WO1996006625A1

WO1996006625A1 - Antibody constructs with cdr switched variable regions

Info

Publication number: WO1996006625A1
Application number: PCT/US1995/010791
Authority: WO
Inventors: Charles R. Ill; James Richard Ludwig; Radhakrishnan Rathnachalam
Original assignee: Eli Lilly And Company
Priority date: 1994-08-26
Filing date: 1995-08-25
Publication date: 1996-03-07
Also published as: AU3415395A

Abstract

CDR grafted recombinant antibodies are provided which have at least one CDR switched variable domain wherein one or more of the heavy chain CDRs from one chain of the donor antibody are grafted into the framework regions of the light chain of the acceptor antibody. To enhance the binding of the CDRs as well as the secretion level of multi-chain constructs, the recombinant antibodies are altered using techniques of molecular modeling.

Description

Title

ANTIBODY CONSTRUCTS WITH CDR SWITCHED VARIABLE REGIONS

gacHqrpund of the Invention

This invention relates to antigen binding molecular agents useful as in vitro diagnostic or in vivo imaging and therapy agents. More specifically, the invention relates to the preparation of antibody-derived proteins useful for diagnosing, imaging and therapy of cancer, cardiovascular lesions, infections, and other pathological states. In particular, this invention relates to recombinant antibody¬ like proteins with reduced iπununogenic properties which can be efficiently expressed in eukaryotic cells.

Native antibodies are comprised of four protein chains, two shorter 'light' chains and two longer 'heavy' chains. The chains are associated in a specific three dimensional structure. Each of the four chains consists of a series of linked domain structures.

These domains are structurally related, incorporating a structural unit known as the immunoglobulin fold. Each chain contains one variable domain, encoded by a variable exon, and a number of constant domains, encoded by constant exons, the number being determined by whether the chain is heavy or light and, for heavy chain, determined by the class of heavy chain. The number of heavy chain constant domains is three for the most commonly occurring class of immunoglobulins, IgG. The constant region of the light chain consists of a single domain, Cj_,. When the variable domains are properly folded, according to the dictates of the protein sequence, the intact antibody provides a structure with specific binding properties. If the intact antibody molecule is envisioned aε a Y-shape, the stem of the Y (Fc) is formed by surface complementarity of the C_H_2, hinge, and CH_3 portions of the constant regions of the two heavy chains, which extend beyond the light chains. In addition, the two heavy chains are covalently linked through a number of disulfide linkages, the number of disulfide linkages varying between different antibody classes (i.e. IgG, IgM, IgD, IgE, igA) and subclasses (e.g. IgGj_., IgG2» G3, IgG ). The constant region of the gamma-1 heavy chain, for example, includes three constant domains, C^-i, ^u-2 _' ^an^ ^CH-3» with CH_I linked to CJJ_2 by ^an extended linker region called the hinge. The five classes of antibodies are determined in the main by their differing heavy chains - thus the IgA, IgD, IgE, IgG and IgM classes have alpha, delta, epsilon, gamma and mu type heavy chains, respectively. Each of these types of heavy chain are characterized by having generally conserved amino acid sequences in their constant domains and hinge regions, regardless of the antigen to which they bind. There are additionally two classes of light chains, lambda and kappa, the latter being more abundant in many mammalian species including mouse (ratio of kappa:lambda of 90:10) and human (ratio of kappa:lambda of 60:40). Aε with the heavy chains, each class of light chain has a generally conserved constant domain sequence regardless of the antigen to which the variable domain of the chain binds.

The variable domains are complementary, so that one heavy and light chain pair joins to form each arm of the antibody. Thus, the amino terminus of each arm contains a region (Fv) containing the antigen binding variable domains of one light and one heavy chain. Each variable domain contains three complimentarity determining regions (CDRs) characterized by highly variable protein sequences between different antibodies. Each CDR is framed by two of the four framework regions (FRs) present in each variable region, thus creating an alternating sequence of FR-CDR-FR-CDR-FR-CDR-FR- (constant domain) .

Antibody specificity and affinity are governed by the sequence and structure of the CDRs. Outside of the CDRs (i.e. within the FRs), the variable domains of the light and heavy chains have the same general structure, albeit with noticeable and functionally significant differences in sequence. The four FRs largely adopt a β-sheet conformation and are joined by connecting loops which incorporate the CDRs. The CDRs are held in close proximity by the FRs. Note that it is not always necessary to have complementary pair variable domains from one heavy and light chain to obtain binding, as is found in native antibodies. Ward, et al , , Nature. 3_£1:544-546 (1989), demonstrated that some V_H domains by themselves have the capability of binding antigens. Various forms of antibodies and antibody fragments are known for use in delivering drugs and toxins to specific sites within the body. Similarly, radiolabeled antibodies and antibody constructs can be administered in vivo for detecting and imaging or treating tumors, thrombi, infection, and other disease states. These immunotherapeutic and imaging agents target a binding site on a particular tissue or cell type, for example, a specific antigen associated with a tumor or thrombus. As a result, other tissues or cells do not accumulate the attached radioisotope, drug or toxin to the same extent. Thus, the risk of toxicity to normal tissue during systemic administration of drugs and radiolabels is considerably lessened, and concomitantly the dose of the therapeutic agent may be lowered.

Another approach in the case of antibodies for therapy and diagnosis is to use antigen binding fragments. Antibody fragments display more rapid specific targeting, less non-specific accumulation in the liver and spleen (due to the absence of the Fc portion) , and a faster rate of clearance from the blood stream than intact antibodies. Due to these characteristics, antibody fragments permit the use of radioisotopes with short half lives, such as 99m_TC/ ISrORh, _ancj _th_e like, as well as isotopes with longer half lives such as 90γ and H^m.

The greatest amount of information to date has been obtained with antibody fragments which have been produced by enzymatic digestion of antibodies, with or without chemical reduction. Digestion with papain cleaves the molecule above the hinge region, containing the interchain disulfide bonds linking the two heavy chains. The resultant fragments include two identical F_AB fragments, containing the heavy and light chain variable domains, referred to generally by the abbreviations VJJ and V_jjf respectively, the light chain constant domain, Cj_j, and the first heavy chain constant domain, CH_I, as well as a small portion of the hinge region.

When the intact antibody is digested instead with the proteolytic enzyme pepsin, the cleavage is below the disulfide bonds of the hinge region and results in a bivalent molecule having the F_AB regions from both arms linked by the disulfides in a larger segment of the hinge than in the F_AB. The resulting fragment is called an F(ab'>2 fragment. Upon reduction of the disulfide bonds, the F(ab')2 fragment produces two Fab' fragments. However, the enzymatic cleavage process often results in low yields and a significant loss of binding properties. (See Wahl, et al . , J. Nucl . Med.. 24.:317-325, 1983) . Therefore, the search continues for targeting molecules having specificity, enhanced binding activity, minimal non-specific binding, and a shorter half- life in vivo than intact antibodies. This is especially true for in vivo diagnostic (imaging) applications.

