expressing a library of vectors encoding a plurality of the antigen-binding units. Preferably, the library of vectors comprises a plurality of phage vectors. Finally, the present invention provides a kit comprising a vector of this invention in suitable packaging.

EXPLANATION OF ABBREVIATIONS USED HEREIN l. Nsc: Non-single chain 2. Sc: Sing-chain 3. Abu: Antigen-binding unit 4. Abus: Antigen-binding units 4. L chain: Light chain 5. H chain: Heavy chain 6. VL: Light chain variable region 7. VH: Heavy chain variable region

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic representation depicting various antigen-binding units. Figure 2 depicts the nucleotide and amino acid sequences of GABAB receptor 1 and 2 that were used in constructing the subject Abus. The coiled-coil sequences are derived from human GABAβ-Rl and GABAβ-R2 receptors. The coding amino acid sequences from GABA_B receptor 1 begins with EEKS and ends with QLQS as shown in the top panel of Figure 2 . The coding amino acid sequences from GABA_B receptor 2 begins with TSRL and ends with QLQD as shown in the bottom panel of Figure 2 . A flexible SerArgGlyGlyGlyGly spacer was added to the amino-termini of Rl and R2 heterodimerization sequences to favor the formation of functional Fv heterodimer. To further stabilize the heterodimer, we have introduced a ValGlyGlyCys spacer to lock the heterodimeric coiled-coil pair via the disulfide bond between the cysteine residues (SEQ ID NOS. 2 and 4). The SerArg coding sequences at the N-terminus of the GGGG spacer provides Xbal or Xhol sites for fusion of the GR1 (heterodimerization sequence derived from GABAB 1 receptor) and GR2 (heterodimerization sequence derived from GABAB2) domains to the carboxyl-termini of VH and VL fragments, respectively. Figure 3A is a schematic representation of two expression vectors pABMXl and pABMX2. Both pABMXl and ρABMX2 were derived from pbluescript SK(+) comprising an ampicillin-resistance gene (Amp) for antibiotic selection, a plasmid origin of replication (colEl ori), an fl phage origin of replication (fl ori), and lac promoter/lac OI driven protein expression cassette (plac-RBS-p8Leader-DH tag for pABMXl, plac- RBS-pelB Leader-DH tag for pABMX2 ). The heterologous sequence is expressed as a DH- tag (HA and 6 x His tag) fusion protein, and is directed by the signal peptide either p8 leader or pelB leader into periplasmid space, where the leader sequence is cleaved Figure 3B depicts the sequences (SEQ ID NOS. 5-8) after the lac promoter between Agel and Bglll sites of pABMXl and pABMX2. The Hindlll/Xbal or Hindlll/Notl or Xbal/Notl sites can be used for inserting heterologous sequences that are to be expressed in pABMXl vector. Additional cloning sites included in the pABMX2 vector are Ncol, Pstl, Xbal and Notl. Figure 4 A is a schematic representation of the phagemid vectors pABMDl and pABMD2 useful for displaying antigen-binding units. pABMDl and ABMD2 were derived from pABMXl and pABMX2 respectively. They comprise all of the functional elements of pABMXl and pABMX2 vectors, and pill gene of a filamentous phage. The pill gene was inserted immediately adjacent to the 3' end of DH-tag. The lac promoter drives expression of a heterologous sequence as pill capsid fusion protein, which in turn is displayed on a phage particle upon superinfection of a helper phage such as KO7 (Amersham Pharmacia Biotech) or R408 helper phage (Stratagene). This vector can also be used for soluble protein expression in a non-suppressor bacterial strain. Figure 4B depicts the sequence (SEQ ID NOS. 9-12) after the lac promoter between Agel and Sail sites of pABMDl and pABMD2. Figure 5 A is a schematic representation of the vectors, pABMX5 and pABMX6. pABMX5 and ABMX6 were derived from pABMXl and pABMX2 respectively. Different leader sequences were incorporated into pABMX5 and pABMX6. The subcloning sites for insertion of heterologous sequence, e.g. VH gene, also differ in these two vectors. pABMX5 contains the p8 leader, and pABMX6 contains the pelB leader. Two protein expression cassettes employing the lac promoter were engineered into these two vectors. The first cassette is used to express VH-GR1 (VH- heterodimerization sequence of GABA_B receptor 1) and the second is used to express VL-GR2 (VL- heterodimerization sequence of GABAB receptor 2). The DH-tag was fused to GR2 domain to facilitate purification of the resulting heterodimers. Figure 5B depicts the sequences (SEQ ID NOS. 13-16) between the leader sequence and DH-tag in pABMX5 and pABMX6 vectors. In addition, the ribosome- binding site, DH-tag, subcloning sites for insertion of VH, VL, GR1 and GR2, are also indicated. Figure 6 A is a schematic representation of phagemid vectors, pABMD5 and pABMD6, which are useful for expressing and displaying ccFv on a phage particle. pABMD5 and ABMD6 were derived from pABMX5 and pABMX6 respectively, pill gene derived from the filamentous phage was inserted immediately after DH-tag. The VL-GR2 proteins were linked to pill capsid protein to facilitate display of the ccFv heterodimer. Figure 6 B depicts the sequences (SEQ ID NOS. 17-20) between the leader sequence and pill for pABMD5 and pABMD6 vectors. In addition, ribosome binding site, DH-tag, partial pill , subcloning sites for insertion of VH, VL, GR1 and GR2, are also indicated. Figure 7 depicts the vector pAMEX7 useful for expressing ccFv fragment in yeast. Figure 8 depicts the results of an ELIS A assay using AM2-scFv fragments that were expressed by the pABMXl vector. The results show a dose-dependent AM2-scFv binding to its antigen AM2. Figure 9 depicts the results of an ELIS A assay using AM2-scFv fragments that were displayed on phage particles. The results demonstrate the assembly of functional scFv fragments on phage particles using the phagemid vector pABMDl. Figure 10A depicts the results of SDS-PAGE analysis of AMl-ccFv expressed in E. coli under reducing and non-reducing condition. The results demonstrate successful expression and assembly of heterodimeric ccFv in E. coli. Figure 1 OB depicts the results of an ELIS A assay using soluble AMI -ccFv expressed in E. coli. The results indicate successful assembly of functional ccFv with expected binding specificity to its corresponding antigen. Figure 11 A depicts a comparison of the antigen binding capability of AMl-ccFv expressing phage and that of AMl-scFv expressing phage. The results demonstrate that the phage particles displaying the AMl-ccFv fragments exhibit slightly higher binding capability than the phages displaying conventional scFv fragments. Figure 1 IB depicts a comparison of the antigen binding capability of AM2-ccFv expressing phage and that of AM2-scFv expressing phage. The results indicate that the binding capability of the phage particles displaying the AM2-ccFv fragments is about one order of magnitude higher than that of the AM2-scFv expressing phages. Figure 12 depicts three multivalent Abu configurations, each comprising more than one basic ccFv unit. Figure 13 depicts four bivalent Abu configurations, each comprising a basic ccFv unit and a scFv or a dsFv fragment. Figure 14 depicts three trivalent Abu configurations, each comprising one or more basic ccFv unit, one or more scFv or dsFv fragment. Figure 15 depicts four bispecific Abu configurations, each comprising one or more basic ccFv unit with distinct binding specificities, and/or a scFv or dsFv fragment. Figure 16 depicts three additional bispecific Abu configurations. Figure 17 depicts three trispecific Abu configurations, each comprising at least one basic ccFv unit, and at least one scFv or dsFv fragment. Figure 18 depicts two exemplary single Abus, in which the heterodimerization sequences are arranged in either a parallel or anti-parallel configuration. Figure 19 is a schematic representation of ccFv displayed on the surface of a prokaryotic or eukaryotic cell. The top panel depicts ccFv displayed by a phage particle that is adhered to the surface of a host cell. Figure 20 depicts the results of panning a "model phage library" using ELIS A involving immobilized protein antigen for AM2 from panning of a "model library". The "model library" contained a mixture of AM2-ccFv phages and unrelated phages, AM1- ccFv, in a ratio of 1 : 10⁶ and 1 : 10⁷, respectively. The readings of the ELIS A, since it detected only AM2-ccFv reactivity, would represent the population of AM2-ccFv in the mixture. After the first round of panning, no ELIS A signal was detected, indicating that the AM2-ccFv fraction present in the mixture was still low. However, after two rounds of panning, the AM2-ccFv fraction measured becomes predominant, as evidenced by a dramatic increase in ELSIA signal. Figure 21 depicts PCR analysis to confirm enrichment of AM2-ccFv via panning the model library shown in Figure 12. Before panning, one AM2-ccFv phage was detected in every 10⁷ unrelated phages (0.00001%). After the first round of panning, the occurrence of AM2-ccFv reached 4.4% and then 100% after the second round of panning. Figure 22 depicts the design of CDR3 of VH (heavy chain) of AM2-ccFv. The degenerative DNA oligo was amplified by two PCR primers that franked by two subcloning restriction sites. The oligo was first amplified in PCR and subsequently restriction digested before being subcloned to replace the wild type CDR3 of VH of AM2-ccFv. Figure 23 depicts AM2-ccFv phage ELISA results from individual clones that were randomly picked from the 5th (upper panel) and the 7th (lower panel) pannings. Each bar represented the reading at OD405 of a particular AM2-ccFv variant. Figure 24 depicts the sequences of variants selected from the AM2-ccFv CDR3

VH library panning. Shown on the top is the amino acid sequence of the library (X indicated multiple options of residues derived from the degenerative DNA oligo described in Figure 22); shown below it are the selected residues at each "X" position from panning and their hit rates (occurrence rate of a particular residue). Positions that were not changed by the library design are shown in blank. The consensus derived from the panning is shown on the bottom. Figure 25 depicts a comparison of the Ko_ff rates of an AM2-ccFv variant and the wild type AM2-ccFv. The slope of the curves represents how fast the test antibody dissociates from the antigen.

MODEfS FOR CARRYING OUT THE INVENTION Throughout this disclosure, various publications, patents and published patent specifications are referenced by an identifying citation. The disclosures of these publications, patents and published patent specifications are hereby incorporated by reference into the present disclosure.

General Techniques: The practice of the present invention will employ, unless otherwise indicated, conventional techniques of immunology, biochemistry, chemistry, molecular biology, microbiology, cell biology, genomics and recombinant DNA, which are within the skill of the art. See, e.g., Matthews, PLANT VIROLOGY, 3^rd edition (1991); Sambrook, Fritsch and Maniatis, MOLECULAR CLONING: A LABORATORY MANUAL, 2^nd edition (1989); CURRENT PROTOCOLS IN MOLECULAR BIOLOGY (F. M. Ausubel, et al. eds., (1987)); the series METHODS IN ENZYMOLOGY (Academic Press, Inc.): PCR 2: A PRACTICAL APPROACH (M.J. MacPherson, B.D. Hames and G.R. Taylor eds. (1995)), Harlow and Lane, eds. (1988) ANTIBODIES, A LABORATORY MANUAL, and ANIMAL CELL CULTURE (R.I. Freshney, ed. (1987)). As used in the specification and claims, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. For example, the term "a cell" includes a plurality of cells, including mixtures thereof.

Definitions: The terms "polypeptide", "peptide" and "protein" are used interchangeably herein to refer to polymers of amino acids of any length. The polymer may be linear, cyclic, or branched, it may comprise modified amino acids, and it may be interrupted by non-amino acids. The terms also encompass amino acid polymers that have been modified, for example, via sulfation, glycosylation, lipidation, acetylation, phosphorylation, iodination, methylation, oxidation, proteolytic processing, phosphorylation, prenylation, racemization, selenoylation, transfer-RNA mediated addition of amino acids to proteins such as arginylation, ubiquitination, or any other manipulation, such as conjugation with a labeling component. As used herein the term "amino acid" refers to either natural and/or unnatural or synthetic amino acids, including glycine and both the D or L optical isomers, and amino acid analogs and peptidomimetics. A polypeptide or amino acid sequence "derived from" a designated protein refers to the origin of the polypeptide. Preferably, the polypeptide have an amino acid sequence that is essentially identical to that of a polypeptide encoded in the sequence, or a portion thereof wherein the portion consists of at least 10-20 amino acids, preferably at least 20-30 amino acids, more preferably at least 30-50 amino acids, or which is immunologically identifiable with a polypeptide encoded in the sequence. This terminology also includes a polypeptide expressed from a designated nucleic acid sequence. A "chimeric" protein contains at least one fusion polypeptide comprising regions in a different position in the sequence than that occurs in nature. The regions may normally exist in separate proteins and are brought together in the fusion polypeptide; or they may normally exist in the same protein but are placed in a new arrangement in the fusion polypeptide. A chimeric protein may be created, for example, by chemical synthesis, or by creating and translating a polynucleotide in which the peptide regions are encoded in the desired relationship.

74 A "multimeric protein" as used herein refers to a globular protein containing more than one separate polypeptide or protein chain associated with each other to form a single globular protein in vitro or in vivo. The multimeric protein may consist of more than one polypeptide of the same kind to form a "homomultimer." Alternatively, the multimeric protein may also be composed of more than one polypeptide of distinct sequences to form a "heteromultimer." Thus, a "heteromultimer" is a molecule comprising at least a first polypeptide and a second polypeptide, wherein the second polypeptide differs in amino acid sequence from the first polypeptide by at least one amino acid residue. The heteromultimer can comprise a "heterodimer" formed by the first and second polypeptide or can form higher order tertiary structures where more than two polypeptides are present. Exemplary structures for the heteromultimer include heterodimers (e.g. Fv and Fab fragments, diabodies, GABA_B receptors 1 and 2 complexes), trimeric G-proteins, heterotetramers (e.g. F(ab')₂ fragments) and further oligomeric structures.

75 The "first recombinant polypeptide" of a chimeric heteromultimer refers to any polypeptide which is or was associated with a "second recombinant polypeptide" via the pairwise affinity of two dimerization sequences that are linked to the first and second polypeptides, respectively. Preferably, the first and second polypeptides contain sequences derived from a light or a heavy chain of an immunoglobulin. More preferably, the first and second polypeptides form a Nsc Abu that confers binding specificity to a desired antigen.

76 A "first heterodimerization sequence" refers to any dimerization sequence which is or was associated with a "second heterodimerization sequence," wherein the second heterodimerization sequence differs in amino acid sequence by at least one amino acid residue. A "heterodimerization pair" refers to two heterodimerization sequences capable of forming a heterodimer. 77 The term "antibody" as used herein refers to immunoglobulin molecules and imrnunologically active portions of immunoglobulin molecules, i.e., molecules that contain an antigen-binding site which specifically binds ("immunoreacts with") an antigen. Structurally, the simplest naturally occurring antibody (e.g., IgG) comprises four polypeptide chains, two heavy (H) chains and two light (L) chains inter-connected by disulfide bonds. The immunoglobulins represent a large family of molecules that include several types of molecules, such as IgD, IgG, IgA, IgM and IgE. The term "immunoglobulin molecule" includes, for example, hybrid antibodies, or altered antibodies, and fragments thereof. It has been shown that the antigen binding function of an antibody can be performed by fragments of a naturally-occurring antibody. These fragments are collectively termed "antigen-binding units" ("Abus"). Abus can be broadly divided into "single-chain" ("Sc") and "non-single-chain" ("Nsc") types based

. on their molecular structures.

78 Also encompassed within the terms "antibodies" and "Abus" are immunoglobulin molecules of a variety of species origins including invertebrates and vertebrates. The term "human" as applies to an antibody or an Abu refers to an immunoglobulin molecule expressed by a human gene or fragment thereof. The term "humanized" as applies to non-human (e.g. rodent or primate) antibodies refer to hybrid immunoglobulins, immunoglobulin chains or fragments thereof which contain minimal sequence derived from non-human immunoglobulin. For the most part, humanized antibodies are human immunoglobulins (recipient antibody) in which residues from a complementary determining region (CDR) of the recipient are replaced by residues from a CDR of a non-human species (donor antibody) such as mouse, rat, rabbit or primate having the desired specificity, affinity and capacity. In some instances, Fv framework region (FR) residues of the human immunoglobulin are replaced by corresponding non-human residues. Furthermore, the humanized antibody may comprise residues which are found neither in the recipient antibody nor in the imported CDR or framework sequences.

These modifications are made to further refine and optimize antibody performance and minimize immunogenicity when introduced into a human body. In general, the humanized antibody will comprise substantially all of at least one, and typically two, variable domains, in which all or substantially all of the CDR regions correspond to those of a non-human immunoglobulin and all or substantially all of the FR regions are those of a human immunoglobulin sequence. The humanized antibody may also comprise at least a portion of an immunoglobulin constant region (Fc), typically that of a human immunoglobulin.