While antibody fragments have advantages for many applications, the intact antibody has advantages for many therapeutic approaches. Naked antibody therapy (i.e. therapy utilizing antibody molecules which are not coupled to drugs, radioisotopes, or toxins) often requires effector functions located in the Fc portion for action. This Fc portion is absent from most fragments. In addition, radioimmunotherapy may be more effective with intact molecules as the total dose delivered is a function of residence time at the tumor, which is uniformly higher for intact antibody molecules over fragments due to the same factors that cause fragments to be more rapidly cleared from the blood stream.

It is also possible to directly express immunoglobulin deletion mutants such as Fab or F(ab')2^_like fragmentε, using recombinant DNA techniques. In one such procedure, an Fd' fragment (i.e. the portion of the immunoglobulin heavy chain found in the F_AB' molecule) waε expresεed in E. coli (Cabilly, et al . , Proc. Natl. USA £1:3273-3277 (1984)). Ward, et al . . in "Binding Activities of a Repertoire of Single Immunoglobulin Variable Domains

Secreted from Escheria coli " , Nature 341:544-546 (1989), also describe expression of isolated heavy chain variable domain genes from E. coli to form a type of binding fragment known as a "single domain antibody." In another method described by Gillies, PCT Patent Application No. PCT/US91/00633, specific constant region domains of the human gamma heavy chain, such as the Cjj_2 domain, were eliminated to enhance the binding activity and eliminate effector functions (such as complement activation and Fc receptor binding) of the recombinant molecule over that of the native antibody.

Ideally, human antibodies and antibody fragments would be used for immunotherapy and im unodiagnosis of humans in order to avoid the undesired immune responses often caused by administering non-human immunoglobulins to them. However, human antibodies of appropriate specificity and affinity are difficult to obtain. For instance, conventional hybridoma techniques yield species hybrid cell lines that are frequently unstable and often produce IgM antibodies, instead of the more desirable IgG class of antibodies. An IgM molecule is expressed primarily as a pentamer made up of five identical subunits (IgM monomers) , each containing two heavy and two light chains. IgM monomers have, as a rule, 96/06625 PC17US95/10791

affinities that are too low for therapeutic and imaging applications. Therefore, methods utilizing genetic engineering have been developed for "humanizing" non-human immunoglobulinε. In the initial attemptε, chimeric antibodies were fashioned by replacing the entire variable domain of human antibodies with those of another species (usually urine) . (See Morrison, et al . . European Patent Application No. EP 0 173 494 and PCT Patent Application No. PCT/US91/01844. ) However, many chimeric antibodies have proven immunogenic because they still contain sufficient non- human protein sequences to generate an immune response.

In a further step towards "humanization, " the CDR' s of human (acceptor) antibody species have been replaced by those of another (donor) species, so that the framework regions and the conεtant domains are entirely or predominantly human immunoglobulin, and only the CDR portion of the recombinant antibody is non-human (See European Patent Application Publication No. EP 0 239 400 by Winter, et al . ) . These constructs, commonly known as "CDR-grafted antibodies," can also be made as antibody fragments, (See Winter, et al . , European Patent Application Publication No. EP 0 239 400 and Adair, et al . , PCT Patent Application No. PCT/GB91/01108) or as single chain antibodies (U.S. Patent No. 4,946,778 (8/7/91) isεued to Ladner, et al . , and U.S. Patent Noε. 5,132,405 (7/21/92) and 5,091,513 (2/25/92) issued to Huston, et al . ) . However, grafting of the donor CDR regions into the acceptor protein framework can displace the donor binding regions out of their optimal conformations and impair binding affinity in the resultant product. Adair et al . , in PCT Patent Application No. PCT/GB90/02017, disclose a method for restoring the CDRs to their native conformation by replacing certain key amino acid residues in the acceptor antibody framework regions to agree with those residues in corresponding regions of the native donor antibody. This procedure increases the binding efficiency of the donor CDRs but at the same time can increase the immunogenicity of the construct, since non-human residues are introduced into the human part of the construct.

In addition to the above problems in humanizing a non-human antibody, the process of producing vectors containing genomic DNA for encoding humanized antibodies has proven difficult due to the size of these human genes.

Accordingly, there exists in the art a need for more and better genetically engineered antibodies with lowered immunogenicity but with sufficient antigen-binding affinity and specificity to be useful for in vivo detection of disease, for therapy, and for a combination thereof, such as for tumor imaging and cancer therapy. The need also exists for recombinant antibodieε that are easily expressed in prokaryotic or eukaryotic host cells in commercially useful quantitieε and which accumulate in normal tiεsue in acceptably low amounts. Particularly of interest are recombinant antibodies with reduced immunogenicity (for instance a CDR-grafted antibody, and fragments thereof, comprised of human framework regions and constant domains) that bind quickly to their target sites and have other preferred pharmacokinetic properties.

Since smaller forms of antibodies, such as fragments, are less immunogenic than large intact antibodies, the combination of CDR grafting with small molecular size offers significant advantages for most in vivo applications. However, intact forms also have advantages in applications, such as radioimmunotherapy, where long residence times at the tumor are essential for maximum therapeutic effect.

Another approach to overcoming immunogenicity is the development of multiple reagents having common binding characteristicε, but different structureε. For example, uεe of different human frameworks with the same CDRs provides a different overall surface to the host immune system. More directly related to the current invention, use of frameworks from different human immunoglobulin chains provides unique molecular structures, either light chain CDRs with heavy chain frameworks or vice versa . The multiple reagents described above can be used in at least three ways. First, employing different molecular forms in consecutive rounds of therapy can decreaεe the likelihood of generating an immune reεponεe to any one form. Similarly, administering a cocktail combining various forms, decreases the amount of any individual form administered, again decreaεing the likelihood of a specific immune response. Finally, alternate molecular forms can be held in reserve, to be administered after an immune response develops to the first form administered.

There existε a need for recombinant antibodies with increased specificity. These higher specificity antibodies should be expressed from mammalian cells in order to have the proper glycosylation, and should be expressed by the cells in practical amounts. In order to impart desirable pharmacokinetic properties, it is further desirable that the recombinant antibodies be fragments of whole antibodies. Finally, it is desirable that these recombinant antibodies be as non-immunogenic as possible. This goal can be accomplished by reducing the size of the construct, by humanizing the construct to the extent posεible, and by replacing heavy chain framework regionε with light chain framework regionε.

Many of the novel molecules embraced by the present invention provide multiple small, humanized forms, which are structurally distinct from native and other recombinant types of humanized antibodies and their fragments, but conserve affinity and specificity.

SUMMARY OF THE INVENTION

The present invention encompasεeε a recombinant antibody or fragment thereof, and DNA and RNA sequences therefor, comprised of at least one light chain variable domain, which domain, in turn, comprises three CDRs wherein one or more of the CDRs is derived from [identical to or closely resemble (s)] the amino acid sequence of the corresponding CDR(s) of a heavy chain variable domain of one (donor) antibody and further comprises four framework regions wherein one or more of the amino acid sequence of framework regions are derived from the amino acid sequence of the corresponding framework region(ε) from the light chain variable domain of the same or a different (acceptor) antibody, and pharmaceutical compositions containing such antibodies or fragments.