79 "Non-single-chain antigen-binding unit" ("Nsc Abus") are heteromultimers comprising a light-chain polypeptide and a heavy-chain polypeptide. Examples of the Nsc Abus include but are not limited to (i) a ccFv fragment (Figure 1) stabilized by the heterodimerization sequences disclosed herein; (ii) any other monovalent and multivalent molecules comprising at least one ccFv fragment as described herein; (iii) a Fab fragment consisting of the VL, VH, CL and CHI domains; (iv) an Fd fragment consisting of the VH and CHI domains; (v) an Fv fragment consisting of the VL and VH domains of a single arm of an antibody; (vi) an F(ab')2 fragment, a bivalent fragment comprising two Fab fragments linked by a disulfide bridge at the hinge region; (vii) a diabody; and (viii) any other Nsc Abus that are described in Little et al. (2000) Immunology Today.

80 As noted above, a Nsc Abus can be either "monovalent" or "multivalent." Whereas the former has one binding site per antigen-binding unit, the latter contains multiple binding sites capable of binding to more than one antigen of the same or different kind. Depending on the number of binding sites, a Nsc Abus may be bivalent (having two antigen-binding sites), trivalent (having three antigen-binding sites), tetravalent (having four antigen-binding sites), and so on.

81 Multivalent Nsc Abus can be further classified on the basis of their binding specificities. A "monospecific" Nsc Abu is a molecule capable of binding to one or more antigens of the same kind. A "multispecific" Nsc Abu is a molecule having binding specificities for at least two different antigens. While such molecules normally will only bind two distinct antigens (i.e. bispecific Abus), antibodies with additional specificities such as trispecific antibodies are encompassed by this expression when used herein (see, e.g. Figures 15-17). Examples of bispecific antigen binding units include those with one arm directed against a tumor cell antigen and the other arm directed against a cytotoxic trigger molecule such as anti-FcγRI/anti-CD15, anti-pl85 ^HE /FcγRIII (CD16), anti-CD3/anti-malignant B-cell (1D10), anti-CD3/anti-ρl85 ^HER2, anti- CD3/anti-p97, anti-CD3/anti-renal cell carcinoma, anti-CD3/anti-OVCAR-3, anti- CD3/L-D1 (anti-colon carcinoma), anti-CD3/anti-melanocyte stimulating hormone analog, anti-EGF receρtor/anti-CD3, anti-CD3/anti-CAMAl, anti-CD3/anti-CD19, anti- CD3/MoV18, anti-neural cell ahesion molecule (NCAM)/anti-CD3, anti-folate binding protein (FBP)/anti-CD3, anti-pan carcinoma associated antigen (AMOC-31)/anti-CD3; bispecific Abus with one arm which binds specifically to a tumor antigen and one arm which binds to a toxin such as anti-saporin/anti-Id-1, anti-CD22/anti-saporin, anti- CD7/anti-saporin, anti-CD38/anti-saporin, anti-CEA/anti-ricin A chain, anti-interferon-α (IFN-α)/anti-hybridoma idiotype, anti-CEA/anti-vinca alkaloid; BsAbs for converting enzyme activated prodrugs such as anti-CD30/anti-alkaline phosphatase (which catalyzes conversion of mitomycin phosphate prodrug to mitomycin alcohol); bispecific Abus which can be used as fibrinolytic agents such as anti-fibrin/anti-tissue plasminogen activator (tPA), anti-fibrin/anti-urokinase-type plasminogen activator (uPA); bispecific antigen-binding untis for targeting immune complexes to cell surface receptors such as anti-low density lipoprotein (LDL)/anti-Fc receptor (e.g. Fcγ RI, FcγRII or FcγRIII); bispecific Abus for use in therapy of infectious diseases such as anti-CD3/anti-herpes simplex virus (HSV), anti-T-cell receptor: CD3 complex/anti-influenza, anti-FcγR/anti- HIV; bispecific Abus for tumor detection in vitro or in vivo such as anti-CEA/anti- EOTUBE, anti-CEA/anti-DPTA, anti-pl85 ^HER2 /anti-hapten; BsAbs as vaccine adjuvants (see Fanger et al, supra); and bispecific Abus as diagnostic tools such as anti- rabbit IgG/anti-ferritin, anti-horse radish peroxidase (HRP)/anti-hormone, anti- somatostatin/anti-substance P, anti-HRP/anti-FITC, anti-CEA/anti-.beta.-galactosidase (see Nolan et al, supra). Examples of trispecific antibodies include anti-CD3/anti- CD4/anti-CD37, anti-CD3/anti-CD5/anti-CD37 and anti-CD3/anti-CD8/anti-CD37. 82 "Single-chain antigen-binding unit" ("Sc Abu") refers to a monomeric Abu. Although the two domains of the Fv fragment are coded for by separate genes, a synthetic linker can be made that enables them to be made as a single protein chain (i.e. single chain Fv ("scFv") as described in Bird et al. (1988) Science 242:423-426 and Huston et al. (1988) PNAS 85:5879-5883) by recombinant methods. Other Sc Abus include antigen-binding molecules stabilized by the subject heterodimerization sequences (see e.g. Figure 18), and dAb fragments (Ward et al, (1989) Nature 341:544- 546 ) which consist of a VH domain and an isolated complimentarity determining region (CDR). An example of a linking peptide is (GGGGS)₃, which bridges approximately 3.5 nm between the carboxyl terminus of one V region and the amino terminus of another V region. Other linker sequences can also be used, and can provide additional functions, such as a means for attaching a drug or a solid support. A preferred single-chain antigen- binding unit contains VL and VH regions that are linked together and stabilized by a pair of subject heterodimerization sequences. The scFvs can be assembled in any order, for example, VH — (first heterodimerization sequence)-(second heterodimerization sequence) — VL or V_L — (first heterodimerization sequence)-(second heterodimerization sequence) — VH.

83 A "repertoire of antigen-binding units" refers to a plurality of antigen-binding units, at least two of which exhibit distinct binding specificities. A genetically diverse repertoire of antigen-binding units refers to a plurality of antigen-binding units, the majority and if not all of the antigen-binding units exhibit unique binding specificities with respect to each other. Genetically diverse repertoire typically has a complexity of at least 10⁶ to 10¹³, preferably between 10⁷to 10⁹, more preferably between 10⁸ to 10¹⁰, even more preferably between 10⁸ to 10¹¹ distinct antigen-binding units.

84 An antibody or Abu "specifically binds to" or "immunoreactive with" an antigen if it binds with greater affinity or avidity than it binds to other reference antigens including polypeptides or other substances.

85 An Abu is displayed "on the surface of a host cell" when the Abu is presented at the outer surface of a host cell. The displayed Abu may be directly attached to the outer surface of the host cell, or may be indirectly attached to the host cell via a host cell bound genetic package such as phage particle.

86 "Surface presenting sequences" refers to sequences that facilitate display of heterologous sequences. Typically, the surface presenting sequences are present on the outer surface of a genetic package, e.g. phage or bacteria. Preferred surface presenting sequences of phage is pill of Ml 3 filamentous phage.

87 "Antigen" as used herein means a substance that is recognized and bound specifically by an antibody. Antigens can include peptides, proteins, glycoproteins, polysaccharides and lipids; portions thereof and combinations thereof.

88 As used herein, the term "surface antigens" refers to the plasma membrane components of a cell. It encompasses integral and peripheral membrane proteins, glycoproteins, polysaccharides and lipids that constitute the plasma membrane. An "integral membrane protein" is a transmembrane protein that extends across the lipid bilayer of the plasma membrane of a cell. A typical integral membrane protein consists of at least one "membrane spanning segment" that generally comprises hydrophobic amino acid residues. Peripheral membrane proteins do not extend into the hydrophobic interior of the lipid bilayer and they are bound to the membrane surface by noncovalent interaction with other membrane proteins.

89 The terms "membrane", "cytosolic", "nuclear" and "secreted" as applied to cellular proteins specify the extracellular and/or subcellular location in which the cellular protein is mostly, predominantly, or preferentially localized.

90 "Cell surface receptors" represent a subset of membrane proteins, capable of binding to their respective ligands. Cell surface receptors are molecules anchored on or inserted into the cell plasma membrane. They constitute a large family of proteins, glycoproteins, polysaccharides and lipids, which serve not only as structural constituents of the plasma membrane, but also as regulatory elements governing a variety of biological functions.

91 A "heterodimeric receptor" encompasses cellular proteins composed of two proteinaceous subunits which exhibits binding affinity to a ligand. The two proteinaceous subunits are distinct molecules which differ in amino acid sequence by at least one amino acid residue. Non-limiting illustrative heterodimeric receptors are those that bind to growth factors (e.g. heregulin), neurotransmitters (e.g. γ-Aminobutyric acid), and other organic or inorganic small molecules (e.g. mineralocorticoid, glucocorticoid). Preferred heterodimeric receptors are nuclear hormone receptors (Belshaw et al. (1996) Proc. Natl. Acad. Sci. U S. A 93(10):4604-4607), erbB3 and erbB2 receptor complex, and G-protein-coupled receptors including but not limited to opioid (Gomes et al. (2000) J. Neuroscience 20(22): RC110); Jordan et al. (1999) Nature 399:697-700), muscarinic, dopamine, serotonin, adenosine/dopamine, and GABA_B families of receptors.

92 "Domain" refers to a portion of a protein that is physically or functionally distinguished from other portions of the protein or peptide. Physically-defined domains include those amino acid sequences that are exceptionally hydrophobic or hydrophilic, such as those sequences that are membrane-associated or cytoplasm-associated. Domains may also be defined by internal homologies that arise, for example, from gene duplication. Functionally-defined domains have a distinct biological function(s). The ligand-binding domain of a receptor, for example, is that domain that binds ligand. An antigen-binding domain refers to the part of an antigen-binding unit or an antibody that binds to the antigen. Functionally-defined domains need not be encoded by contiguous amino acid sequences. Functionally-defined domains may contain one or more physically-defined domain. Receptors, for example, are generally divided into the extracellular ligand-binding domain, a transmembrane domain, and an intracellular effector domain. A "membrane anchorage domain" refers to the portion of a protein that mediates membrane association. Generally, the membrane anchorage domain is composed of hydrophobic amino acid residues. Alternatively, the membrane anchorage domain may contain modified amino acids, e.g. amino acids that are attached to a fatty acid chain, which in turn anchors the protein to a membrane.

93 A "host cell" includes an individual cell or cell culture which can be or has been a recipient for the subject vectors. Host cells include progeny of a single host cell. The progeny may not necessarily be completely identical (in morphology or in genomic of total DNA complement) to the original parent cell due to natural, accidental, or deliberate mutation. A host cell includes cells transfected in vivo with a vector of this invention.

94 A "cell line" or "cell culture" denotes bacterial, plant, insect or higher eukaryotic cells grown or maintained in vitro. The descendants of a cell may not be completely identical (either morphologically, genotypically, or phenotypically) to the parent cell.

95 A "defined medium" refers to a medium comprising nutritional and hormonal requirements necessary for the survival and/or growth of the cells in culture such that the components of the medium are known. Traditionally, the defined medium has been formulated by the addition of nutritional and growth factors necessary for growth and/or survival. Typically, the defined medium provides at least one component from one or more of the following categories: a) all essential amino acids, and usually the basic set of twenty amino acids plus cysteine; b) an energy source, usually in the form of a carbohydrate such as glucose; c) vitamins and/or other organic compounds required at low concentrations; d) free fatty acids; and e) trace elements, where trace elements are defined as inorganic compounds or naturally occurring elements that are typically required at very low concentrations, usually in the micromolar range. The defined medium may also optionally be supplemented with one or more components from any of the following categories: a) one or more mitogenic agents; b) salts and buffers as, for example, calcium, magnesium, and phosphate; c) nucleosides and bases such as, for example, adenosine and thymidine, hypoxanthine; and d) protein and tissue hydrolysates.

96 As used herein, the term "isolated" means separated from constituents, cellular and otherwise, in which the polynucleotide, peptide, polypeptide, protein, antibody, or fragments thereof, are normally associated with in nature. As is apparent to those of skill in the art, a non-naturally occurring the polynucleotide, peptide, polypeptide, protein, antibody, or fragments thereof, does not require "isolation" to distinguish it from its naturally occurring counterpart. In addition, a "concentrated", "separated" or "diluted" polynucleotide, peptide, polypeptide, protein, antibody, or fragments thereof, is distinguishable from its naturally occurring counterpart in that the concentration or number of molecules per volume is greater than "concentrated" or less than "separated" than that of its naturally occurring counterpart.

97 Enrichment can be measured on an absolute basis, such as weight per volume of solution, or it can be measured in relation to a second, potentially interfering substance present in the source mixture. Increasing enrichments of the embodiments of this invention are increasingly more preferred. Thus, for example, a 2-fold enrichment is preferred, 10-fold enrichment is more preferred, 100-fold enrichment is more preferred, 1000-fold enrichment is even more preferred. A substance can also be provided in an isolated state by a process of artificial assembly, such as by chemical synthesis or recombinant expression.

98 "Linked" and "fused" or "fusion" are used interchangeably herein. These terms refer to the joining together of two more chemical elements or components, by whatever means including chemical conjugation or recombinant means. An "in-frame fusion" refers to the joining of two or more open reading frames (OFRs) to form a continuous longer OFR, in a manner that maintains the correct reading frame of the original OFRs. Thus, the resulting recombinant fusion protein is a single protein containing two ore more segments that correspond to polypeptides encoded by the original OFRs (which segments are not normally so joined in nature.) Although the reading frame is thus made continuous throughout the fused segments, the segments may be physically or spatially separated by, for example, in-frame linker sequence (e.g. "flexon"), as described infra.

99 In the context of polypeptides, a "linear sequence" or a "sequence" is an order of amino acids in a polypeptide in an amino to carboxyl terminus direction in which residues that neighbor each other in the sequence are contiguous in the primary structure of the polypeptide. A "partial sequence" is a linear sequence of part of a polypeptide which is known to comprise additional residues in one or both directions.

100 "Heterologous" means derived from a genotypically distinct entity from the rest of the entity to which it is being compared. For example, a promoter removed from its native coding sequence and operatively linked to a coding sequence other than the native sequence is a heterologous promoter. The term "heterologous" as applied to a polynucleotide, a polypeptide, means that the polynucleotide or polypeptide is derived from a genotypically distinct entity from that of the rest of the entity to which it is being compared. For instance, a heterologous polynucleotide or antigen may be derived from a different species origin, different cell type, and the same type of cell of distinct individuals.

101 The terms "polynucleotides", "nucleic acids", "nucleotides" and "oligonucleotides" are used interchangeably. They refer to a polymeric form of nucleotides of any length, either deoxyribonucleotides or ribonucleotides, or analogs thereof. Polynucleotides may have any three-dimensional structure, and may perform any function, known or unknown. The following are non-limiting examples of polynucleotides: coding or non-coding regions of a gene or gene fragment, loci (locus) defined from linkage analysis, exons, introns, messenger RNA (mRNA), transfer RNA, ribosomal RNA, ribozymes, cDNA, recombinant polynucleotides, branched polynucleotides, plasmids, vectors, isolated DNA of any sequence, isolated RNA of any sequence, nucleic acid probes, and primers. A polynucleotide may comprise modified nucleotides, such as methylated nucleotides and nucleotide analogs. If present, modifications to the nucleotide structure may be imparted before or after assembly of the polymer. The sequence of nucleotides may be interrupted by non-nucleotide components. A polynucleotide may be further modified after polymerization, such as by conjugation with a labeling component.

102 "Recombinant" as applied to a polynucleotide means that the polynucleotide is the product of various combinations of cloning, restriction and/or ligation steps, and other procedures that result in a construct that is distinct from a polynucleotide found in nature.

103 The terms "gene" or "gene fragment" are used interchangeably herein. They refer to a polynucleotide containing at least one open reading frame that is capable of encoding a particular protein after being transcribed and translated. A gene or gene fragment may be genomic or cDNA, as long as the polynucleotide contains at least one open reading frame, which may cover the entire coding region or a segment thereof.

104 "Operably linked" or "operatively linked" refers to a juxtaposition wherein the components so described are in a relationship permitting them to function in their intended manner. For instance, a promoter sequence is operably linked to a coding sequence if the promoter sequence promotes transcription of the coding sequence.