The invention also encompasses DNA sequences encoding such recombinant antibodies or fragments thereof, and vectors containing these DNA sequences in addition to host cellε transfected by these vectors.

BRIEF DESCRIPTION OF THE FIGURES

Figure 1 is a schematic representation depicting a recombinant fragment defined herein as a CSVL fragment. In the example depicted, the CSVL fragment iε fuεed at the fragment'ε carboxy terminuε to a peptide that chelates metal ions. The illustrative CSVL fragment also conεiεts of all four framework regions from the V domain of an acceptor antibody and all three CDR regions from the V_H domain of a donor antibody.

Figure 2 is a schematic representation depicting a recombinant fragment defined herein as a Heavybody. The Heavybody consiεts of a CSV fragment and a CL domain.

Figure 3 is a schematic representation depicting a recombinant fragment defined herein as a Kappabody fragment. The Kappabody fragment has two chains: one a Heavybody and the other a CDR-grafted light chain. Preferably, the two chains are connected by a disulfide bond.

Figure 4 is a schematic representation depicting a recombinant molecule defined herein as an Intact Kappabody. This molecule comprises two heavy chains, wherein both of the heavy chain variable domainε have been replaced by CSVL fragments, and two light chains, wherein both light chains are CDR-grafted.

Figure 5 is a schematic representation depicting a recombinant molecule defined herein as an ScFy-CSV fragment. As the title implies, the Figure depicts a CSVL ra men bound by a short peptide linker to a CDR-grafted VL domain.

Figure 6 provides a linear array of the sequences of light chain variable regions of eight antibodies whose atomic coordinates have been deposited in the Brookhaven Protein Data Bank (PDB) . The identifiers uεed in this Figure correspond to PDB file names and antibody names as shown in Table 2. The sequences contained within bold boxes represent consensus SCRs. The light boxeε associated with SCR5 enclose the SCRs common only to FB4 and each individual sequence of the array. The NSCRs in each εequence are found in the εequence segments outside of (and between, except for NSCR N,l and NSCR 7,C) the bold boxes. Dots represent gaps introduced into the sequenceε in order to align the columns in the array.

Figure 7 provides a linear array of the sequences of heavy chain variable regions of eight antibodies from the Brookhaven Data Base. The Brookhaven antibodies are referred to by the identifiers of Table 2. The sequences contained within bold boxes represent consensus SCRs. The light boxes associated with SCRl enclose the SCRs common to only FB4 and each individual sequence of the array. The NSCRs in each εequence are found in the εequence εegments outεide of (and between, except for NSCR N,l and NSCR 10,C) the bold boxeε. Dotε repreεent gapε introduced into the εequences in order to align the SCRs in the array.

Figure 8 showε the sequence array of the ZCE025 light chain variable region aligned with the Brookhaven sequences shown in Figure 6. The segments of ZCE025 corresponding to the consensus SCRs are contained within bold boxes. Kabat defined CDR residues are in bold. CDR- aεεociated residues are in bold italics.

Figure 9 provides the sequence array of the ZCE025 heavy chain variable region aligned with the Brookhaven sequenceε εhown in Figure 7. The segments of ZCE025 corresponding to the consensus SCRs are contained within bold boxes. Kabat-defined CDR residues are in bold. CDR- aεεociated residues are in bold italics.

Figure 10 provides a sequence array in which the sequence of IM9 light chain variable region has been aligned with the Brookhaven sequences shown in Figure 6. The IM9 segmentε correεponding to the conεensus SCRε are contained within bold boxes.

Figure 11 depicts a sequence array in which the sequence of IM9 heavy chain variable region has been aligned with the Brookhaven sequences shown in Figure 7. The IM9 segments corresponding to the consensus SCRs are contained within bold boxes.

Figure 12 The variable region of the CSVL(HB) containing the light chain variable region of IM9 grafted with the Kabat-defined CDRs from the heavy chain of ZCE025, aligned with the heavy and light chain variable regionε of IM9 and ZCE025. Structurally homologouε regions between pairs of antibodies are enclosed by boxes.

Figure 13 shows the amino acid sequence of the IM9 light chain variable domain CDR-grafted with CDR's derived from the heavy chain of ZCE025. Lower case letters repreεent reεidues from IM9 human VK domain; upper case letters represent residues from ZCE025 murine VH domain; @ represents a glycosylation site; * designates CDR-supporting framework reεidueε from the donor antibody; $ designates residueε involved in domain association and β designates residueε that are common to both the VH domain of ZCE025 and the V domain of IM9.

Figure 14 is a restriction map of the 9 Kb BamHI fragment containing the IM9 kappa gene in bacteriophage lambda EMB 3. The Mbol termini generated by the partial genomic digest, were reconstructed as BamHI sites. The left and right lambda arms are 20 and 9 Kb, respectively. The exons are represented by solid boxes.

Figure 15 is a restriction map of pBluescript ®KS-

(commercially available from Stratagene Cloning Syεtems, 11099 North Torrey Pines Road, La Jolla, CA 92037) containing

IM9 kappa BamHI/BstEII insert from the 5' -end of the IM9 kappa gene subcloned from the 9 Kb BamHI fragment of Figure

14. The B≥tEII εite waε eliminated by filling in the 5' overhang and cloning into the EcoRV εite of pBlueεcript®KS-. The exons are represented with solid boxeε and the Amp^r gene is represented with a box.

Figure 16 is a map showing the primers for overlap PCR mutagenesiε of the IM9 kappa gene 5'-end from BamHI to BstEll. The two εetε of primerε flanking the variable exon εpecify the addition of Sfi εites on each side of the exon. The location of the Mstll site ablation is indicated 5' to the open box representing signal exon I.

Figure 17 is a restriction map of the IM9 kappa expresεion vector pGIM9kappa. Coding regionε are repreεented by εtippled boxes with arrows indicating the direction of transcription. In clockwise order from the Clal εite, the vector conεiεtε of the following fragmentε: a Clal - BamHI fragment containing the ampicillin reεiεtance gene, the SV 40 promoter, the mycophenolic acid reεiεtance gene, and the SV 40 polyadenylation site; and a BamHI - Clal fragment containing the IM9 kappa promoter, the IM9 kappa εignal exon, the IM9 kappa signal intron, the IM9 kappa variable exon, the IM9 kappa major intron, including the kappa enhancer, the IM9 kappa constant region exon, and the IM9 kappa polyadenylation site plus 3 Kb of downstream sequence.

Figure 18 showε a reεtriction map of the pGIM9k/hZCE(CSV_L) -kappa expreεsion vector. In clockwiεe order 5' to 3 ' are: the BamHI to Sfil fragment containing the IM9 light chain promoter and εignal exon; the Sfil to Sfil fragment containing the CSVL ^eχon and the 3' end of the major intron; the Sfil to Mεtll fragment containing the remainder of the major intron (including the IM9 light chain enhancer) , the IM9 Ck conεtant exon, and the IM9 kappa 3 ' untranslated region; and the Mstll to BamHI fragment containing the pSV2gpt (enhancer minus) vector. The solid boxes with arrows indicate open reading frames.

A DETAILED DESCRIPTION OF THE INVENTION

The present invention embraces genetically engineered CDR-grafted recombinant antibodies or antigen- binding fragments comprised of at least one CDR switched light chain variable domain (hereafter referred to as a "CSV-^" fragment or domain) , which domain, in turn, comprises three CDRs wherein the amino acid sequence of one or more of the CDRs is derived from the amino acid sequence of the corresponding CDR(s) of a heavy chain variable domain of one (donor) antibody and further comprises four framework regions wherein one or more of the framework regions are derived from the amino acid sequence as the corresponding framework regions(s) from the light chain variable domain of the εame or a different (acceptor) antibody. The recombinant antibodies, and the corresponding antigen-binding fragments thereof, will be referred to collectively herein as "CSV_L recombinant antibodies'". It will be understood by one skilled in the art that the CSV^- recombinant antibodies can contain CDRs and FRs from donor and acceptor antibodieε of widely divergent origins. Thus, the donor and acceptor antibodies do not have to be from the same species, and whether they are from the same species or not they certainly do not have to be of the same claεε or εubclaεε . Thuε, one could uεe a murine Ig-alpha donor antibody and a rabbit Ig- gamma acceptor antibody to conεtruct a CSVL fragment of the inεtant invention. Similarly, one could use a murine IgG-2a donor antibody and a human IgG-4 acceptor antibody to construct such a fragment.

Five types of CSV_L recombinant antibodies comprise the preferred embodiments of the present invention. The first is the CSV_L fragment itself (see Figure 1); the second iε a εingle chain derivative termed a "heavybody" (see Figure 2), which is composed of a CSV_L-containing fragment fused through the C-terminus to the N-terminuε of a light chain conεtant domain. A third preferred embodiment iε termed a kappabody fragment, which compriεeε a heavybody chain combined with a CDR-grafted light chain, preferably covalently linked by a diεulfide bridge between the two light chain conεtant domains (see Figure 3) . The latter light chain differs in general from its CSV_L counterpart in that the CDR-grafted chain has CDRs derived from a donor light chain variable domain substituted for the native CDRs in the acceptor light chain variable domain, versuε εubεtitution with donor heavy chain CDRs in the case of a CSV_L domain. A further preferred embodiment is termed an intact kappabody (see Figure 4). The intact kappabody resembleε an intact CDR-grafted antibody (with all four variable domainε having at leaεt one CDR replaced with a non-native CDR of the same type of chain (i.e. heavy or light) ; differing in that the two CDR-grafted heavy chain variable domains are replaced by two CSVL domains. The fifth preferred embodiment is termed a single chain chain-switched variable fragment and is defined aε a CSV_L domain bonded to a CDR-grafted light chain variable domain throughout a εhort peptide linker, generally no more than 25 amino acid reεidueε (εee Figure 5) . The εymbol uεed in the Specification for this embodiment is "SCFV(CSVL) "• The C-terminal end of the CDR-Grafted VL domain can be fused to the N-terminus of the CSVL domain through the peptide linker, or vice versa. As with the CSVL fragment, one skilled in the art will realize that the Heavybody, the kappabody fragment, the intact kappabody and the SCFV(CSVL) fragment offer a wide array of choices for donor and acceptor antibodies. Thus, taking the heavybody as an example, the donor antibody could be a murine lgAι_, the Framework Region(s) and the CL could be from a sheep IgM acceptor antibody. Taking thiε principle one εtep further, in the case of an intact kappabody, the present invention contemplates the expresεion of a molecule having one lambda and one kappa chain, regardleεε of whether they were of the εame species, or a molecule having two kappa or two lambda chains of different specieε. To inεure proper diεulfide bridging, heavy chain acceptor antibodieε of an intact kappabody are preferably of the εame species, class and subclasε. The five illustrative generalized preferred embodiments have several common, more preferred embodiments. For instance, it is preferred that the donor and acceptor antibodies for these five constructs have donor and acceptor antibodies that are different and that are chosen from murine, rabbit, or primate monoclonal or antibodies.

Furthermore, it is preferred that all of the CDRs in the various CSV-^ and CDR-grafted V_L domains, as the case may be, are identical in amino acid sequence to the corresponding CDRs of donor antibody CDRs; that all of the framework regions are derived from the same amino acid sequence as,

(i.e., being at least about 75% and preferably at least 85% homologous to) the corresponding framework regions of the acceptor antibody(ies) ; and that any constant domains, whether light chain or heavy chain, as the case may be, are identical in amino acid sequence to the corresponding domains of the acceptor antibody(ies) . In order to make these preferred constructs less immunogenic, it is further preferred that the acceptor antibody (ies) be human, especially a human antibody that has light chains of the kappa class, and more so when the human heavy chains are of the gamma class. (It is understood that the class and subclaεε of the two heavy chains in an intact kappabody are preferably the same in order to obtain optimal disulfide bridging between the two chains.) With regard to the SCFV(CSVL) fragment, when the acceptor antibody iε human, it iε preferred that the linking peptide be from about 12 to about 18 amino acid residues, and especially so when the CDR- grafted V_L domain is fused to the N-terminus of the polypeptide linker, and wherein the C-terminuε of the polypeptide linker iε fuεed to the N-terminuε of the CSV_L domain.

Further preferred embodimentε of this invention occur when murine monoclonal antibodies are used as donor antibodies, and more so when these murine antibodies have binding affinity, and thus were raised against, tumor antigens and antigens on thrombi; but especially so for tumor antigens of human but alεo of any other vertebrate origin. Preferred tumor antigens; (or markers as they are sometimes called), are AFP, CA-125, CEA, Neuron Specific Enolase, C-erb2/Her-2/NEU protein, Cathepsin D, Chromagranins A, B, and C, the Cytokeratins, Epidermal Growth Factor Receptor,

Epithelial Membrane Antigen, Estrogen Receptor, Progesterone Receptor, Prostatic Acid Phoεphataεe, Proεtate Specific Antigen, Ki-67, PGP-170 (a multiple drug resistance marker) , Proliferating Cell Nuclear Antigen, Vimentin, and the proteins expressed by the c-myc, N-myc, N-ras, Ki-ras and Ha-ras oncogeneε. An eεpecially important tumor antigen iε CEA, with preferred murine donor antibodieε being the anti- CEA antibodieε ZCE 025 (C. M. Haskell, et al . , Cancer Research. 41,3857 (1983), who refers to the antibody as "MAB 035") and CEM 231 (C.B. Beidler, et al., J. Immunology.