105 A "fusion gene" is a gene composed of at least two heterologous polynucleotides that are linked together.

106 A gene "database" denotes a set of stored data which represent a collection of sequences including nucleotide and peptide sequences, which in turn represent a collection of biological reference materials.

107 As used herein, "expression" refers to the process by which a polynucleotide is transcribed into mRNA and/or the process by which the transcribed mRNA (also referred to as "transcript") is subsequently being translated into peptides, polypeptides, or proteins. The transcripts and the encoded polypeptides are collectively referred to as gene product. If the polynucleotide is derived from genomic DNA, expression may include splicing of the mRNA in a eukaryotic cell.

108 A "subject" as used herein refers to a biological entity containing expressed genetic materials. The biological entity is preferably plant, animal, or microorganisms including bacteria, viruses, fungi, and protozoa. Tissues, cells and their progeny of a biological entity obtained in vivo or cultured in vitro are also encompassed.

109 A "vector" is a nucleic acid molecule, preferably self-replicating, which transfers an inserted nucleic acid molecule into and/or between host cells. The term includes vectors that function primarily for insertion of DNA or RNA into a cell, replication of vectors that function primarily for the replication of DNA or RNA, and expression vectors that function for transcription and/or translation of the DNA or RNA. Also included are vectors that provide more than one of the above functions.

110 An "expression vector" is a polynucleotide which, when introduced into an appropriate host cell, can be transcribed and translated into a polypeptide(s). An "expression system" usually connotes a suitable host cell comprised of an expression vector that can function to yield a desired expression product.

111 A "replicon" refers to a polynucleotide comprising an origin of replication

(generally referred to as an ori sequence) which allows for replication of the polynucleotide in an appropriate host cell. Examples of replicons include episomes (such as plasmids), as well as chromosomes (such as the nuclear or mitochondrial chromosomes).

Chimeric Heteromultimer of the Present Invention

112 As noted above, proper assembly of polypeptide subunits to form a stable complex is required to ensure the biological function of a multimeric protein. Accordingly, a central aspect of the present invention is the design of a technique that enables specific assembly of selected monomeric polypeptides to effect efficient production of heteromultimers. The experimental design is particularly useful for generating and screening for heteromultimers such as Abus whose binding specificities depend on the assembly of specific subunits in a specific manner. Distinguished from the previously reported chimeric Abus, the subject Abus have one or more of the following unique features. First, the Abus are reconstititued via pairwise affinity of two heterodimerization sequences, at least one of which and preferably both of which, lack(s) detectable propensity to form homodimers. Unlike the previously reported dimerization sequences such as Fos and Jun leucine zippers that are known to form homodimers under both physiological buffer conditions and physiological body temperature (O'Shea et al. (1992) Cell 68: 699-708; Vidal et al. (1996) Proc. Natl. Acad. Sci. U.S.A), the subject heterodimerization sequences are essentially incapable of forming homodimers either under the specified buffer conditions and/or at the specified body temperatures. The subject heterodimerization sequences may be also distinguished from the previously employed sequences at the structural level as detailed below.

113 In one embodiment, the present invention provides a chimeric heteromultimer displayed on the surface of the host cell, wherein heteromultimer comprises: (i) a first polypeptide fused to a first heterodimerization sequence and a surface presenting sequence; (ii) a second polypeptide fused to a second heterodimerization sequence; wherein the first and second polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures.

114 In another embodiment, the present invention provides a Nsc Abu that comprises: (a) a light-chain polypeptide comprising a light-chain variable region linked to a first heterodimerization sequence; (b) a heavy-chain polypeptide comprising a heavy-chain variable region linked to a second heterodimerization sequence; wherein the light-chain and the heavy-chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences, at least one of which is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures. In another aspect, the present invention provides a Nsc Abu whose the light-chain and the heavy-chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences that are derived from heterodimeric receptors. In one aspect, the first and second heterodimerization sequences comprising heterodimerization receptor sequences that mediate heterodimerization of the receptors.

115 In a separate embodiment, the present invention provides a Sc Abu which comprises a light-chain variable region and a heavy-chain variable region connected by a first and a second heterodimerization sequence spanning the distance between the C- terminus of one of the region to the N-terminus of the other region, wherein the two regions form an intra-molecular dimer via pairwise affinity of the first and second heterodimerization sequences, at least one of which is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures. In another aspect of this embodiment, the present invention provides a Sc Abu wherein the light-chain variable region and the heavy-chain variable region form an intra-molecular dimer via pairwise affinity of two heterodimerization sequences that are derived from heterodimeric receptors. In one aspect, the first and second heterodimerization sequences comprising heterodimerization receptor sequences that mediate heterodimerization of the receptors. Selection of Heterodimerization Sequences:

116 Several factors apply to the design of Abus having one or more of the above- mentioned characteristics. First, the heterodimerization sequences must exhibit pairwise affinity to effect formation of a stable complex. By "stable" is meant that the complex or dimer is sufficiently long-lasting to persist between the formation of the complex or dimer, and its subsequent detection and/or purification. The complex or dimer must be able to withstand whatever conditions exist or are introduced between the moment of formation and the moment of detection, these conditions being a function of the assay or reaction which is being performed. Preferably, the formation of the complex or dimer is carried out under physiological buffer conditions and at physiological body temperatures ranging from approximately room temperature to approximately 37°C. Intervening conditions which may optionally be present and which may dislodge a complex or dimer include washing, heating, adding additional solutes or solvents to the reaction mixture (such as denaturants), and competing with additional reacting species. Stable complex or dimer may be irreversible or reversible, but must meet the other requirements of this definition. Thus, a transient complex or dimer may form in a reaction mixture, but it does not constitute a stable complex if it dissociates spontaneously under physiological buffer conditions or as a result of a newly imposed condition or manipulation introduced before detection.

117 Second, the selected heterodimerization sequences must exhibit pairwise affinity resulting in predominant formation of heterodimers to a substantial exclusion of homodimers. Preferably, the predominant formation yields a heteromultimeric pool that contains at least 60% heterodimers, more preferably at least 80% heterodimers, more preferably between 85-90% heterodimers, and more preferably between 90-95% heterodimers, and even more preferably between 96-99% heterodimers that are allowed to form under physiological buffer conditions and/or physiological body temperatures.

In certain embodiments of the present invention, at least one of the heterodimerization sequences employed to reconstitute an Abu is essentially incapable of forming a homodimer in a physiological buffer and/or at physiological body temperature. By "essentially incapable" is meant that the selected heterodimerization sequences when tested alone do not yield detectable amounts of homodimers in an in vitro sedimentation experiment as detailed in Kammerer et al. (1999) Biochemistry 38: 13263-13269), or in the in vivo two-hybrid yeast analysis (see e.g. White et al. Nature (1998) 396: 679-682). Specifically, Kammerer et al. have demonstrated by sedimentation experiments that the heterodimerization sequences of GABAB receptor 1 and 2, when tested alone, sediment at the molecular mass of the monomer under physiological conditions and at physiological body temperatures (e.g. at 37°C). When mixed in equimolar amounts, GABA_B receptor 1 and 2 heterodimerization sequences sediment at the molecular mass corresponding to the heterodimer of the two sequences (see Table 1 of Kammerer et al.). In addition, individual heterodimerization sequences can be expressed in a host cell and the absence of homodimers in the host cell can be demonstrated by a variety of protein analyses including but not limited to SDS-PAGE, Western blot, and immunoprecipitation. The in vitro assays must be conducted under a physiological buffer conditions, and/or preferably at physiological body temperatures. Generally, a physiological buffer contains a physiological concentration of salt and at adjusted to a neutral pH ranging from about 6.5 to about 7.8, and preferably from about 7.0 to about 7.5. A variety of physiological buffers is listed in Sambrook et al. (1989) supra and hence is not detailed herein. Preferred physiological conditions are described in Kammerer et al, supra.

118 The specific association of heterodimerization sequences typically involves noncovalent interactions. Such interactions encompass every exiting stable linkage that does not result in the formation of a covalent bond. Non-limiting examples of noncovalent interactions include electrostatic bonds, hydrogen bonding, Van der Waal's forces, steric interdigitation of amphiphilic peptides.

119 A further consideration in designing the subject Abu is to minimize any structural interference between the heterodimerization sequences and the antigen-binding site of the resulting heteromultimer. A variety of techniques is available in the art for designing a chimeric heteromultimer with minimal internal structural interference. For instance, one approach involves the use of minimal heterodimerization sequences containing only amino acid residues that are required for heterodimerization. The second approach is to link the heterodimerization sequences to either N-terminus or C-terminus of the resulting heteromultimer. The choice of either terminus will depend on the location of the biological active domain of the heteromultimer. For constructing a chimeric Abu whose antigen-binding site resides in the N-terminal half of the light and heavy chain variable regions, it is preferable to link the heterodimerization sequences to the C-terminus of a light or heavy chain. Another alternative design employs a "flexon" incorporated between the antigen-binding site and the heterodimerization sequence of the heteromultimer. "Flexon" as used herein, refers to a flexible polypeptide linker (or a nucleic acid sequence encoding such a polypeptide) which typically comprises amino acids having small side chains (e.g. glycine, alanine, valine, leucine, isoleucine, and serine). Incorporating flexons between one or more sites of the subject Abu is believed to promote functionality by allowing them to assume a conformations relatively independent of each other. Such a construction generally provides additional flexibility to the antigen-binding domain. Suitable flexons preferably comprise between about four and about one hundred amino acids, more preferably about four to fifty amino acids, and even more preferably about four to fifteen amino acids.

120 Heterodimerization sequences applicable for constructing the subject Abus can be derived from a variety of sources. Generally, any protein sequences involved in the formation of stable heteromultimers are candidate heterodimerization sequences. As such, these sequences may be derived from any heteromultimeric protein complexes. Representative candidate sequences are viral proteins such as the capsid proteins of adeno-associated viruses, protein kinase phosphorylation sites that interact with SH2- domain-containing proteins (Cantely et al. (1993) Cell 72: 767-778; Cantely et al. (1995) J Biol. Chem.270(44): 26029-26032), domains of transcription factors and heterodimeric receptors, which mediate heteromer formation.

121 Preferred heteromultimeric transcription factors are a-Pal/ Max complexes and

Hox/Pbx complexes. Hox represents a large family of transcription factors involved in patterning the anterior-posterior axis during embryogenesis. Hox proteins bind DNA with a conserved three alpha helix homeodomain. In order to bind to specific DNA sequences, Hox proteins require the presence of hetero-partners such as the Pbx homeodomain. Wolberger et al. solved the 2.35A crystal structure of a HoxBl-Pbxl- DNA ternary complex in order to understand how Hox-Pbx complex formation occurs and how this complex binds to DNA. The structure shows that the homeodomain of each protein binds to adjacent recognition sequences on opposite sides of the DNA. Heterodimerization occurs through contacts formed between a six amino acid hexapeptide N-terminal to the homeodomain of HoxBl and a pocket in Pb l formed between helix 3 and helices 1 and 2. A C-terminal extension of the Pbxl homeodomain forms an alpha helix that packs against helix 1 to form a larger four helix homeodomain (Wolberger et al. (1999) Cell 96: 587-597; Wolberger et al. J Mol Biol 291: 521-530).

122 A vast number of heterodimeric receptors have also been identified. They include but are not limited to those that bind to growth factors (e.g. heregulin), neurotransmitters (e.g. γ-Aminobutyric acid), and other organic or inorganic small molecules (e.g. mineralocorticoid, glucocorticoid). Preferred heterodimeric receptors are nuclear hormone receptors (Belshaw et al. (1996) Proc. Natl. Acad. Sci. U S. A 93(10):4604-4607), erbB3 and erbB2 receptor complex, and G-protein-coupled receptors including but not limited to opioid (Gomes et al. (2000) J. Neuroscience 20(22):

RC110); Jordan et al. (1999) Nature 399:697-700), muscarinic, dopamine, serotonin, adenosine/dopamine, and GABA_B families of receptors. For majority of the known heterodimeric receptors, their C-terminal sequences are found to mediate heterodimer formation.

123 Where desired, sequences from novel heterodimeric receptors can be employed in constructing the subject Abus. In such situation, the identification of a candidate heterodimerization sequences in a given receptor pair can be determined by any genetic or biochemical assays without undue experimentation. Additionally, computer modeling and searching technologies further facilitates detection of heterodimerization sequences based on sequence homologies of common domains appeared in related and unrelated genes. Non-limiting examples of programs that allow homology searches are Blast (http://www.ncbi.nlm.nih.gov/BLAST/), Fasta (Genetics Computing Group package, Madison, Wisconsin), DNA Star, Clustlaw, TOFFEE, COBLATH, Genthreader, and MegAlign. Any sequence databases that contains DNA sequences corresponding to a target receptor or a segment thereof can be used for sequence analysis. Commonly employed databases include but are not limited to GenBank, EMBL, DDB J, PDB, SWISS-PROT, EST, STS, GSS, and HTGS. Another preferred class of heterodimerization sequences consists of amphiphilic peptides that adopt a coiled-coil helical structure. The helical coiled coil is one of the principal subunit oligomerization sequences in proteins. Primary sequence analysis reveals that approximately 2-3% of all protein residues form coiled coils (Wolf et al. ( 1997) Protein Sci. 6:1179-1189). Well-characterized coiled-coil-containing proteins include members of the cytoskeletal family (e.g. α-keratin, vimentin), cytoskeletal motor family (e.g. myosine, kinesins, and dyneins), viral membrane proteins (e.g. membrane proteins of Ebola or HIV), DNA binding proteins, and cell surface receptors (e.g. GABA_B receptors 1 and 2). Coiled-coil heterodimerization sequences of the present invention can be broadly classified into two groups, namely the left-handed and right- handed coiled coils. The left-handed coiled coils are characterized by a heptad repeat denoted "abcdefg" with the occurrence of apolar residues preferentially located at the first (a) and fourth (d) position. The residues at these two positions typically constitute a zig-zag pattern of "knobs and holes" that interlock with those of the other stand to form a tight-fitting hydrophobic core. In contrast, the second (b), third (c) and sixth (f) positions that cover the periphery of the coiled coil are preferably charged residues. Examples of charged amino acids include basic residues such as lysine, arginine, histidine, and acidic residues such as aspartate, glutamate, asparagine, and glutamine. Uncharged or apolar amino acids suitable for designing a heterodimeric coiled coil include but are not limited to glycine, alanine, valine, leucine, isoleucine, serine and threonine. While the uncharged residues typically form the hydrophobic core, inter-helical and intra-helical salt-bridge including charged residues even at core positions may be employed to stabilize the overall helical coiled-coiled structure (Burkhard et al. (2000) J. Biol. Chem. 275:11672-11677). Whereas varying lengths of coiled coil may be employed, the subject heterodimerization sequences preferably contain two to ten heptad repeats. More preferably, the heterodimerization sequences contain three to eight heptad repeats, even more preferably contain four to five heptad repeats.

125 In designing optimal coiled-coil heterodimerization sequences, a variety of existing computer software programs that predict the secondary structure of a peptide can be used. An illustrative computer analysis uses the COILS algorithm which compares an amino acid sequence with sequences in the database of known two-stranded coiled coils, and predicts the high probability coiled-coil stretches (Kammerer et al(l999) Biochemistry 38:13263-13269).

126 While a diverse variety of coiled coils involved in hetero-oligomerization can be employed in the subject invention, preferred coiled coils are derived from heterodimeric receptors. Accordingly, the present invention encompasses the coiled-coil dimeric sequences derived from GABA_B receptors 1 and 2. In one aspect, the subject coiled coils comprise the C-terminal sequences of GABA_B receptor 1 and GABAB receptor 2. In another aspect, the subject coiled coils are further linked to cysteine residues. The coiled coils are GABAB receptor 1 and 2 polypeptides of at least 30 amino acid residues, one of which is essentially identical to a linear sequence of comparable length depicted in SEQ ID NO. 2, and the other is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4.

127 A linear sequence of peptide is "essentially identical" to another linear sequence, if both sequences exhibit substantial amino acid sequence homology. Generally, essentially identical sequences are at least about 60% identical with each other, after alignment of the homologous regions. Preferably, the sequences are at least about 70% identical; more preferably, they are at least about 80% identical; more preferably, they are at least about 90% identical; more preferably, the sequences are at least about 95% identical; still more preferably, the sequences are 100% identical. 128 In determining whether polypeptide sequences are essentially identical, a sequence that preserves the functionality of the polypeptide with which it is being compared is particularly preferred. Functionality may be established by different criteria, such as ability to form a heterodimer with a pairing coiled-coil sequence, and inability to form a homodimer under physiological buffer conditions and/or physiological body temperatures.