141, (11), 4053 (1988)) . Regarding the ScFv(CSV ) fragment, when the donor murine antibody is an anti-CEA antibody, it is further preferred that the peptide linker be composed of εerine and glycine reεidues.) With the latter two anti-CEA donor antibodies, it is preferred that the acceptor antibody be the human IM9 antibody, (Reference under IM-9 in ATCC #159) wherein the framework regions in the CSVL ^an^ CDR- grafted light chain domains, aε the case may be, are moεtly the same in amino acid sequence as the corresponding IM9 framework regions. The most preferred donor antibody is ZCE025. Finally, with regard to the ScFv(CSV_L) fragment, when the donor antibody is ZCE025, it is preferred that the peptide linker have the amino acid sequence -GGSGGSGGSGGSGG-

(Sequence I.D. No. 1) .

Each of the above five preferred embodiments can optionally have fused to its C- or N- terminus a metal- chelating peptide sequence. The chelating peptide sequence can be up to about twenty-five amino acid residues in length.

In the case of the CSV_L and the ScFv(CSV_L) fragment, only one such peptide chain is bound to either available terminus. In the case of the kappabody fragment and the heavybody, the chelating peptide can be bound to either one or the other, or both, chains, and when bound to both chains, can be bound to either the N-termini, the C- ermini, the C-terminus of one chain and N-terminus of the other, or to both termini of both chains. With the intact kappabody, a chelating peptide such as that described above can be bound to any number of the four chains comprising the molecule, with any and all combinations of N-termini and C-termini bonding envisioned. For any one of the five preferred constructε, it is further preferred that metal chelating peptide consist of about ten amino acid residues or less and chelate to either nickel(+2), zinc(+2), copper(+2), or cobal (+2) ions and be bonded to one or more, as the case may be, of the c-termini of the molecule. More preferred is the case where one (or more) of the C-termini is fused to a metal chelating peptide of the sequence HWHHHP (Sequence I.D. No. 2) through the peptide's N-terminal histidine residue. Regarding the preferred embodiment of any of the five preferred constructs as described above, when the embodiment iε narrowed to a murine donor antibody, it iε preferred that a metal-chelating εpecieε be bonded to the C-terminuε (or possibly more than one termini , as is applicable) , consiεt of ten or less amino acid residueε, and chelate with either nickel (+2), copper (+2), zinc (+2) or copper (+2) ions. Finally, in the most preferred embodiment of the above five constructε , that is, wherein the donor antibody is the murine monoclonal antibody ZCE 025, it is preferred that the optional metal chelating peptide have the sequence HWHHHP and be fused to the C-terminal (or one or more termini , as is applicable) of the molecule.

The present invention also comprises the RNA and DNA sequenceε coding for any molecule therein, including but not limited to the five preferred constructs and their corresponding preferred embodiments.

The present invention also compriseε antigen- binding fragmentε of any of the above moleculeε that can be obtained by routine chemical and enzymatic manipulation, εuch aε the fragmentε resulting from the chemical cleavage of bridging disulfide bonds, (e.g. using 2-mercaptoethanol and iodoacetate) , and from enzymatic digestion with routine reagentε εuch as pepsin and papain. For instance, it is within the scope of the present invention to have an F(ab')2 fragment obtained from the digestion of an intact kappabody, and any of its preferred embodiments described above, with pepεin, or an Fab fragment obtained from the digeεtion of it with papain.

The CSV_L recombinant antibodieε of the present invention contain one or more heavy chain CDR(s) from a donor antibody grafted into a kappa or lambda chain variable domain. The immunoglobulin chain containing the CSV_L can further contain either a kappa or lambda constant region, or one or more alpha, delta, epεilon, gamma or mu constant region, depending upon its intended use. As mentioned above, gamma constant regions are preferred for this invention, and eεpecially preferred are the conεtant regionε of the gamma-1 εubclass.

As used to define and delineate the scope of the present invention, the term "CSVL recombinant antibodies" shall mean both a CSV fragment and a CSVL~containing antibody or fragment thereof, including a Heavybody, an ScFv(CSV_L) fragment, an Intact Kappabody, and a Kappabody fragment.

In the various constructε of the present invention the antibody that provides the framework regions into which are grafted CDRε from another antibody iε referred to as the "acceptor antibody." The antibody that provides the CDRs grafted into the acceptor antibody is referred to as the "donor antibody". In one embodiment, the amino acid sequence in the four framework regions of the acceptor antibody are subεtantially homologous (i.e. at least about 75% homology) to the corresponding regions of the native acceptor antibodies. In another embodiment, the protein sequences in the framework regions of the acceptor antibody are altered, for example, by means of computer modeling, to preserve certain amino acidε from the donor antibody that are neceεsary to conserve the binding affinity of the CSV_h domains and the CDR-grafted light chain domain and the ability of the hybrid immunoglobulin chains containing the altered variable domains to associate and assemble with other such immunoglobulin chains into antibody-like constructs.

Since a single alteration in the protein sequence of a CDR can subεtantially decrease the binding affinity of the construct for its antigen, the grafted CDRs are preferably homologous to those of the donor antibody; however, it is intended that one or more residues of a donor CDR can optionally be changed or omitted. The donor and acceptor antibodies can be polyclonal or monoclonal and can be of any antibody class or species. Preferably, however, the acceptor light chains are derived from a human antibody, most preferably IgG, and the CDRε are derived from a donor antibody from a non-human species selected from the group consiεting of rodent, rabbit, and primate antibodieε. Human donor antibodies may also be used and in one embodiment of the invention the CSVL recombinant antibodies are made using the same antibody as both donor and acceptor, i.e., the heavy chain CDRs are grafted into a kappa light chain and asεociated with a native kappa light chain to make an engineered light chain dimer fragment.

A SV-^ recombinant antibody may have attached to it an effector or reporter molecule. For instance, a macrocycle or chelating peptide may be attached for chelating a heavy metal atom. Similarly, a toxin, such aε ricin, can be attached to the recombinant antibodieε of thiε invention by any of a number of covalent binding structures known in the art. Alternatively, a fusion protein comprising a CSV_L recombinant antibody joined by a peptide linkage to a chelating peptide or functional non-immunoglobulin protein, such as an enzyme or toxin molecule, can be produced using the procedures of recombinant DNA technology, for instance, the general methods of Neuberger, et al . , in PCT Patent Application No. PCT/GB85/00392.

The term "antigen" aε uεed herein shall encompass large protein antigens, such as carcinoembryonic antigen, in addition to haptens, εuch aε metal-binding haptenε. The ability to bind with an antigen or hapten is determined by assays well known in the art, such as antibody capture assays (See, for example, Harlow and Lane, Antibodies. A Laboratory Manual . Cold Spring Harbor Laboratory, Cold Spring Harbor,

New York (1988) ) . The CSV_L recombinant antibodies are made using techniques of genetic engineering that are well known in the art. (See for example European Patent Application EP 0 239 400 to Winter, et al . , PCT Patent Application PCT/GB91/-1108 to Adair, and U.S. Patent Nos. 5,132,405 and 5,091,513 to Huston, et al . ) The terms "CDR grafted", "grafted with", and "grafted into", and the like, as used herein shall have the meaning well known in the art that, using the techniques of genetic engineering, in one antibody, called the acceptor antibody, the CDRs are removed and replaced with those of another antibody, usually of another species, called the donor antibody. In the CSVL recombinant antibodies taught herein a CDR from the donor antibody can be grafted into a CDR locus in the acceptor immunoglobulin other than the one from which it is derived in the donor immunoglobulin. That iε, CDRl in the acceptor immunoglobulin can be replaced with CDR2 or CDR3 from a donor antibody, and εo forth. The CSVj_^ recombinant antibodies may comprise only one or two donor- derived CDRs, though preferably all three CDRs are derived from the donor antibody and are grafted into the acceptor frameworks so as to replace the native CDRs therein, i.e., donor CDRl of the opposite chain is grafted into the locus of CDRl in the acceptor immunoglobulin chain. As used herein the terms "CDR" and "framework region" shall have the meanings and their locations shall be determined according to the method of Wu and Kabat, J. Exo. Med. 132:211-250 (1970), unless crystallographic analysis or homology modeling dictate that they have slightly modified locations.