129 The invention includes modified GABA_B heterodimerization sequences which are functionally equivalent to the sequences exemplified herein. Modified polypeptides providing improved stability to the resulting Abus are preferred. Examples of modified polypeptides include those with conservative substitutions of amino acid residues, and one or more deletions or additions of amino acids which do not significantly deleteriously alter the heterodimerization specificity. Substitutions can range from changing or modifying one or more amino acid residues to complete redesign of a region as long as the pairwise affinity is maintained. Amino acid substitutions, if present, are preferably conservative substitutions that do not deleteriously affect folding or functional properties of the peptide. Groups of functionally related amino acids within which conservative substitutions can be made are glycine/alanine; valine/isoleucine/leucine; asparagine/glutamine; aspartic acid/glutamic acid; serine/threonine/methionine; lysine/arginine; and phenylalanine/tryosine/tryptophan. Polypeptides of this invention can be in glycosylated or unglycosylated form, can be modified post-translationally (e.g., acetylation, and phosphorylation) or can be modified synthetically (e.g., the attachment of a labeling group).

Configurations and Modifications of Antigen-Binding Units (Abus):

130 The Abus of the present invention can adopt a variety of configurations. The smallest non-single chain Abu is a monovalent ccFv fragment. The ccFv fragment is a dimeric protein composed of VL and VH regions, which dimerize via the pairwise affinity of the first and second heterodimerization sequences fused in-frame with the VL and VH regions, respectively. Preferably, the ccFv contains a short flexon sequence that provides additional flexibility to the VL and VH regions (see an exemplary ccFv in Figure 1). A more complex Nsc Abu is a multivalent molecule capable of binding to more than one antigen of the same kind (i.e. multivalent but monospecific) or different kind (i.e. multivalent and multispecific Abus). Typically, a multivalent Abu is a heteromultimer composed of more than one L and H chain polypeptides, in which either the L or H polypeptide or both contain more than one V region. For instance, an exemplary bivalent Abus assumes the configuration of (ccFv)₂ as depicted in Figure 12. The H chain polypeptide in this illustrative bivalent Abus contains two VH regions, each of which dimerizes with a VL region to constitute two antigen-binding sites. Alternatively, the L chain polypeptide may provide two VL regions, each of which dimerizes with a VH region to reconstitute the two binding sites. As shown in Figure 12, the multivalent Abu is stabilized via pairwise affinity of the two heterodimerization sequences linked to the VL and VH regions. The Abu is assembled efficiently because at least one, and preferably both, of the heterodimerization sequences is or are incapable of forming homodimers, thus minimizing intra-molecular dimerization to form nonfunctional VH/VH or VL/VL dimers. Applying this general antibody engineering scheme, trivalent and tetravalent Abus can be constructed (see, e.g. Figure 12).

131 A variant approach to construct multivalent Abus employs a scFv or dsFv fragment as illustrated in Figure 13. In addition to the building unit ccFv that provides one antigen-binding site, Abus of this configuration contains one or more scFv or dsFv fragments that are linked to the ccFv. The linked scFv or dsFv provides the additional binding sites. For instance, a bivalent Abus may adopt the ccFv-scFv or ccFv-dsFv configuration (Figure 13). Whereas one of the antigen-binding site is assembled via the pairwise affinity of the heterodimerization sequences linked to the VL region and VH region (as in ccFv), the other is provided by the scFv or dsFv fragment that is fused in- frame with the VL region. Alternatively, the scFv or dsFv fragment can be linked to the VH region.

132 The same approach can be employed to generate trivalent ccFv-scFv or ccFv- dsFv Abus as shown in Figure 14. In one aspect, the trivalent Abus assumes the configuration of ccFv-(scFv)₂, in which two polypeptides "VH-first heterodimerization sequence-scFv" and "VL-second heterodimerization sequence-scFv" dimerize via the pairwise affinity of the two heterodimerization sequences to constitute three binding sites. One of the binding sites is composed of the VL and VH regions of the ccFv building units; the remaining two are provided by the scFv fragments linked to the respective VL an VH polypeptides. Alternatively, the multivalent Abus can be configured as ccFv-scFv-dsFv. In this configuration, one of the antigen-binding sites is assembled and stabilized via an inter-molecular disulfide bond between a pair of cysteine residues that is located within the VH and VL regions of the dsFv fragment. A further variant of this configuration is a trivalent ccFv-(dsFv)₂, in which two of the binding sites assume the dsFv format (see, e.g. Figure 14). Any other multivalent Abus variants employing the basic ccFv building unit, whether being monospecific or multispecifc, are encompassed by this invention.

133 Accordingly, this invention further provides multispecific Abus. They are multivalent molecules capable of binding to at least two distinct antigens. Preferred multispecific Abus are bispecific and trispecific molecules exhibiting binding specificities to two and three distinct antigens, respectively. Distinguished from previously characterized multispecific antibodies (see e.g. U.S. Patent No. 5,932,448), the subject multispecific Abus comprises one or more ccFv building unit with distinct binding specificities. The subject multispecific Abus can also incorporate one or more scFv or dsFv fragments as detailed above. Preferred bispecific and trispecific Abus are configured according to the general structures depicted in Figures 15-17.

134 Aside from non-single chain Abus, the present invention includes single-chain

Abus that is stabilized by the subject heterodimerization sequences. Typically, the Sc

Abus comprises a VL and a VH region forming an intra-molecular dimer via the pairwise affinity of the heterodimerization sequences connected to these two regions.

The heterodimerization sequences may be configured in either a parallel or anti-parallel manner (see, e.g. Figure 18). In a parallel configuration, the two heterodimerzation sequences are aligned such that they have the same orientation (amino-terminal to carboxyl-terminal). In an anti-parallel configuration, the heterodimerization sequences are arranged such that the amino-terminal end of one sequence is aligned with the carboxyl-terminal end of the other sequence, and vice versa. Generally, the heterodimerization sequences are linked together via a flexon sequence. As described herein, flexon is a flexible polypeptide linker (or a nucleic acid sequence encoding such a polypeptide) which typically comprises amino acids having small side chains (e.g. glycine, alanine, valine, leucine, isoleucine, and serine). Incorporating flexons between the two heterodimerization sequences generally provides spatial flexibility for them to form an intra-molecular dimer. Suitable flexons for the anti-parallel configuration preferably comprise between about four to about one hundred amino acids, more preferably about four to fifty amino acids, and even more preferably about four to fifteen amino acids. Flexons for the parallel configuration are generally longer, preferably ranging from about ten to about one hundred amino acids, more preferably from about fifty to about thirty amino acid residues.

135 Where desired, one or more pairs of cysteine residues may be incorporated at the N- or C- terminus of the heterodimerization sequences to further stabilize the Abus of the present invention.

136 The Abus of this invention may contain sequences derived from the constant regions of an L chain or a H chain. Such sequences derived from the constant regions are generally placed between a light-chain or a heavy-chain variable region and the heterodimerization sequence to which it is linked. In addition, the light and heavy chains may contain partly or entirely human sequences.

137 Methods for humanizing non-human antibodies are well known in the art.

"Humanized" antibodies are antibodies in which at least part of the sequence has been altered from its initial form to render it more like human immunoglobulins. In one version, the H chain and L chain C regions are replaced with human sequence. This is a fusion polypeptide comprising a V region and a heterologous immunoglobulin C region.

In another version, the CDR regions comprise non human antibody sequences, while the

V framework regions have also been converted human sequences. See, for example, EP

0329400. In a third version, V regions are humanized by designing consensus sequences of human and mouse V regions, and converting residues outside the CDRs that are different between the consensus sequences.

138 In making humanized antibodies, the choice of framework residues can be critical in retaining high binding affinity. In principle, a framework sequence from any HuAb can serve as the template for CDR grafting; however, it has been demonstrated that straight CDR replacement into such a framework can lead to significant loss of binding affinity to the antigen. Glaser et al. (1992) J. Immunol. 149:2606; Tempest et al. (1992) Biotechnology 9:266; and Shalaby et al. (1992) J. Exp. Med. 17:217. The more homologous a HuAb is to the original muAb, the less likely that the human framework will introduce distortions into the murine CDRs that could reduce affinity. Based on a sequence homology search against an antibody sequence database, the HuAb IC4 provides good framework homology to muM4TS.22, although other highly homologous HuAbs would be suitable as well, especially kappa L chains from human subgroup I or H chains from human subgroup III. Kabat et al. (1987). Various computer programs such as ENCAD (Levitt et al. (1983) J. Mol. Biol. 168:595) are available to predict the ideal sequence for the V region. The invention thus encompasses HuAbs with different V regions. It is within the skill of one in the art to determine suitable V region sequences and to optimize these sequences. Methods for obtaining antibodies with reduced immunogenicity are also described in U.S. Patent No. 5,270,202 and EP 699,755.

139 It is important that antibodies be humanized with retention of high affinity for the antigen and other favorable biological properties. To achieve this goal, according to a preferred method, humanized antibodies are prepared by a process of analysis of the parental sequences and various conceptual humanized products using three dimensional models of the parental and humanized sequences. Three dimensional immunoglobulin models are familiar to those skilled in the art. Computer programs are available which illustrate and display probable three-dimensional conformational structures of selected candidate immunoglobulin sequences. Inspection of these displays permits analysis of the likely role of the residues in the functioning of the candidate immunoglobulin sequence, i.e., the analysis of residues that influence the ability of the candidate immunoglobulin to bind its antigen. In this way, FR residues can be selected and combined from the consensus and import sequence so that the desired antibody characteristic, such as increased affinity for the target antigen(s), is achieved.

140 The invention also encompasses Abus conjugated to a chemically functional moiety. Typically, the moiety is a label capable of producing a detectable signal. These conjugated Abus are useful, for example, in detection systems such as quantitation of tumor burden, and imaging of metastatic foci and tumor imaging. Such labels are known in the art and include, but are not limited to, radioisotopes, enzymes, fluorescent compounds, chemiluminescent compounds, bioluminescent compounds substrate cofactors and inhibitors. See, for examples of patents teaching the use of such labels, U.S. Patent Nos. 3,817,837; 3,850,752; 3,939,350; 3,996,345; 4,277,437; 4,275,149; and 4,366,241. The moieties can be covalently linked to Abus, recombinantly linked, or conjugated to Abus through a secondary reagent, such as a second antibody, protein A, or a biotin-avidin complex.

141 Other functional moieties include signal peptides, agents that enhance immunologic reactivity, agents that facilitate coupling to a solid support, vaccine carriers, bioresponse modifiers, paramagnetic labels and drugs. Signal peptides is a short amino acid sequence that directs a newly synthesized protein through a cellular membrane, usually the endoplasmic reticulum in eukaryotic cells, and either the inner membrane or both inner and outer membranes of bacteria. Signal peptides are typically at the N-terminal portion of a polypeptide and are typically removed enzymatically between biosynthesis and secretion of the polypeptide from the cell. Such a peptide can be incorporated into the subject Abus to allow secretion of the synthesized molecules.

142 Agents that enhance immunologic reactivity include, but are not limited to, bacterial superantigens. Agents that facilitate coupling to a solid support include, but are not limited to, biotin or avidin. Immunogen carriers include, but are not limited to, any physiologically acceptable buffers. Bioresponse modifiers include cytokines, particularly tumor necrosis factor (TΝF), interleukin-2, interleukin-4, granulocyte macrophage colony stimulating factor and γ-interferons. 143 Suitable drug moieties include antineoplastic agents. Non-limiting examples are radioisotopes, vinca alkaloids such as the vinblastine, vincristine and vindesine sulfates, adriamycin, bleomycin sulfate, carboplatin, cisplatin, cyclophosphamide, cytarabine, dacarbazine, dactinomycin, duanorubicin hydrochloride, doxorubicin hydrochloride, etoposide, fluorouracil, lomustine, mechlororethamine hydrochloride, melphalan, mercaptopurine, methotrexate, mitomycin, mitotane, pentostatin, pipobroman, procarbaze hydrochloride, streptozotocin, taxol, thioguanine, and uracil mustard.

144 Immunotoxins, including single chain molecules, can be produced by recombinant means. Production of various immunotoxins is well-known in the art, and methods can be found, for example, in "Monoclonal Antibody-toxin Conjugates: Aiming the Magic Bullet," Thorpe et al. (1982) Monoclonal Antibodies in Clinical Medicine, Academic Press, pp. 168-190; Vitatta (1987) Science 238:1098-1104; and Winter and Milstein (1991) Nature 349:293-299. Suitable toxins include, but are not limited to, ricin, radionuclides, pokeweed antiviral protein, Pseudomonas exotoxin A, diphtheria toxin, ricin A chain, fungal toxins such as restrictocin and phospholipase enzymes. See, generally, "Chimeric Toxins," Olsnes and Pihl, Pharmac. Ther. 15:355- 381 (1981); and "Monoclonal Antibodies for Cancer Detection and Therapy," eds. Baldwin and Byers, pp. 159-179, 224-266, Academic Press (1985).

145 The chemically functional moieties can be made recombinantly for instance by creating a fusion gene encoding the Abu and the functional moiety. Alternatively, the Abu can be chemically bonded to the moiety by any of a variety of well-established chemical procedures. For example, when the moiety is a protein, the linkage may be by way of heterobifunctional cross linkers, e.g., SPDP, carbodiimide glutaraldehyde, or the like. The moieties may be covalently linked, or conjugated, through a secondary reagent, such as a second antibody, protein A, or a biotin-avidin complex. Paramagnetic moieties and the conjugation thereof to antibodies are well-known in the art. See, e.g., Miltenyi et al (1990) Cytometry 11:231-238. Preparation of Antigen-Binding Units (Abus):

146 The subject Abus can be prepared by recombinant DNA technology, synthetic chemistry techniques, or a combination thereof. For instance, sequences encoding the desired components of the Abus, including VL, VH and the heterodimerization sequences are typically assembled and fragments ligated into an expression vector. These sequences may be assembled from other vectors encoding the desired protein sequence, from PCR-generated fragments using respective template nucleic acids, or by assembly of synthetic oligonucleotides encoding the desired sequences. However, all nucleic acid sequences encoding the Abus are preferably assembled by in-frame fusion of coding sequences. Flexons, described above, can be included between various components and domains in order to enhance the ability of the individual components to assume a configurations relatively independently of each other. To produce Nsc Abus, the L and H chain can be formed separately and then assembled, or assembled in vivo by an expression system for both chains. Such expression systems can be created by transfecting a suitable cell with a vector comprising separate transcribable regions for the L and H chain, or by co-transfecting the same cell with vectors for each chain.

147 The assembled Abus can be isolated using a variety of protein purification techniques known in the art. Generally, the Abu is isolated from culture media as secreted polypeptides, although they can be recovered from host cell lysates or bacterial periplasm, when directly produced without signal peptides. If the Abus are membrane- bound, they be solubilized by suitable detergent solutions commonly employed by artisans in the field. The recovered Abus may be further purified by salt precipitation (e.g., with ammonium sulfate), ion exchange chromatography (e.g. on a cationic or anionic exchange column run at neutral pH and eluted with step gradients of increasing ionic strength), gel filtration chromatography (including gel filtration HPLC), and chromatography on tag-affinity column, or on affinity resins such as protein A, protein G, hydroxyapatite, and anti-immunoglobulin. Polynucleotides and Vectors of the Present Invention

148 The invention provides various polynucleotides that encode the Abus of the invention. The invention polynucleotides are characterized, in part, by the unique heterodimerization sequences contained therein as detailed above. Such heterodimerization sequences enable efficient assembly of and screening for Abus, such as those that specifically bind to a desired antigen. Such sequences also facilitate display of heteromultimers on living biological entities including phages, bacteria, other prokaryotic or eukaryotic cells. Preferred heterodimerization sequences shown in SEQ ID NOS. 2 and 4.

149 In one embodiment, this invention provides isolate polynucleotides that encode the subject Nsc Abus. In one aspect of this embodiment, the recombinant polynucleotide comprises a coding sequence that encodes the light-chain polypeptide of a subject Nsc Abu. In another aspect, the recombinant polynucleotide comprises a coding sequence that encodes the heavy-chain polypeptide of a Nsc Abu. In yet another aspect, the recombinant polynucleotide comprises two separate coding sequences, one of which codes for the light-chain polypeptide, and the other codes for the heavy-chain.

150 Nucleotide sequences corresponding to various regions of L or H chains of an existing antibody can be readily obtained and sequenced using convention techniques including but not limited to hybridization, PCR, and DNA sequencing. Hybridoma cells that produce monoclonal antibodies serve as a preferred source of antibody nucleotide sequences. A vast number of hybridoma cells producing an array of monoclonal antibodies may be obtained from public or private repositories. The largest depository agent is American Type Culture Collection (http://www.atcc.org), which offers a diverse collection of well-characterized hybridoma cell lines. Alternatively, antibody nucleotides can be obtained from immunized or non-immunized rodents or humans, and form organs such as spleen and peripheral blood lymphocytes. Specific techniques applicable for extracting and synthesizing antibody nucleotides are described in Orlandi et α/.(1989) Proc. Natl. Acad. Sci. U.S.A 86: 3833-3837; Larrick et al. (1989) Biochem. Biophys. Res. Commun. 160:1250-1255; Sastry et al (1989) Proc. Natl. Acad. Sci, U.S.A. 86: 5728-5732; and U.S Patent No. 5,969,108.

151 The antibody nucleotide sequences may also be modified, for example, by substituting the coding sequence for human heavy and light chain constant regions in place of the homologous non-human sequences. In that manner, chimeric antibodies are prepared that retain the binding specificity of the original antibody.

152 It is also understood that the polynucleotides embodied in the invention include those coding for functional equivalents and fragments thereof of the exemplified polypeptides. Functionally equivalent polypeptides include those that enhance, decrease or not significantly affect properties of the polypeptides encoded thereby. Functional equivalents may be polypeptides having conservative amino acid substitutions, analogs including fusions, and mutants.

153 Due to the degeneracy of the genetic code, there can be considerable variation in nucleotides of the L and H sequences, as well as the heterodimerization sequences suitable for construction of the polynucleotide and vectors of the present invention.

Sequence variants may have modified DNA or amino acid sequences, one or more substitutions, deletions, or additions, the net effect of which is to retain the desired antigen-binding activity. For instance, various substitutions can be made in the coding region that either do not alter the amino acids encoded or result in conservative changes.

These substitutions are encompassed by the present invention. Conservative amino acid substitutions include substitutions within the following groups: glycine, alanine; valine, isoleucine, leucine; aspatic acid, glutamic acid; asparagine, glutamine; serine, threonine; lysine, arginine; and phenylalanine, tyrosine. While conservative substitutions do effectively change one or more amino acid residues contained in the polypeptide to be produced, the substitutions are not expected to interfere with the antigen-binding activity of the resulting Abus to be produced. Nucleotide substitutions that do not alter the amino acid residues encoded are useful for optimizing gene expression in different systems. Suitable substitutions are known to those of skill in the art and are made, for instance, to reflect preferred codon usage in the expression systems. 154 Where desired, the recombinant polynucleotides may comprise heterologous sequences that facilitate detection of the expression and purification of the gene product. Examples of such sequences are known in the art and include those encoding reporter proteins such as β-galactosidase, β -lactamase, chloramphenicol acetyltransferase (CAT), luciferase, green fluorescent protein (GFP) and their derivatives. Other heterologous sequences that facilitate purification may code for epitopes such as Myc, HA (derived from influenza virus hemagglutinin), His-6, FLAG, or the Fc portion of immunoglobulin, glutathione S-transferase (GST), and maltose-binding protein (MBP).

155 The polynucleotides can be conjugated to a variety of chemically functional moieties described above. Commonly employed moieties include labels capable of producing a detectable signal, signal peptides, agents that enhance immunologic reactivity, agents that facilitate coupling to a solid support, vaccine carriers, bioresponse modifiers, paramagnetic labels and drugs. The moieties can be covalently linked polynucleotide recombinantly or by other means known in the art.

156 The polynucleotides of the invention can comprise additional sequences, such as additional encoding sequences within the same transcription unit, controlling elements such as promoters, ribosome binding sites, and polyadenylation sites, additional transcription units under control of the same or a different promoter, sequences that permit cloning, expression, and transformation of a host cell, and any such construct as may be desirable to provide embodiments of this invention.

157 The polynucleotides embodied in this invention can be obtained using chemical synthesis, recombinant cloning methods, PCR, or any combination thereof. Methods of chemical polynucleotide synthesis are well known in the art and need not be described in detail herein. One of skill in the art can use the sequence data provided herein to obtain a desired polynucleotide by employing a DNA synthesizer or ordering from a commercial service.

158 Polynucleotides comprising a desired sequence can be inserted into a suitable vector which in turn can be introduced into a suitable host cell for replication and amplification. Accordingly, the invention encompasses a variety of vectors comprising one or more of the polynucleotides of the present invention. Also provided is a selectable library of expression vectors comprising at least one vector encoding the subject Abus.

159 Vectors of the present invention are generally categorized into cloning and expression vectors. Cloning vectors are useful for obtaining replicate copies of the polynucleotides they contain, or as a means of storing the polynucleotides in a depository for future recovery. Expression vectors (and host cells containing these expression vectors) can be used to obtain polypeptides produced from the polynucleotides they contain. Suitable cloning and expression vectors include any known in the art, e.g., those for use in bacterial, mammalian, yeast, insect and phage display expression systems.

160 Suitable cloning vectors can be constructed according to standard techniques, or selected from a large number of cloning vectors available in the art. While the cloning vector selected may vary according to the host cell intended to be used, useful cloning vectors will generally have the ability to self-replicate, may possess a single target for a particular restriction endonuclease, or may carry marker genes. Suitable examples include plasmids and bacterial viruses, e.g., pBR322, pMB9, ColEl, pCRl, RP4, pUC18, mpl8, mpl9, phage DNAs (including filamentous and non-filamentous phage DNAs), and shuttle vectors such as pSA3 and pAT28. These and other cloning vectors are available from commercial vendors such as Clontech, BioRad, Stratagene, and Invitrogen.

161 Expression vectors containing these nucleic acids are useful to obtain host vector systems to produce proteins and polypeptides. It is implied that these expression vectors must be replicable in the host organisms either as episomes or as an integral part of the chromosomal DNA. Suitable expression vectors include plasmids, viral vectors, including phagemids, adenoviruses, adeno-associated viruses, retroviruses, cosmids, etc.

A number of expression vectors suitable for expression in eukaryotic cells including yeast, avian, and mammalian cells are known in the art. One example of an expression vector is pcDNA3 (Invitrogen, San Diego, CA), in which transcription is driven by the cytomegalo virus (CMV) early promoter/enhancer. Two types of particularly useful expression vectors for expressing the subject Abus are the phage display vector and bacterial display vector.

162 The techniques for constructing phage display vectors are well established in the art (see review article by Winter G. et al. (1994) Ann. Rev. Immunol. 12:433-55). Both filamentous and non-filamentous phage sequences are applicable for constructing a display vector. Filamentous phage vectors are preferred because the genomes of many representative phages of this class have been sequenced, and their genomes are found to be much smaller than that of non-filamentous phages. Representative phages of this class include Ml 3, fl, fd, Ifl, Ike, Xf, Pfl, and Pf3. The phage vector typically is constructed to express heteromultimers, e.g. antibody peptides, by fusion to a part or all of a phage coat protein. Suitable coat proteins include pill, VIII, VI, VII and IX of Ml 3. The heteromultimer sequence must be inserted into the phage vector in such a way that the integrity of the expressed phage coat is not undermined, and the heteromultimer is preferably biologically functional.

163 For constructing pill fusion vector, commonly employed fusion sites are located at the amino terminus, in between the flexible spacer between the two domains of pill (Smith et al. Science 288:1315-17), or any other alternative fusion sites described in U.S. Patent Nos. 5,969,108, 5,837,500. The pill fusion and other proteins of the phage can be encoded entirely within the same page replicon or on different replicons. When at least two replicons are used, the pill fusion is generally encoded on a phagemid, a plasmid containing a phage origin of replication. Phagemids can be packaged into phage particles by "rescue" with a helper phage such as M13KO7, which provides all the phage proteins, including pill, but due to a defect origin is itself poorly packaged in competition with the phagemids. Other multivalent helper phages (e.g. M13ΔgIII) that lack or contain altered pill to enhance the package efficiency can also be employed (Rondot et al. Nature Biotechnology 19:75-78).

164 Similar constructions can be made with other filamentous phage. Pf3 is a well- known filamentous phage that infects Pseudomonas aerugenosa cells that harbor an IncP-

1 plasmid. The entire genome has been sequenced and the genetic signals involved in replication and assembly are characterized. The major coat protein of PF3 is unusual in having no signal peptide to direct its secretion. The sequence has charged residues ASP₇, ARG₃₇, LYS ₀, and PHE^-COO^" which is consistent with the amino terminus being exposed. To construct a display Pf3 vector, it is generally desirable to engineer a signal sequence known to cause secretion in P. aerugenosa fused in-frame to a gene fragment encoding a heterologous polypeptide, which in turn is fused in-frame with a DNA encoding the mature Pf3 coat protein.

165 The same general construction scheme applies to generating display vectors containing sequences derived from non-filamentous phages including bacteriophage XI 74, λ, T4 and T7 phages. A wealth of information on the structures of these non- filamentous phages is known in the art. One skilled in the art can readily generate a corresponding display vector that expresses the subject heteromultimers using the unique heterodimerization sequences without undue experimentation.

166 In addition to phage display vector, another class of preferred vector is bacterial display vector. The general scheme outlined above is equally applicable for constructing such vectors. Briefly, the vectors facilitate expression of a heteromultimer, Abus in particular, as a fusion with a bacterial surface protein. Prior research has revealed a vast number of bacterial surface proteins applicable for expressing such fusions. Non- limiting examples of bacterial surface proteins are LamB (Bremer et al Proc. Natl. Acad. Sci U.S.A. (1984) 81:3830-34; Gene (1987) 52:165-73); OmpA (Prog Biophys Molec Biol (1987) 49:89-115); OmpC; OmpF (Pages et al. Biochemimie (1990) 72:169- 76); PhoE (van der Ley et al. J. Biol. Chem. 261:12222-5); pilin (So et al. Curr Top in Microbiol & Immunol (1985) 118:13-28); pldA (de Geus et al EMBOJ. (1984) 3(8): 1799-1802) and their homologs. Characterization of these and other surface proteins, and the methods of using these proteins for displaying heterologous polypeptides are detailed in U.S. Patent No. 5,837,500 as well as the references cited therein.

167 The vectors of the present invention generally comprises a transcriptional or translational control sequences required for expressing the Abus. Suitable transcription or translational control sequences include but are not limited to replication origin, promoter, enhancer, repressor binding regions, transcription initiation sites, ribosome binding sites, translation initiation sites, and tenmnation sites for transcription and translation.

168 As used herein, a "promoter" is a DNA region capable under certain conditions of binding RNA polymerase and initiating transcription of a coding region located downstream (in the 3' direction) from the promoter. It can be constitutive or inducible. In general, the promoter sequence is bounded at its 3' terminus by the transcription initiation site and extends upstream (5¹ direction) to include the minimum number of bases or elements necessary to initiate transcription at levels detectable above background. Within the promoter sequence is a transcription initiation site, as well as protein binding domains responsible for the binding of RNA polymerase. Eukaryotic promoters will often, but not always, contain "TATA" boxes and "CAT" boxes.

169 The choice of promoters will largely depend on the host cells in which the vector is introduced. For animal cells, a variety of robust promoters, both viral and non- viral promoters, are known in the art. Non-limiting representative viral promoters include CMV, the early and late promoters of SV40 virus, promoters of various types of adenoviruses (e.g. adenovirus 2) and adeno-associated viruses. It is also possible, and often desirable, to utilize promoters normally associated with a desired light or heavy chain gene, provided that such control sequences are compatible with the host cell system.

170 Suitable promoter sequences for other eukaryotic cells include the promoters for 3-phosphoglycerate kinase, or other glycolytic enzymes, such as enolase, glyceraldehyde-3 -phosphate dehydrogenase, hexokinase, pyruvate decarboxylase, phosphofructokinase, glucose-6-phosphate isomerase, 3-phosphoglycerate mutase, pyruvate kinase, triosephosphate isomerase, phosphoglucose isomerase, and glucokinase. Other promoters, which have the additional advantage of transcription controlled by growth conditions, are the promoter regions for alcohol dehydrogenase 2, isocytochrome C, acid phosphatase, degradative enzymes associated with nitrogen metabolism, and the aforementioned glyceraldehyde-3 -phosphate dehydrogenase, and enzymes responsible for maltose and galactose utilization.

171 In certain preferred embodiments, the vectors of the present invention use strong enhancer and promoter expression cassettes. Examples of such expression cassettes include the human cytomegalovirus immediately early (HCMV-IE) promoter (Boshart et al, Cell 41: 521,(1985)), the β-actin promoter (Gunning et al (1987) Proc. Natl. Acad. Sci. (USA) 84: 5831), the histone H4 promoter (Guild et α/.(1988), J. Viral 62: 3795), the mouse metallothionein promoter (Mclvor et al. (1987), Mol, Cell. Biol. 7: 838), the rat growth hormone promoter (Millet et al. (1985), Mol. Cell Biol. 5: 431), the human adenosine deaminase promoter (Hantzapoulos et al. (1989) Proc. Natl. Acad. Sci. USA 86: 3519), the HSV tk promoter 25 (Tabin et al. (1982) Mol. Cell. Biol 2: 426), the α-1 antitrypsin enhancer (Peng et al. (1988) Proc. Natl. Acad. Sci. USA 85: 8146), and the immunoglobulin enhancer/promoter (Blankenstein et al. (1988) Nucleic Acid Res. 16: 10939), the SV40 early or late promoters, the Adeno virus 2 major late promoter, or other viral promoters derived from polyoma viris, bovine papilloma virus, or other retroviruses or adenoviruses. The promoter and enhancer elements of immunoglobulin (Ig) genes confer marked specificity to B lymphocytes (Banerji et al. (1983) Cell 33: 729; Gillies et al. (1983) Cell 33: 717; Mason et al. (1985) Cell 41: 479), while the elements controlling transcription of the B-globin gene function only in erythroid cells (van Assendelft et al. (1989) Cell 56:969).

172 Cell-specific or tissue-specific promoters may also be used. A vast diversity of tissue specific promoters have been described and employed by artisans in the field. Exemplary promoters operative in selective animal cells include hepatocyte-specific promoters and cardiac muscle specific promoters. Depending on the choice of the recipient cell types, those skilled in the art will know of other suitable cell-specific or tissue-specific promoters applicable for the construction of the expression vectors of the present invention.

173 Using well-known restriction and ligation techniques, appropriate transcriptional control sequences can be excised from various DNA sources and integrated in operative relationship with the intact selectable fusion genes to be expressed in accordance with the present invention.

174 In constructing the subj ect vectors, the termination sequences associated with the exogenous sequences are also inserted into the 3' end of the sequence desired to be transcribed to provide polyadenylation of the mRNA and/or transcriptional termination signal. The terminator sequence preferably contains one or more transcriptional termination sequences (such as polyadenylation sequences) and may also be lengthened by the inclusion of additional DNA sequence so as to further disrupt transcriptional read- through. Preferred terminator sequences (or termination sites) of the present invention have a gene that is followed by a transcription termination sequence, either its own termination sequence or a heterologous termination sequence. Examples of such termination sequences include stop codons coupled to various polyadenylation sequences that are known in the art, widely available, and exemplified below. Where the terminator comprises a gene, it can be advantageous to use a gene which encodes a detectable or selectable marker; thereby providing a means by which the presence and/or absence of the terminator sequence (and therefore the corresponding inactivation and/or activation of the transcription unit) can be detected and/or selected.

175 In addition to the above-described elements, the vectors may contain a selectable marker (for example, a gene encoding a protein necessary for the survival or growth of a host cell transformed with the vector), although such a marker gene can be carried on another polynucleotide sequence co-introduced into the host cell. Only those host cells into which a selectable gene has been introduced will survive and/or grow under selective conditions. Typical selection genes encode protein(s) that (a) confer resistance to antibiotics or other toxins, e.g., ampicillin, neomycyin, G418, methotrexate, etc.; (b) complement auxotrophic deficiencies; or (c) supply critical nutrients not available from complex media. The choice of the proper marker gene will depend on the host cell, and appropriate genes for different hosts are known in the art.