As used herein the phrase "derived from" and "altered" shall encompass the meaning that certain amino acids (lesε than or equal to 25% and preferably less than or equal to 15% of the total amino acid residueε) in the acceptor framework regions of the CDR grafted constructs are switched to match the corresponding amino acids from the donor antibody as needed to facilitate the dual goals of preserving the binding affinity of the donor antibody and the expression levels of the acceptor antibody. The CSVL recombinant antibodies of this invention can be engineered to have the size, function and general design of an intact antibody or of any antibody fragments, such as Fv, Fab' , single chain Fv, or single domain antibody (for example, an isolated heavy chain variable region), so long as each contains at least one CSVL domain. The CSVL recombinant antibodies can be labeled for use in in vivo diagnoεiε and therapy. For inεtance radioactive ions having suitable properties for use in n vivo regimens can be attached to the recombinant antibodies under conditions similar to those known in the art.

In the kappabody (Figure 3), the two chains are joined by one or more, preferably one or two, sulfhydryl bridges at the C-terminus of the light chain constant domain. In the native kappa chain, there is one sulfhydryl bridge to the heavy chain, but additional sulfhydryl-bearing cysterne residueε could be added by incorporating all or part of the hinge region of an IgG heavy chain or by fusing an appropriate metal-binding protein containing cysteine.

Kappa and lambda dimer fragments occur in nature and result from spontaneous combination of light chains within the host cell upon expression. Like these naturally occurring light chain dimer fragments, those of the invention asεociate naturally within the hoεt cell and are held together by weak bonding interactions between the two chains, (i.e., hydrogen bonding and Van der Waalε forceε), by a spontaneouεly formed diεulfide bridge at the C terminuε of the chainε, aε well aε by any natural forceε of attraction of the heavy CDRε for the light CDRs. Unlike the naturally occurring light chain dimer fragments, however, it iε believed that the CSVL recombinant antibodieε of the invention may experience diεlocation of εome of the εites of weak bonding interaction in the kappa chainε (aε compared to native kappa dimer fragmentε) due to strain caused by the εplicing of foreign CDR'ε into the acceptor kappa chains. Therefore, in the kappabody fragments of the present invention certain residueε in the acceptor framework regions holding the donor CDRs are preferably altered to overcome the effects upon affinity and εpecificity of the foreign CDR[ε) and to ensure the ability of the engineered proteins to properly assemble upon translation. Theεe small (50 kd) , humanized molecules offer several advantages over Fab antibody fragments. First, they are readily expresεed from the same vector due to uniformity of the two chains, thus allowing for rapid construction and more equivalent expresεion of both chains. Second, since they are recombinantly expressed, the native carboxy-terminus is present on both chains; whereas fragments created by treatment of whole antibody with enzyme lack the native terminus and therefore can be more immunogenic. And third, these molecules, which have a structure distinct from Fab antibody fragments, are expreεεed at high levels and are highly stable.

It is known that during trafficking of immunoglobulin proteins within the eukaryotic cell, the heavy chain binds to the chaperon protein complex Bip/GRP94 located within the rough endoplasmic reticulum, and is thereby prevented from passage into the Golgi apparatus and thus is prevented from expresεion by the cell. A heavy chain is not secreted in eukaryotic cells unless or until it is displaced from the chaperon protein by a light chain, with which the heavy chain combines, thereby leading to secretion of intact antibody. For potentially similar reasonε, a chimeric construct comprising the variable domain of a heavy chain and the constant region of a light chain (i.e., a VHC^ fragment) will not be secreted by itself in mammalian host cells.

However, the instant invention discloses that a genetically engineered gene encoding a CSVL fragment, when operably linked to the required transcriptional and translational sequences functional in eukaryotic host cells suitable for expression of immunoglobulin genes, will be transcribed, translated and secreted. The secretable CSVL can be incorporated into constructs that also contain a light chain constant region and will convey upon the resulting the similar ability to be secreted in eukaryotic cells. Indeed, just such a single chain fragment has been mentioned above as a preferred embodiment of this invention. A species of this "heavybody" fragment is depicted in 96/06625 PC17US95/ 10791

- 24 -

Figure 2. As with the kappabody fragment, two different acceptor antibodies can be used, one for the light chain constant domain and another for the FRs, although it iε preferred that the same acceptor antibody be used for both areas. As illustrated below, the heavybody fragment iε εecreted in mammalian cellε aε a homodimer (an assembly of two identical chains) in the absence of the expresεion of native light chain by the hoεt cell. However, in the preεence of light chain, a light-heavy heterodimer (an aεεembly of two εignificantly different chainε) iε preferentially formed. The binding affinity of the heavybody homodimer can readily be aεεayed, uεing methodε known in the art, such as a competition ELISA.

Unlike isolated native light chains of antibodies or native light chain homodimers, which do not poεsesε binding affinity by themselves, the instant heavybodies (i.e., the εingle chain monomer) retain the ability to bind antigen. If it iε desired to assay the binding affinity of an isolated heavybody, the sulfhydryl bridge(s) that join the chains of the heavybody homodimer can be reduced by treatment with enzyme under conditions mild enough to preserve the binding affinity of the isolated heavybody monomer using techniques well known in the art, or as is illustrated in the Exampleε. The heavybody iε a very small (25 kd) humanized molecule of different structure from a native kappa or lambda chain. And, unlike a chimeric heavy chain, the heavybody molecule is secreted from mammalian cells with high levels of expresεion.

As one skilled in the art will appreciate, the present invention enables production of recombinant antibodies of smaller size. For instance, fragments analogous to Fv fragmentε can be made from the variable do ainε of two acceptor light chains by grafting at least one light chain CDR into one copy of the light chain variable region and at least one heavy chain CDR into another copy of the light chain variable domain of the donor antibody. Like Fv fragmentε, these smaller constructs lack the natural sulfhydryl bridge that connects naturally occurring kappa dimers (and would leave out the light chain conεtant domain.) It is also possible to rapidly engineer and secrete in mammalian host cells at high expression levels a single domain construct comprising an acceptor light chain variable region with one or more donor antibody heavy chain CDRs grafted between the framework regions referred to herein as a CSV_L fragment (aε diεcuεεed above) , aε illuεtrated in

Figure 1 and deεcribed further in the Exampleε. Thiε (embodiement of the preεent invention) further evidences that the DNA εequenceε effective for expression of the heavybody fragments in mammalian cells are contained in the framework regions of the light chain variable domain.