176 In a preferred embodiment, the vector is a shuttle vector, capable of replicating in at least two unrelated expression systems. In order to facilitate such replication, the vector generally contains at least two origins of replication, one effective in each expression system. Typically, shuttle vectors are capable of replicating in a eukaryotic expression system and a prokaryotic expression system. This enables detection of protein expression in the eukaryotic host (the expression cell type) and amplification of the vector in the prokaryotic host (the amplification cell type). Preferably, one origin of replication is derived from SV40 and one is derived from pBR322 although any suitable origin known in the art may be used provided it directs replication of the vector. Where the vector is a shuttle vector, the vector preferably contains at least two selectable markers, one for the expression cell type and one for the amplification cell type. Any selectable marker known in the art or those described herein may be used provided it functions in the expression system being utilized

177 The vectors embodied in this invention can be obtained using recombinant cloning methods and/or by chemical synthesis. A vast number of recombinant cloning techniques such as PCR, restriction endonuclease digestion and ligation are well known in the art, and need not be described in detail herein. One of skill in the art can also use the sequence data provided herein or that in the public or proprietary databases to obtain a desired vector by any synthetic means available in the art.

Host Cells of the Present Invention:

178 The invention provides host cells transfected with the vectors or a library of the expression vectors described above. The expression vectors can be introduced into a suitable prokaryotic or eukaryotic cell by any of a number of appropriate means, including electroporation, microprojectile bombardment; lipofection, infection (where the vector is coupled to an infectious agent), transfection employing calcium chloride, rubidium chloride, calcium phosphate, DEAE-dextran, or other substances. The choice of the means for introducing vectors will often depend on features of the host cell.

179 For most animal cells, any of the above-mentioned methods is suitable for vector delivery. Preferred animal cells are vertebrate cells, preferably mammalian cells, capable of expressing exogenously introduced gene products in large quantity, e.g. at the milligram level. Non-limiting examples of preferred cells are NIH3T3 cells, COS, HeLa, and CHO cells. The animal cells can be cultured in a variety of media. Commercially available media such as Ham's F10 (Sigma), Minimal Essential Medium (MEM, Sigma), RPMI- 1640 (Sigma), and Dulbecco's Modified Eagle's Medium (DMEM, Sigma) are suitable for culturing the host cells. In addition, animal cells can be grown in a defined medium that lacks serum but is supplemented with hormones, growth factors or any other factors necessary for the survival and/or growth of a particular cell type. Whereas a defined medium supporting cell survival maintains the viability, morphology, capacity to metabolize and potentially, capacity of the cell to differentiate, a defined medium promoting cell growth provides all chemicals necessary for cell proliferation or multiplication. The general parameters governing mammalian cell survival and growth in vitro are well established in the art. Physicochemical parameters which may be controlled in different cell culture systems are, e.g., pH, pO₂, temperature, and osmolarity. The nutritional requirements of cells are usually provided in standard media formulations developed to provide an optimal environment. Nutrients can be divided into several categories: amino acids and their derivatives, carbohydrates, sugars, fatty acids, complex lipids, nucleic acid derivatives and vitamins. Apart from nutrients for maintaining cell metabolism, most cells also require one or more hormones from at least one of the following groups: steroids, prostaglandins, growth factors, pituitary hormones, and peptide hormones to proliferate in serum-free media (Sato, G.H., et al. in "Growth of Cells in Hormonally Defined Media", Cold Spring Harbor Press, N.Y., 1982). In addition to hormones, cells may require transport proteins such as transferrin (plasma iron transport protein), ceruloplasmin (a copper transport protein), and high-density lipoprotein (a lipid carrier) for survival and growth in vitro. The set of optimal hormones or transport proteins will vary for each cell type. Most of these hormones or transport proteins have been added exogenously or, in a rare case, a mutant cell line has been found which does not require a particular factor. Those skilled in the art will know of other factors required for maintaining a cell culture without undue experimentation. 181 For plant cells, a variety of vector delivery techniques is available in the art. The host cells may be in the form of whole plants, isolated cells or protoplasts. Illustrative procedures for introducing vectors into plant cells include Agrobacterium-mediated plant transformation, protoplast transformation, gene transfer into pollen, injection into reproductive organs and injection into immature embryos. As is evident to one skilled in the art, each of these methods has distinct advantages and disadvantages. Thus, one particular method of introducing vectors into a particular plant species may not necessarily be the most effective for another plant species.

182 Agrobacterium tumefaciens-mediated transfer is a widely applicable system for introducing vectors into plant cells because the vector can be introduced into whole plant tissues, bypassing the need for regeneration of an intact plant from a protoplast. The use of Agrobacterium-mediated expression vectors to introduce vector into plant cells is well known in the art. This technique makes use of a common feature of Agrobacterium which colonizes plants by transferring a portion of their DNA (the T-DNA) into a host cell, where it becomes integrated into nuclear DNA. The T-DNA is defined by border sequences which are 25 base pairs long, and any DNA between these border sequences is transferred to the plant cells as well. The insertion of a recombinant plant viral nucleic acid between the T-DNA border sequences results in transfer of the recombinant plant viral nucleic acid to the plant cells, where the recombinant plant viral nucleic acid is replicated, and then spreads systemically through the plant.

183 Because not all plants are natural hosts for Agrobacterium, alternative methods such as transformation of protoplasts may be employed to introduce the subject vectors into the host cells. For certain monocots, transformation of the plant protoplasts can be achieved using methods based on calcium phosphate precipitation, polyethylene glycol treatment, electroporation, and combinations of these treatments.

184 In addition to protoplast transformation, particle bombardment is an alternative and convenient technique for delivering the invention vectors into a plant host cell.

Specifically, the plant cells may be bombarded with microparticles coated with a plurality of the subject vectors. Bombardment with DNA-coated microprojectiles has been successfully used to produce stable transformants in both plants and animals (see, for example, Sanford et al. (1993) Methods in Enzymology, 217:483-509). Microparticles suitable for introducing vectors into a plant cell are typically made of metal, preferably tungsten or gold. These microparticles are available for example, from BioRad (e.g., Bio-Rad's PDS-1000/He). Those skilled in the art will know that the particle bombardment protocol can be optimized for any plant by varying parameters such as He pressure, quantity of coated particles, distance between the macrocarrier and the stopping screen and flying distance from the stopping screen to the target.

185 Vectors can also be introduced into plants by direct DNA transfer into pollen as described by Zhou et al, Methods in Enzymology, 101:433 (1983); D. Hess, Intern Rev. Cytol, 107:367 (1987); Luo et al, Plant Mol. Biol. Reporter, 6:165 (1988). Alternatively, the vectors can be injected into reproductive organs of a plant as described by Pena et al, Nature, 325:274 (1987).

186 Other techniques for introducing nucleic acids into a plant cell include:

(a) Hand Inoculations. Hand inoculations are performed using a neutral pH, low molarity phosphate buffer, with the addition of celite or carborundum (usually about 1%). One to four drops of the preparation is put onto the upper surface of a leaf and gently rubbed.

(b) Mechanized Inoculations of Plant Beds. Plant bed inoculations are performed by spraying (gas-propelled) the vector solution into a tractor- driven mower while cutting the leaves. Alternatively, the plant bed is mowed and the vector solution sprayed immediately onto the cut leaves.

(c) High Pressure Spray of Single Leaves. Single plant inoculations can also be performed by spraying the leaves with a narrow, directed spray (50 psi, 6-12 inches from the leaf) containing approximately 1% carborundum in the buffered vector solution.

(d) Vacuum Infiltration. Inoculations may be accomplished by subjecting a host organism to a substantially vacuum pressure environment in order to facilitate infection. 187 Other suitable host cells for cloning and expressing the subject vectors are prokaryotes and eukaryotic microbes such as fungi or yeast cells. Suitable prokaryotes for this purpose include bacteria including Gram-negative and Gram-positive organisms. Representative members of this class of microorganisms are Enterobacteriaceae (e.g E. coli), Enterobacter, Erwinia, Klebsiella, Proteus, Salmonella (e.g. Salmonella typhimurium), Serratia (e.g., Serratia marcescans), Shigella, Neisseria (e.g. Neisseria meningitidis) as well as Bacilli (e.g. Bacilli subtilis and Bacilli licheniformis). Preferably, the host cell secretes minimal amounts of proteolytic fragments of the expressed Abus. Commonly employed fungi (including yeast) host cells are S. cerevisiae, Kluyveromyces lactis (K. lactis), species of Candida including C. albicans and C. glabrata, C. maltosa, C. utilis, C. stellatoidea, C. parapsilosis, C. tropicalus, Neurospora crassas, Aspergillus nidulans, Schizosaccharomyces pombe (S. pombe), Pichia pastor is, and Yarowia lipolytica.

188 Once introduced into a suitable host cell, expression of the Abus can be determined using any nucleic acid or protein assay known in the art. For example, the presence of transcribed mRNA of L or H chain, or the Sc Abu can be detected and/or quantified by conventional hybridization assays (e.g. Northern blot analysis), amplification procedures (e.g. RT-PCR), SAGE (U.S. Patent No. 5,695,937), and array- based technologies (see e.g. U.S. Pat. Nos. 5,405,783, 5,412,087 and 5,445,934), using probes complementary to any region of Abu polynucleotide.

189 Expression of the vector can also be determined by examining the Abu expressed. A variety of techniques are available in the art for protein analysis. They include but are not limited to radioimmunoassays, ELIS A (enzyme linked immunoradiometric assays), "sandwich" immunoassays, immunoradiometric assays, in situ immunoassays (using e.g., colloidal gold, enzyme or radioisotope labels), western blot analysis, immunoprecipitation assays, immunoflourescent assays, and PAGE-SDS. Uses of the polynucleotides, vectors and host cells of the present invention:

190 The polynucleotides and vectors of this invention have several specific uses. They are useful, for example, in expression systems for the production of both Sc and Nsc Abus. The polynucleotides are also useful as primers to effect amplification of desired polynucleotides. Furthermore, the polynucleotides of this invention are also useful in pharmaceutical compositions including vaccines, diagnostics, and drugs.

191 The host cells of this invention can be used, inter alia, as repositories of the subject polynucleotides, vectors, or as vehicles for producing and screening desired Abus based on their antigen binding specificities.

192 Accordingly, the invention provides a method of identifying a Nsc Abu that is immunoreactive with a desired antigen. The method involves the following steps: (a) preparing a genetically diverse repertoire of Abus, wherein the repertoire comprises at least one subject Abu; (b) contacting the repertoire of antigen binding units with the desired antigen; (c) detecting a specific binding between Abus and the antigen, thereby identifying the Abu that is immunoreactive with the desired antigen.

193 The ability of an Abu to specifically bind to a desired antigen can be tested by a variety of procedures well established in the art. See Harlow and Lane (1988) Antibodies: A Laboratory Manual, Cold Spring Harbor Laboratory, New York; Gherardi et al. (1990) J. Immunol. Meth. 126:61-68. Typically, Abus exhibiting desired binding specificities can be detected directly by immunoassays, for example, by reacting labeled Abus with the antigens that are immobilized on a solid support or substrate. In general, the substrate to which the antigen is adhered is fabricated with material exhibiting a low level of non-specific binding during immunoassay. A preferred solid support is made from one or more of the following types of materials: plastic polymers, glass, cellulose, nitrocellulose, semi-conducting material, and metal. Preferably, the substrate is petri dish, chromatography beads, magnetic beads, and the like.

194 For such solid-phase assay, the unreacted Abus are removed by washing. In a liquid-phase assay, however, the unreacted Abus are removed by some other separation technique, such as filtration or chromatography. After binding the antigen to the labeled Abus, the amount of bound label is determined. A variation of this technique is a competitive assay, in which the antigen is bound to saturation with an original binding molecule. When a population of the subject Abu is introduced to the complex, only those that exhibit higher binding affinity will be able to compete, and thus remain bound to the antigen.

195 Alternatively, specific binding to a given antigen can be assessed by cell sorting, which involves presenting the desired antigen on the cells to be sorted, then labeling the target cells with Abus that are coupled to detectable agents, followed by separating the labeled cells from the unlabeled ones in a cell sorter. A sophisticated cell separation method is fluorescence-activated cell sorting (FACS). Cells traveling in single file in a fine stream are passed through a laser beam, and the fluorescence of each cell bound by the fluorescently labeled Abu is then measured.

196 Subsequent analysis of the eluted Abus may involve protein sequencing for delineating the amino acid sequences of the L and H chains. Based on the deduced amino acid sequences, the cDNA encoding the antibody polypeptides can then be obtained by recombinant cloning methods including PCR, library screening, homology searches in existing nucleic acid databases, or any combination thereof. Commonly employed databases include but are not limited to GenBank, EMBL, DDBJ, PDB, SWISS-PROT, EST, STS, GSS, and HTGS.

197 When the repertoire of Abu is displayed on phage or bacterial particles, selection is preferably performed using affinity chromatography. The method typically proceeds with binding a repertoire of phage Abus to an antigen coated plates, column matrices, cells or to biotinylated antigen in solution followed by capture. The phages or bacteria bound to the solid phase are washed and then eluted by soluble hapten, acid or alkali. Alternatively, increasing concentrations of antigen can be used to dissociate the Abus from the affinity matrix. For certain Abus with extremely high affinity or avidity to the antigen, efficient elution may require high pH or mild reducing solution as described in WO 92/01047. 198 To avoid potential difficulties in recovering the bound Abus with the desired binding specificities, protease cleavage sites may be introduced between the heterodimerization sequences and the phage coat protein employed for displaying the Abus. Cleavage sites applicable for this purpose include but are not limited to Factor X, trypsin, and thrombin recognition sites. After binding the phage repertoire to an affinity matrix and washing the non-specific phages, the remaining phages that display the Abus with the desired affinity can be collected by washing the antigen-affinity matrix with protease under conditions suitable for digestion at the cleavage site. Such digestion would release the Abus from the phage particles.

199 An alternative procedure to the above is to take the affinity matrix which has retained the strongly bound phage or bacterial particles and extract their nucleic acids, for example by boiling in SDS solution. Extracted nucleic acids can be used to directly transform E. coli host cells or alternatively the antibody encoding sequences can be amplified by PCR using suitable primers.

200 The efficiency of selection is likely to depend on a combination of several factors, including the kinetics of dissociation during washing, and whether multiple Abus on a single phage or bacterium can simultaneously bind to antigens on a solid support. For example, antibodies with fast dissociation kinetics (and weak binding affinities) should be retained by use of short washes, multivalent display and a high coating density of antigen at the solid support. Conversely, the selection of Abus with slow dissociation kinetics (and good binding affinities) should be favored by use of long washes, monovalent phages, and a low coating density of antigen.

201 Where desired, the repertoire of Abus can be pre-selected against an unrelated antigen to counter-select the undesired Abus. The repertoire may also be pre-selected against a related antigen in order to isolate, for example, anti-idiotypic Abus.

202 The subject Abu repertoire enables rapid isolation of Abus with desired specificities. Many of the isolated Abus would be expected to be difficult or impossible to obtain through conventional hybridoma or transgenic animal technology. Kits comprising the vectors of the present invention

203 The present invention also encompasses kits containing the vectors of this invention in suitable packaging. Kits embodied by this invention include those that allow generation of Abus reconstituted via pairwise affinity of a unique heterodimerization sequence pair as described herein.

204 Each kit necessarily comprises the reagents which render the delivery of vectors into a host cell possible. The selection of reagents that facilitate delivery of the vectors may vary depending on the particular transfection or infection method used. The kits may also contain reagents useful for generating labeled polynucleotide probes or proteinaceous probes for detection of Abus. Each reagent can be supplied in a solid form or dissolved/suspended in a liquid buffer suitable for inventory storage, and later for exchange or addition into the reaction medium when the experiment is performed. Suitable packaging is provided. The kit can optionally provide additional components that are useful in the procedure. These optional components include, but are not limited to, buffers, capture reagents, developing reagents, labels, reacting surfaces, means for detection, control samples, instructions, and interpretive information.

205 Further illustration of the development and use of Abus, polynucleotides, vectors and host cells according to this invention are provided in the Example section below. The examples are provided as a guide to a practitioner of ordinary skill in the art, and are not meant to be limiting in any way.