Not all antibodies or fragments with useful affinities for their antigen have heavy chain variable domains with sufficient affinity to bind with the antigen. However, by proper screening of the genome of a lymphoid cell, a heavy chain variable domain having CDRs with sufficient antigen affinity to bind as a single domain antibody can be found using techniques well known in the art. For instance, Ward, et al . , in "Binding Activities of a Repertoire of Single Immunoglobulin Variable Domains Secreted from Escherichia coli , " Nature 341:544-546. (1989) disclose a method for screening to obtain an antibody heavy chain variable region (VH single domain antibody) with sufficient affinity for its target epitope to bind thereto in the single domain format.

Alternatively, a phage expression library can be prepared from VH DNA fragments using methods well known in the art. (See for instance, Garrard, L.J., et al., PCT Patent Application PCT/US91/09133, assigned to Genentech. Proteins expressed on the phage head can be screened using an affinity column having bound antigen or a polypeptide probe constructed from the peptide sequence of the desired target epitope or antigen. Single domain VH antibodies that bind with the antigen can be selected and ranked to obtain those with the highest affinity for the antigen. These single domain VH antibodieε, however, cannot be secreted in mammalian cells. By adopting these and other techniques, εuch as molecular modeling, known in the art and/or discloεed herein, heavy chain variable domains showing antigen-binding affinity can be obtained and used as the donor antibody to make a single domain CSV_L~containing recombinant antibody fragment according to the present invention, i.e., having one or more CDRs from a high affinity donor heavy chain variable domain grafted into the framework regions of a acceptor light chain variable domain, and preferably wherein the acceptor antibody haε kappa light chains and is of human origin.

As illustrated below in the Exampleε, a preferred embodiment of the εingle domain fragment of this invention, namely the CSVL fragments, can be expressed in mammalian cells. In contrast, a conventional single domain antibody, (i.e., one consisting of a VH domain) cannot.

With a molecular weight of 12.5 kd, approximately one sixteenth that of intact antibody, the CSVL fragmentε of the invention bind to target antigen with the specificity of the donor antibody, and with the potentially greater binding ability than the variable domain of a light chain alone. Yet these extremely small peptides will clear from the circulation more rapidly with decreased normal tissue retention and decreased immunogenicity, and penetrate tumor more extensively than any other size of antibody fragment. Even when the framework sequences of the CSVL fragment have been altered in accordance with this invention to facilitate folding of the molecule into a three-dimensional geometry that provides the specificity and a sufficient affinity for use in in vivo imaging and therapeutic applications, the CSVL fragment proteins are generally approximately thirty to thirty five percent human when three non-human CDRε have been grafted into them. Therefore, theεe very εmall recombinant fragmentε, which can be rapidly engineered to improve affinity or εpecificity due to their small, single chain format, are particularly useful for in vivo applications that require rapid clearance of the unbound binding fragment from the blood, εuch as in vivo radiotherapy using strong beta- emitting particles attached to the binding fragment.

As mentioned above, insertion of donor CDRs into the acceptor framework regions can displace the CDRs out of their preferred spatial alignment. Association siteε between the heavy and light chainε can also be disrupted by introduction of the foreign CDRs so that the expression level of the CDR grafted construct is impaired relative to that of the intact acceptor antibody. In the CSV-^ recombinant antibodies of the present invention, an additional problem is encountered. Grafting of heavy chain CDRs into light chain framework regions in the making of a SV-^ , can produce either different or additional diεlocationε of the εites in the framework regionε that are necessary to support the CDRs in their preferred spatial orientations and dislocationε of the association siteε between the light and heavy chains that contribute to assemblage of the recombinant antibody chains during expression.

To accomplish the dual goals of (1) preserving the spatial orientation of the CDR loops as it appears in the donor antibody, and (2) maintaining to the greatest extent possible the expression levels and reduced immunogenicity of the acceptor antibody, any of a number of available methods based on computer-assisted molecular modeling procedures can be used or modified for effectively identifying and replacing amino acids in the acceptor framework regionε to create CSV_L recombinant antibody of this invention.

For instance, Adair, J., et al., PCT Patent Application PCT/GB90/02017, assigned to Celltech, disclose a method for introducing mutations into acceptor framework regions of CDR-grafted antibody chains of anti-CEA antibodies to match the corresponding donor residueε. In the Celltech method, in addition to the Kabat-defined CDRε from the donor antibody (CDRl: poεitions 24-34; CDR2: positions 50-56; CDR3: positionε 89-97) the structural loop residues (positionε 89- 97) in CDR3 and residues at one or more of positions 1, 2, and/or 3, 46, 47, 49, 60, 70, 84, 85 and 87 are replaced by the correεponding donor residues, if they differ.

According to the Celltech method, in the heavy chain, in addition to the Kabat-defined CDRs, the amino acid residues of the acceptor variable domain are replaced at poεitionε 23 and 24 and 71 and/or 73 with thoεe of the donor antibody, if they differ. Additionally, in the heavy chain, the acceptor residues can be replaced by donor residues at some or all of positions 48 and/or 49, 69, 76 and/or 78, 80, 88 and/or 91 and 96. The definitions of the CDRs can alεo be εhifted to accommodate idioεyncratic regionε in any given donor antibody.

Ideally, commercially available computer programε are used with actual crystal structureε of the donor and acceptor antibodieε (bound to their antigenε) to determine which amino acidε in the CDRε (and framework regionε) contain atoms that are close enough to atoms in the amino acids of the antigen to interact.

Yet another method, generally referred to as homology modeling, is uεeful when a crystal structure cannot be obtained for the antibody to be used in making the antibodieε of thiε invention. Several fully automated algorithmε to align cryεtal structures and define structurally conεerved regions are known. The loop regions are modeled by two basic methods: 1) use of a data base of available structures to provide the best possible loop conformations or, 2) use of distance-geometry based mathematical model to generate further poεεible conformerε. The beεt conformer choεen by either method of modeling iε choεen on the baεiε of εome type of energy function, uεually an energy calculation. For instance, computer programs such as Insight II, Homology and Discover (Biosym, San Diego, CA) are employed in conjunction with a database containing the known cryεtal εtructureε of proteinε, such as the Brookhaven Protein Data Bank, to construct a three dimensional representation of the immunoglobulion of interest. This three dimenεional repreεentation iε baεed upon homology between structurally conεerved regionε (SCRε) of the known structures and corresponding regions in the protein whoεe cryεtal structure is unknown. A loop search algorithm is used to identify protein loops from the database with the right number of residues and correct three dimensional disposition of backbone atoms of the regions flanking the loop to splice between the structurally conserved regions. In this way the three dimensional model of the donor and acceptor antibodies iε conεtructed by the computer so that the residueε of the donor antibody frameworks that are involved in εupporting the CDRε and the reεidueε of the acceptor antibody frameworks that are involved in chain association can be conserved in the CSVL recombinant antibody. The preferred method of making the CSVL recombinant antibodies of this invention, when actual crystal structures of the donor and acceptor antibodies are not known, employs molecular modeling. Molecular modeling can be used to locate the three dimensional structurally conserved regions (SCRs) common among all antibodies. Separate computer modelε of the donor and acceptor immunoglobulinε are constructed by a technique of homology modeling based upon a database of known protein crystal structures, such as the Brookhaven Protein Data Bank of known protein crystal εtructureε, uεing the computer modeling programε Inεight II, Homology and Discover, Version 2.1.2. From computer models of the donor and acceptor antibodies, the amino acid residues in each structure involved in association of the immunoglobulin chains in the acceptor antibody are determined and conserved in the CDR grafted construct. In addition, the amino acid residues involved in support of the CDRs in the donor antibody are conserved in the CSVL recombinant antibodieε.