EXAMPLES Construction of Non-Single-Chain Antigen-Binding Units: Coiled-Coil Fv (ccFv)

206 As described above, the Fv fragment is the smallest antibody fragment containing the whole antigen-binding site. Composed of the two variable regions of heavy and light chain (VH and VL), Fv is located at the "upper" tips of the Y-shaped immunoglobulin molecule. The Fv fragments have very low interaction energy between their two VH and

VL fragments, and are often too unstable for many applications at physiological condition. In a naturally occurring immunoglobulin (e.g. Ig), an interchain disulfide bond located in the constant domains CHI and CL is used to link the VH and VL. This linkage makes a stabilized antigen-binding fragment Fab with a molecular weight 50 kDa. It has been shown that the VH and VL fragments can also be artificially held together by a short peptide linker between the carboxyl terminus of one fragment and amino-terminus of another to form a single-chain Fv antibody fragment (scFv). The scFv antigen-binding unit is only half the size of Fab. However, some scFv proteins are also unstable. The polypeptide linker in scFv can interfere with binding in some cases. An interchain disulfide bond has also been introduced into the framework regions in VH and VL to form a disulfide-stabilized Fv (dsFv). The dsFv configuration also has profound limitations. The introduction of two Cys residues into the antigen-binding variable regions may change the intrachain disulfide bond in VH or VL, therefore interfere with antigen binding. We have devised a new strategy to stabilize VH and VL heterodimer. We have designed and used a unique heterodimerization sequence pair to create a Fab-like, functional artificial Fv fragment, namely coiled-coil Fv fragment (ccFv). The heterodimerization pair were derived from the heterodimeric receptors GABA_B receptors 1 and 2. The pair of sequences forms a coiled-coil structure and mediates the functional heterodimerization of GABA_B-R1 and GABA_B-R2 receptors. Distinguished from previously characterized coiled-coil leucine zippers from the

Fos and Jun proteins, the C-terminal coiled coil of GABAβ-Rl and GABAβ-R2 receptors do not form detectable homodimers under physiological conditions (e.g. in vivo); nor do they form homodimers at physiological body temperatures. Research by

Kuner et al. and White et al (Science (1999) 283: 74-77); Nature (1998) 396: 679-682)) have demonstrated the heterodimerization specificity of GABAβ-Rl and GABAβ-R2 in vivo. In fact, White et al. were able to clone GABAβ-R2 from yeast cells based on the exclusive specificity of this heterodimeric receptor pair. In vitro studies by Kammerer et al. supra has shown that neither GABAB-R1 nor GABAβ-R2 C-terminal sequences is capable of forming homodimers in physiological buffer conditions when assayed at physiological body temperatures (see Table 1 of Kammerer). However, none of these researchers who were involved in the original isolation of the GABAβ-R2 gene and the characterization of the coiled-coil sequences describe or even suggest the use of this unique heterodimerization sequences for construction of heteromultimers such as antigen-binding units. We have modified the carboxyl terminus of GR1 and GR2 domains by adding a flexon "SerArgGlyGlyGlyGly" to the amino-terminus of GR1 and GR2 domains to provide additional flexibility to the V regions. To further stabilize ccFv, we have introduced a pair of cysteine residues by adding "ValGlyGlyCys" spacer at the C-termini of the coiled coil. The GR1 and GR2 domains are fused to the carboxyl terminus of VH and VL fragment, respectively. TheVH-GRl and VL-GR2 fusions were expressed in E. coli and displayed by phage. As shown in Figures 10-11, functional heterodimeric ccFv Abus stabilized by the parallel coiled-coil helix were generated. Since the coiled-coil heterodimerization sequences are about half the size of CHI and CL domains, the ccFv (approximately 35 kDa) is smaller than the conventional Fab fragments (approximately 50 kDa). Because of the small size, the ccFvs and its derivatives are potentially more useful for clinical applications such as tumor and tissue penetration. More efficient expression and display of ccFv is expected. Furthermore, the specific assembly of VH and VL regions due to the pairwise affinity of the unique heterodimerization sequences makes the construction of a robust, vast diverse repertoire of Abs more feasible.

Materials and Methods: Bacterial and phage strain: Escherichia coli TGI ( supE A(hsdM-mcrB)5(r_k ^'m_k McrB^')thi A(lac-proAB / F' traD36, LacF A(lacZ)M15] was used for plasmid DNA and phage production; KO7 helper phage and HRP-conjugated anti-M13 antibody from Phamersham Pharmacia Biotech; pbluescript SK(+) from Stratagene; Anti-HA antibody from Santa Cruz Biotechnology. Example 1: Vector construction

pABMXl and pABMX2 Vectors:

The phagemid display vector pABMXl and pABMX2 were derived from pbluescript SK(+). A unique Agel restriction site was introduced immediately after lac promoter by PCR-based site-directed mutagenesis with a set of primer (pBS-SKa: 5'GGAATTGTGAGCGGATAACAATTTACCGGTCACACAGGAAACAGCTATGA CCATG-3' and pBS-SKb

5'CATGGTCATAGCTGTTTCCTGTGTGACCGGTAAATTGTTATCCGCTCACAAT TCC-3'), and the Xhol and Kpnl sites were deleted by cutting and blunt end ligation. Afterwards, the synthetic DNA fragments flanked by Agel site at 5'and Bglll/EcoRI sites at 3', containing translation enhance sequence EP from T7 phage gene 10 (TTAACTTTA), ribosome binding sequence S/D (TAAGGAGG), fd phage gene8 leader sequence with Hindlll site

(ATGAAAAAGTCTTTAGTCCTCAAAGCCTCCGTAGCCGTTGCTCCCTCGTTCC GATGCTAAGCTTCGCT, for pABMXl) or pelB leader sequence (ATGAAATACCTATTGCCTACGGCAGCCGCTGGATTGTTA TTACTCGCGGCCCAGCCGGCCATGGCG, for pABMX2) and HA-(His)₆-tag (DHtagXTATCCATACGACGTACCAGACTACGCAGGAGGTCATCACCATCATCA CC ATTAG), were cloned into modified pbluescript SK(+). The resulting vectors are designated pABMXl and pABMX2 (see Fig. 3A-B for restriction maps and sequences). Heterologous sequences encoding heteromultimers such as Nsc Abus were further subcloned into these vectors for periplasmic expression.

pABMDl and pABMD2 vectors: 212 A PCR-amplified fd gene III (or gene 3) fragment flanked by Bglll and Sail sites was inserted into pABMXl and pABMX2 vectors (see Fig. 4). The heterologous sequence to be displayed can be inserted after the leader sequences. The lac promoter drives expression of the pill capsid fusion, which in turn can be displayed on phage surface after superinfection by helper phage such as KO7.

PABMX5 and pABMX6 vectors:

213 These two vectors were derived from pABMXl and pABMX2. A synthetic

DNA fragment flanked by Xbal/Ascl site at 5'and Mlul/Xhol/Notl sites at 3', containing the ribosome binding sequence S/D (TAAGGAGG) and Gene 3 leader sequence (ATGAAA AAATTATTATTCGCAATTCCTTT AGTTGTTC CTT TCTATTCTCACTCCGCT), was inserted into pABMXl and pABMX2 by Xbal/Notl sites. Afterward, the GRl domain coding sequence (Fig.2) was subcloned into Xba/Ascl sites, and GR2 domain coding sequence (Fig. 2) was inserted into Xhol/Notl site. Then the VH and VL domain were inserted before GRl and GR2 sequence respectively. A schematic representation of the vectors pABMX5 and pABMX6 are shown as Fig. 5 A. These vectors express two proteins: VH-GR1 and VL-GR2 under one lac promoter.

pABMD5 and pABMD6 vector:

214 The ccFv DNA fragments from vector p ABMX5 and ABMX6 vectors were subcloning into pABMDl and pABMD2 to yield vectors pABMD5 and pABMD6 (see Fig. 6A-B for restriction maps and sequences). These vectors express two proteins: VH- GR1 and VL-GR2-pIII fusions. The expressed VH-GR1 and VL-GR2-pIII fusions are secreted into periplasmic space, where dimerization takes place via the coiled-coil domain heterodimerization. The assembled Abu is then displayed on phage surface upon superinfection of helper phages such as KO7.

Example 2: Expression of Functional ccFv

215 Antibody variable domains from antibody AMI were subcloned into pABMX6 vector to expression ccFv fragment. The vector was then introduced into either TGI cells or BL21 cells. The transformed bacteria in 500 ml 2xYT containing approximately 100 υg/ml carbenicillin and 0.1% glucose from a single colony was grew to OD₆oo ⁼ 0.7 (approximately), at 37°C. 1 mM of IPTG was added for 4 hrs inductions at 30°C. The bacterial pellet was collected for periplasmic and osmotic shock preparation. The pellet was resuspended in 12.5 ml PPB buffer (200 mg/ml sucrose, lmM EDTA, 30 mM Tris- HCl, pH 8.0) with 1.25 ml protease inhibitor cocktail from Sigma, and put on ice for 20 min. The supernatant was collected by spinning. The pellet was resuspended in 5 mM MgSO , and incubated on ice for 20 min. The MgSO₄ and PPB supernatants were combined, and dialyzed against PBS. After loading to a 1ml of Ni-NTA column, the His- tag proteins were purified by 350 mM imidozole elution. Figure 10A shows that the purified ccFv has an electrophoretic mobility of 35 kDa on a non-reducing gel. When analyzed under reducing condition, two subunits corresponding to VL and VH were observed. The upper band was confirmed as VL-His-tag fusion by Western blot analysis.

216 To measure the binding specificity of soluble AM 1 -ccFv, ELIS A assay was carried out. The AMI antigens (0.2-lug/well) were coated on ELISA plates for overnight at 4 °C. After 5% milk/PBS blocking, antibody solution in 5% milk/PBS was added to the ELISA plate, and incubated for 1-2 hr at room temperature. The unbound Abus were washed out. Figure 10B shows specific binding of AMl-ccFv to its antigen. The control contains 5% milk in PBS. This result confirms the assembly of functional ccFv by the coiled-coil GABA_B R1 R2 heterodimerization sequences. Example 3: Display of Functional ccFv

217 Display of antibody by a genetic package is a powerful tool to enrich and isolate specific Abu from large libraries. To analyze whether ccFv can be utilized in a phage display system, we have constructed a phagemid vector by subcloning ccFv gene of AMI antibody into pABMD6 vector. TGI cells carrying the phagemid vectors were superinfected by KO7 helper phage. The infected TGI cells were grown in 2xYT/Amp/Kan at 30 °C overnight. The phagemid particles were precipitated by PEG/NaCl from culture supernatants twice, and resuspended in PBS. The antibody displayed on phage was detected by antigen binding activity via phage ELISA assay. Briefly, the antigens were first coated on ELISA plates. After 5% milkPBS blocking, the phage solution was added to ELISA plates. The phage bound to antigen was detected by incubation with HRP-conjugated anti-M13 antibody. The substrate ABTS [2,2'Azino- bis(3-ethylbenzthiazoline-6-sulfonic acid)] was used for measurement of HRP activity. The anti-HA tag antibody was also used to detect the proteins displayed on phage particle. The anti-HA antibody was coated on 96-well plates (2 ug/each well). The phages bound to anti-HA antibody coated on ELISA plate were detected by HRP- conjugated anti-Mi 3 antibody.

218 The single-chain antibody phage was also prepared for comparing the ccFv and scFv phage-display. As shown in Figure 11 A-B, the binding capability of ccFv phages is comparable to that of the conventional scFv phages. For certain ccFv expressing phages, their binding capability is almost one order of magnitude higher than phages expressing conventional scFv (Figure 1 IB). Thus, ccFv is a functional Abu even when displayed on a phage particle.

Expression of Single-Chain Antigen-Binding Units:

Example 4: Expression of conventional scFv

219 The AM2-scFv was subcloning into soluble expression vector pABMXl at the

Hindlll/Notl sites. The periplasmic preparation was carried out as outlined above. A 30 kDa antibody protein purified from NI-NTA column was confirmed by SDS-PAGE analysis, and was tested for its antigen-binding specificity using ELISA. The AM2 antigens were first coated onto ELISA plates at a concentration of 0.2 ug/well. Different amounts of AM2-scFv fragments were incubated with the antigen. Bound AM2-svFv fragments were detected by the anti-HA-tag antibody. The experiment revealed a dose- dependent binding of AM2-scFv to its the AM2 antigen (Figure 8).

Example 5: Display of conventional scFv on phage

220 The AM2-scFv fragment was first subcloning into phagemid vector pABMDl at the Hindlll/Notl sites. TGI cells carrying this phagemid vector were infected by the helper phage KO7. The phages were purified from the supernatants. Phage ELISA assay was subsequently performed to detect AM2-scFv displayed on the phage particles. Because the coat pill gene is tagged with HA-tag, the fusion can be detected with anti- HA antibodies. ELISA assay using AM2 antigen and anti-HA antibodies confirmed that the displayed scFv was capable of specifically binding to the corresponding antigens (Figure 9). The control involves phages displaying unrelated antibodies that are not HA tagged.

Expression of the Subject Antigen-Binding Units in Eukaryotic Cells:

Example 6: Expression of ccFv in yeast

221 The yeast vector pAMEX7 carrying both VL and VH sequences that are linked to the subject heterodimerization sequences is constructed. Competent yeast cells, e.g. AH109 cell, are prepared and transformed with pAMEX7 vectors according to any method known in the art. The transformed yeast cells are cultured under conditions suitable for protein expression. Such conditions are well known to artisans in the field and hence are not detailed herein. The expressed ccFv Abus are harvested using conventional methods known in the art and/or procedures described herein. The antigen binding capability of the harvest ccFv are determined by ELISA according to the protocols described above. Enrichment and Identification of ccFv Phage Exhibiting Desired Binding Specificity:

Example 7: Enrichment of AM2-ccFv phages from a model library

To demonstrate that ccFv-displayed phages with desired binding specificity can be selected and ultimately enriched from background phages, we performed panning of AM2-ccFv phages from a model library. The model library was prepared by mixing AM2-ccFv phages with unrelated AM 1 -ccFv displayed phages at a ratio of 1 : 10⁶ or 1 : 10⁷. Two rounds of panning of the libraries against specific protein antigen that was coated on Nunc Maxisorb 96-well plate were carried out. After blocking with 5% milk in PBS, 1X10¹² library phages in 2% milk/PBS were added to the well, and incubated for 2 hours at room temperature. Phage solution was then discarded, and the wells were washed 5 times with PBST (0.05% Tween-20 in PBS) and 5 times with PBS. Bound phages were eluted with 100 mM triethylamine, and were added to TGI culture for infection. The phages prepared from infected TGI cells were used for next round of panning and ELISA against immobilized protein antigen of AM2-ccFv. After each round panning, the ratio of AM2-ccFv phage to background phage AMl-ccFv phage was also determined by PCR analysis of randomly picked clones via a pair of primers that were specifically designed for AM2-ccFv only (a band of ~1 kb PCR product could be visualized on 1% agarose gel), but not AMl-ccFv (no PCR product). As shown in Figure 20, ELISA against immobilized protein antigen of AM2-ccFv was performed. Phages from second round panning yielded much higher OD405 readings as compared to that of the first round, suggesting that AM2-ccFv phages were successfully selected and enriched during panning. PCR analysis in Figure 21 showed 1% agarose gel image from PCR products of randomly picked clones. The results indicated that the occurrence rate of AM2-ccFv phage was 4.4% from the 1 : 10⁷ library after first round of panning, whereas after second round of panning, the occurrence reached 100%.

Example 8: Library screening using phage library displaying functional ccFv Screening a phage library with a large number of Abus, all of which are capable of binding to a desired antigen but each Abu differs in certain regions of the critical binding domain, allows one to identify and further design antibody with even higher affinity. To demonstrate that ccFv can be used efficiently as a vehicle for antibody library screening, we constructed a library of AM2 in the format of ccFv. The framework structure of AM2 consisted of a VH and a VL region, each was subcloned in to corresponding positions of vector pABMD6 as described above. Briefly, VH was linked to the N-terminus of GRl , and VL was linked to the N-terminus of GR2, respectively. GR2 in turn fused to pill in order to display the ccFv antibody fragment to the surface of the phage particle (See Examples 1, 3, and 7). In order to select phages high affinity binders through ccFv phage library panning, CDR3 of VH was constructed as a library which contains multiple options of residues at multiple positions. The library was prepared through standard synthesis of degenerative DNA oligo franking restriction sites at both ends so that it could be subcloned into its position of AM2 after PCR amplification and restriction digestion. Figure 22 illustrated the design of the degenerative DNA oligo, its amplification primers and directions. The diversity of the library was designed to be approximately lxl 0⁶ species in total. After ligation, the library was electroporated into TGI competent cells. The transformants were then harvested and rescued by KO7 helper phage before phage particles of the library were collected as described in above example. To "pan" the library, recombinant antigen was immobilized onto Nunc Maxisorb 96-well plate in 0.05 M NaHCO₃ pH9.6 buffer at 4°C overnight. An aliquot of the library phage containing 10 phage particles diluted in PBS buffer with 2% milk was added to the well that was coated with the antigen. After incubating at 37°C for 2 hours, the well was washed and phage was eluted as described in example 3 and 7. The eluted phages were used to infect TGI cells, then rescued by

KO7 helper phage. After growing overnight at 30°C, the phages were amplified in TGI cells and ready to be harvested for the next round of panning. Prior studies have shown that the rate of dissociation, k_ofr, contributes most significantly for the affinity of this particular antibody, whereas K_on is relatively constant. Thus, in order to select phages with high binding affinity,, off-rate panning, a procedure of prolonged wash to select survival phages, was designed to specifically select those phages that have slower K_off. A total of 7 consecutive rounds of panning were conducted for this library.

224 After the 5^th and the 7^th panning, individual clones were randomly picked to prepared phages for ELISA using the same protein antigen in phage panning. ELISA was performed using 96-well microtiter plate following the procedures described in example 7 above. As shown in Figure 23, all clones selected at the 5^th and 7^th rounds positively reacted with the antigen. Note that negative control were set at positions H10, 11, 12 of the 5^th round and HI 1, 12 of the 7^th round, all of which were below 0.05 OD₄₀₅.

225 The clones that showed positive reactivity in ELSIA of the 7^th panning were then randomly picked for sequencing and residues on CDR3 of VH were mapped in Figure 24. Consensus on the positions targeted by the library was then identified. Protein expression of a clone that represented the consensus sequence and the wild type (the original AM2) was performed and the proteins were subsequently used in determination of their K_off in binding with immobilized protein antigen using BiaCore (surface plasmon resonance). As shown in Figure 25, the AM2 variant selected from the library has significantly slower Ko_ff.

Claims

CLAIMSWhat is claimed is:

1. A non-single-chain antigen-binding unit comprising:

(a) a light (L) chain polypeptide comprising a light (L) chain variable region fused in-frame to a first heterodimerization sequence;

(b) a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused in-frame to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures.

2. The non-single-chain antigen-binding unit of claim 1 , wherein both of the first and second heterodimerization sequences are essentially incapable of forming homodimers under physiological buffer conditions and at physiological body temperatures.

3. A non-single-chain antigen-binding unit comprising:

(a) a light (L) chain polypeptide comprising a light (L) chain variable region fused in-frame to a first heterodimerization sequence;

(b) a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused in-frame to a second heterodimerization sequence;

wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences, said first and second heterodimerization sequences comprising heterodimeric receptor sequences that mediate heterodimerization of the receptors.

4. The non-single-chain antigen-binding unit of claim 1 or 3, wherein the first and second heterodimerization sequences form a coiled-coil dimer.

5. The non-single-chain antigen-binding unit of claim 1 or 3, wherein the L and the H chain polypeptides dimerize via non-covalent pairwise affinity of the two heterodimerization sequences.

6. The non-single-chain antigen-binding unit of claim 4, wherein the L chain polypeptide further comprises a flexon that is flanked by the L chain variable region and the first heterodimerization sequence.

7. The non-single-chain antigen-binding unit of claim 4, wherein the H chain polypeptide further comprises a flexon sequence that is flanked by the H chain variable region and the second heterodimerization sequence.

8. The non-single-chain antigen-binding unit of claim 4, wherein both the first and the second heterodimerization sequences are linked to at least one cysteine residue.

9. The non-single-chain antigen-binding unit of claim 4, wherein the antigen- binding unit is multivalent.

10. The non-single-chain antigen-binding unit of claim 4, wherein the antigen- binding unit is multispecific.

11. The non-single-chain antigen-binding unit of claim 10, wherein the antigen- binding unit is bispecific.

12. The non-single-chain antigen-binding unit of claim 10, wherein the antigen- binding unit is trispecific.

13. The non-single-chain antigen-binding unit of claim 4, wherein the L chain polypeptide comprises sequences derived from a human light chain.

14. The non-single-chain antigen-binding unit of claim 4, wherein the H chain polypeptide comprises sequences derived from a human heavy chain.

15. The non-single-chain antigen-binding unit of claim 4, wherein the antigen- binding unit is conjugated to a chemically functional moiety.

16. The non-single-chain antigen-binding unit of claim 15, wherein the moiety is selected from the group consisting of signal peptides, agents that enhance immunologic reactivity, agents that facilitate coupling to a solid support, vaccine carriers, bioresponse modifiers, toxins, detectable labels, paramagnetic labels, and drugs.

17. The non-single-chain antigen-binding unit of claim 4, wherein the first and second heterodimerization sequences are derived from C-terminal sequences of GABA_B receptor 1 and GABA_B receptor 2, respectively.

18. The non-single-chain antigen-binding unit of claim 4, wherein the first heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 1 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 2 and wherein the second heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 2 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4, wherein said first and second heterodimerization sequences are linked to cysteine residues.

19. The non-single-chain antigen-binding unit of claim 4, wherein the first heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 1 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4; and wherein the second heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 2 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 2, wherein said first and second heterodimerization sequences are linked to cysteine residues.

20. A single-chain antigen-binding unit comprising a light (L) chain variable region and a heavy (H) chain variable region connected by a first and a second heterodimerization sequence spanning the distance between the C- terminus of one of the region to the N-terminus of the other region, wherein the two regions form an intra-molecular dimer via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures.

21. The single-chain antigen-binding unit of claim 20, wherein both of the first and second heterodimerization sequences are essentially incapable of forming homodimers under physiological buffer conditions and at physiological body temperatures.

22. A single-chain antigen-binding unit comprising a light (L) chain variable region and a heavy (H) chain variable region connected by a first and a second heterodimerization sequence spanning the distance between the C- terminus of one of the region to the N-terminus of the other region, wherein the two regions form an intra-molecular dimer via pairwise affinity of the first and second heterodimerization sequences, said first and second heterodimerization sequences comprising heterodimeric receptor sequences that mediate heterodimerization of the receptors.

23. The single-chain antigen-binding unit of claim 20 or 22, wherein the first and second heterodimerization sequences form a coiled-coil dimer.

24. The single-chain antigen-binding unit of claim 20 or 22, wherein the first and second heterodimerization sequences dimerize via non-covalent pairwise affinity.

25. The single-chain antigen-binding unit of claim 23, wherein.the antigen- binding unit is conjugated to a chemically functional moiety.

26. The single-chain antigen-binding unit of claim 23, wherein the L chain variable region comprises sequences derived from a human light chain.

27. The single-chain antigen-binding unit of claim 23, wherein the H chain variable region comprises sequences derived from a human heavy chain.

28. The single-chain antigen-binding unit of claim 23, wherein the first and second heterodimerization sequences are derived from C-terminal sequences of GABA_B receptor 1 and GABA_B receptor 2, respectively.

29. The single-chain antigen-binding unit of claim 23, wherein the first heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 1 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 2; and wherein the second heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 2 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4, wherein said first and second heterodimerization sequences are linked to cysteine residues.

30. The single-chain antigen-binding unit of claim 23, wherein the first heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 1 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 4; and wherein the second heterodimerization sequence is a heterodimerization sequence comprising a GABA_B receptor 2 polypeptide of at least 30 amino acid residues that is essentially identical to a linear peptide sequence of comparable length depicted in SEQ ID NO. 2, wherein said first and second heterodimerization sequences are linked to cysteine residues.

31. A recombinant polynucleotide comprising a coding sequence that encodes L chain polypeptide of claim 1.

32. A recombinant polynucleotide comprising a coding sequence that encodes the H chain polypeptide of claim 1.

33. A recombinant polynucleotide comprising a first coding sequence that encodes the L chain polypeptide of claim 1, and a second coding sequence that encodes the H chain of polypeptide of claim 1.

34. A recombinant polynucleotide comprising a coding sequence that encodes the L chain polypeptide of claim 3.

35. A recombinant polynucleotide comprising a coding sequence that encodes the H chain polypeptide of claim 3.

36. A recombinant polynucleotide comprising a first coding sequence that encodes the L chain polypeptide of claim 3, and a second coding sequence that encodes the H chain of polypeptide of claim 3.

37. A recombinant polynucleotide comprising a coding sequence that encodes the single-chain antigen-binding unit of claim 20.

38. A recombinant polynucleotide comprising a coding sequence that encodes the single-chain antigen-binding unit of claim 22.

39. A vector comprising the recombinant polynucleotide of any one of claims 31- 38.

40. The vector of claim 39, wherein the vector is an expression vector.

41. The vector of claim 39, wherein the vector is a phage display vector.

42. A selectable library of expression vectors encoding a repertoire of antigen binding units, comprising more than one vector of claim 39.

43. The selectable library of claim 39, wherein the vector is a phage display vector.

44. A host cell comprising the recombinant polynucleotides of any one of claims 31-38.

45. The host cell of claim 44, wherein the recombinant polynucleotide encoding the L chain polypeptide and the polynucleotide encoding the H chain polypeptide, are present in a single vector.

46. The host cell of claim 44, wherein the recombinant polynucleotide encoding the L chain polypeptide and the polynucleotide encoding the H chain polypeptide, are present in separate vectors.

47. The host cell of claim 44, wherein the host cell is a eukaryotic cell.

48. The host cell of claim 44, wherein the host cell is a prokaryotic cell.

49. A method of producing a non-single-chain antigen-binding unit, comprising:

(a) expressing in a host cell a first recombinant polynucleotide encoding a light (L) chain polypeptide comprising a light (L) chain variable region fused in-frame to a first heterodimerization sequence, and a second recombinant polynucleotide encoding a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused in- frame to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures; and optionally

(b) isolating the antigen-binding unit expressed in the host cell.

50. A method of claim 49, wherein both of the first and second heterodimerization sequences are essentially incapable of forming homodimers under physiological buffer conditions and at physiological body temperatures.

51. A method of producing a non-single-chain antigen-binding unit, comprising:

(a) expressing in a host cell a first recombinant polynucleotide encoding a light (L) chain polypeptide comprising a light (L) chain variable region fused to a first heterodimerization sequence, and a second recombinant polynucleotide encoding a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences, said first and second heterodimerization sequences comprising heterodimeric receptor sequences that mediate heterodimerization of the receptors; and optionally

(b) isolating the antigen-binding unit expressed in the host cell.

52. The method of claim 49 or 51, wherein the non-single-chain antigen-binding expressed in step (a) is displayed on surface of the host cell.

53. The method of claim 49 or 51, wherein the non-single-chain antigen-binding expressed in step (a) is displayed on a phage particle.

54. The method of claim 49 or 51 , wherein the host cell is a eukaryotic cell.

55. The method of claim 49 or 51, wherein the host cell is a prokaryotic cell.

56. The method of claim 49 or 51, wherein the first and second heterodimerization sequences form a coiled-coil dimer.

57. The method of claim 49 or 51, wherein the L chain and the H chain polypeptides dimerize via non-covalent pairwise affinity.

58. The method of claim 56, wherein the L chain polypeptide further comprises a flexon that is flanked by the L chain variable region and the first heterodimerization sequence.

59. The method of claim 56, wherein the H chain polypeptide further comprises a flexon sequence that is flanked by the H chain variable region and the second heterodimerization sequence.

60. The method of claim 56, wherein both the first and the second heterodimerization sequences are linked to at least one cysteine residue.

61. The method of claim 56, wherein the non-single-chain antigen-binding unit is multivalent.

62. The method of claim 56, wherein the non-single-chain antigen-binding unit is multispecific.

63. The method of claim 62, wherein the non-single-chain antigen-binding unit is bispecific.

64. The method of claim 62, wherein the non-single-chain antigen-binding unit is trispecific.

65. The method of claim 56, wherein the L chain polypeptide comprises sequences derived from a human light chain.

66. The method of claim 56, wherein the H chain polypeptide comprises sequences derived from a human heavy chain.

67. A method of producing a non-single-chain antigen-binding unit, comprising:

(a) preparing a first recombinant polynucleotide encoding a light (L) chain polypeptide comprising a light (L) chain variable region fused in-frame to a first heterodimerization sequence, and a second recombinant polynucleotide encoding a heavy (H) chain polypeptide comprising a heavy (H) chain variable region fused in-frame to a second heterodimerization sequence; wherein the L chain and the H chain polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; and wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures; and (b) allowing the first and second polypeptides to dimerize via pairwise affinity of the first and second heterodimerization sequences.

68. The method of claim 67, wherein step (b) comprises dimerizing the first and the second polypeptides in vitro.

69. A method of producing a single-chain antigen-binding unit, comprising:

(a) expressing in a host cell a polynucleotide comprising a coding sequence that encodes the single-chain antigen-binding unit of claim 20 or 22; and optionally

(b) isolating the single-chain antigen-binding unit expressed in the host cell.

70. The method of claim 69, wherein the polynucleotide is contained in a phage display vector.

71. A method of displaying a chimeric heteromultimer comprising at least two polypeptides on a surface of a host cell, the method comprising:

expressing in the host cell

(a) a first recombinant polynucleotide encoding a first polypeptide fused in- frame to a first heterodimerization sequence and a surface presenting sequence;

(b) a second recombinant polynucleotide encoding a second polypeptide fused in-frame to a second heterodimerization sequence; wherein the first and second polypeptides dimerize via pairwise affinity of the first and second heterodimerization sequences; wherein at least one of the heterodimerization sequences is essentially incapable of forming a homodimer under physiological buffer conditions and/or at physiological body temperatures.

72. The method of claim 71, wherein both of the first and second heterodimerization sequences are essentially incapable of forming homodimers under physiological buffer conditions and at physiological body temperatures.

73. The method of claim 71, wherein the first and second heterodimerization sequences form a coiled-coil dimer.

74. The method of claim 71, wherein the first and second polynucleotides are expressed by a single phage display vector.

75. The method of claim 71, wherein the first and second polynucleotides are expressed by separate phage display vectors.

76. The method of claim 71, wherein the host cell is a prokaryotic cell.

77. The method of claim 71, wherein the host cell is a eukaryotic cell.

78. The method of claim 71, wherein the chimeric heteromultimer is a non- single-chain antigen-binding unit.

79. A chimeric heteromultimer displayed on the surface of the host cell according to the method of claim 71.

80. A method of identifying a non-single-chain antigen-binding unit that is immunoreactive with a desired antigen, comprising:

(a) preparing a genetically diverse repertoire of antigen-binding units, wherein the repertoire comprises more than one antigen-binding unit of claim 1 or 3;

(b) contacting the repertoire of antigen binding units with the desired antigen;

(c) detecting a specific binding between antigen binding units and the antigen, thereby identifying the antigen-binding unit that is immunoreactive with the desired antigen.

81. The method of claim 80, wherein the repertoire of antigen-binding units are prepared by expressing a library of vectors encoding a plurality of the antigen-binding units.

82. The method of claim 80, wherein the library of vectors comprises a plurality of phage vectors.

83. A method of identifying a single-chain antigen-binding unit that is immunoreactive with a desired antigen, comprising:

(a) preparing a genetically diverse repertoire of single-chain antigen-binding units, wherein the repertoire comprises at least one antigen-binding unit of claim 20 or 22;

(b) contacting the repertoire of antigen-binding units with the desired antigen; detecting a specific binding between antigen-binding units and the antigen, thereby identifying the single-chain antigen-binding unit that is immunoreactive with the desired antigen.

84. The method of claim 83, wherein the repertoire of antigen-binding units are prepared by expressing a library of vectors encoding a plurality of the antigen-binding units.

85. The method of claim 83, wherein the library of vectors comprises a plurality of phage vectors.

86. A kit comprising a vector of claim 39 in suitable packaging.

87. A non-single-chain antigen-binding unit of claim 1 or 3, wherein said first and second heterodimerization sequences comprise heterodimeric receptor sequences of growth factor receptors.

88. A non-single-chain antigen-binding unit of claim 1 or 3, wherein said first and second heterodimerization sequences comprise heterodimeric receptor sequences of G-protein-coupled receptors.

89. A non-single-chain antigen-binding unit of claim 1 or 3, wherein said first and second heterodimerization sequences comprise heterodimeric receptor sequences of neurotransmitters.

90. A non-single-chain antigen-binding unit of claim 1 or 3, wherein said first and second heterodimerization sequences comprise heterodimeric receptor sequences of nuclear hormone receptors.

91. A non-single-chain antigen-binding unit of claim 1 or 3, wherein the antigen- binding unit is a ccFv fragment.

92. A non-single-chain antigen-binding unit of claim 1, wherein the physiological body temperatures are at about 37°C.

93. A non-single-chain antigen-binding unit of claim 1, wherein said first and second heterodimerization sequences are essentially incapable of forming homodimers when mixed in equimolar.