Briefly, for the purpose of modeling the light chain variable region of an antibody, at least two and preferably at least eight antibodies are selected from a protein database, such as the Brookhaven Protein Data Bank, that provides both a linear amino acid sequence and three dimensional atomic coordinates of each antibody variable region. The sequenceε and structures of these antibodies are manipulated by a computer program having the ability to assign the corresponding atomic coordinates from a segment of a known structure to the atomε of any εegment of an amino acid εequence having the εame number of residues. One skilled in the art will know of computer programs and databases that are suitable to work in tandem in thiε faεhion. For example, the Brookhaven Protein Data Bank can be uεed together with the current verεionε of molecular modeling programs Insight II, Homology and Discover (Biosym Technologies, Inc., San Diego, CA) ; aε discussed in the immediately following sections.

Step 2S£ - Definition Q£ S ruc urally Conserve^

Regions UBinσ Known Three-pime gipna;|.

Structures Q£ Antibodies

Using the selected three dimensional protein structureε and εequenceε from the database, the operator uses the computer program to align the sequences of the variable regionε and to superimpose the corresponding structures so that structurally conserved regions can be identified. For instance, the sequenceε are aligned in a linear array, with each εequence constituting one row of the array, i.e., Seq a, Seq b, Seq c, etc.

To facilitate alignment by placing the SCRs into columns using the Insight II software, certain landmark amino acids known to be universally conserved among antibodieε, such as the cysteineε that form the intrachain diεulfide bridge, are identified in each sequence and are aligned in vertical columns. Taking the first two of the linearly aligned sequenceε, one, for inεtance Seq a, is designated to be held constant and the other, for instance Seq b, to be superimposed onto the first. (In practice, the bottom εequence on a computer diεplay iε usually most convenient to hold constant. )

Three-dimensional alignment of the known εtructureε iε further refined by diεcovering additional amino acid εequenceε that correspond to regionε in all the εelected antibodies that preserve almost identically the same three- dimensional conformation, called herein the structurally conserved regions (SCRs) .

Using the already superimposed structures, the first putative SCR, conveniently designated SCRlab, is discovered by visual inspection. Preferably, succesεive SCRs are identified by working from the amino to the carboxy terminus of the molecules. The RMS deviation of the backbone atoms in the two segments of amino acids corresponding to SCRlab is calculated. The exact locations of SCRlab, and hence of the amino acids contained within the segmentε correεponding to the SCRlab, are adjusted by a procedure of trial and error whereby the amino acids in the linear εequenceε of the array that correspond to those in the putative SCRlab are boxed and the RMS deviation iε calculated. The width of the box is maximized and the location of the box is adjusted until the RMS deviation reaches an acceptable maximum, for instance no more than about 0.75 Angstromε. To ensure that spatial alignment of SCRlab at the amino terminus of firεt and second εtructures is not destroyed by establishment of subsequent SCRs along the sequenceε (i.e., SCR2ab and SCR3ab, etc.), preferably after the proceεε has been carried out to define SCR2ab, the two structures are superimposed again using the residueε for the backbone atomε in SCRlab aε well aε SCR2ab. Thiε proceεε iε repeated for each subsequent SCR. Gaps, for example, empty space holders, can be inserted within nonconserved (nonhomologous) regions, referred to herein as NSCRs. Usually the NSCRs are found in the loops and CDRs. Gaps are inserted as needed to accomplish vertical alignment of the SCRε, for example, where any εequence had fewer amino acidε between the SCRε than did the other.

Uεually by the method of thiε invention, from about seven to ten SCRs are established between any two light chain variable regions, with each SCR containing from about three to twenty amino acids from each of the structures, when the RMS deviation of the backbone atomic coordinateε in the SCRε iε no more than about 0.75 Angεtromε.

Once two of the εtructureε have been aligned in thiε manner, the procedure iε repeated, preferably by εelecting the firεt εtructure, for inεtance the bottom εtructure in the array, to be held conεtant (Seq a), and discovering the SCRs between that first structure and each in turn of the other structureε represented in the linear array (Seq b, Seq c, etc.) to yield SCRlac, SCR2ac, SCR3ac, etc. and then SCRlad, SCR2ad, SCR3ad, etc. Alternatively, of course, any other method can be used whereby segmentε having a common εpatial conformation, εuch aε SCRs, are located within the known three dimensional structures of from six to ten antibody variable regions. For instance, one skilled in the art will appreciate that it would be posεible to locate the firεt SCRε in the middle of the moleculeε and work outward therefrom in either direction, or to begin at the carboxy terminuε of moleculeε and work progreεεively towards the amino terminus. The order in which the sequences (and their structureε) are compared with one another can alεo be varied. For inεtance, one skilled in the art will appreciate that it would be possible not to hold a first structure constant, and instead to align any two structureε and then to choεe any one of thoεe two εtructures to be aligned with a third, and so forth.

When all of the structures have been compared with one another by any of the alternative methods described above, for instance when each structure in the array has been in turn superimposed and aligned with the constant first structure, as iε preferred herein, the next εtep is to identify the consensus SCRs. A conεensus SCR compriseε the residues in each linear sequence that are in the intersection of all of the individual SCRs. One skilled in the art will appreciate, however, that the technique of locating the consensus SCRs can be varied so long as structurally conserved regions (SCRs) common to all of the structures in the array are located, and so long as the RMS deviation of the coordinates corresponding to the superimposed backbone atoms in all of the structures is acceptably low, for inεtance no more than 0.75 Angεtroms. A similar procedure is followed to locate and fix in spatial relation to one another the SCRs common to the heavy chain variable domains of antibodies, except that the sequenceε uεed in the linear array are those of the heavy chain variable domainε of the antibodieε in the database whose three dimensional structures are known.

££££ £ £ - Three-dimensional Modeling of Acceptor fv and Identification ^of Chai

Associat on Residues

Now the linear sequence of the acceptor antibody chain to be modeled is displayed as an additional row in the linear array and aligned with the sequences of the eight database antibodies as described above to discover the segmentε of SCRε in the acceptor chain that correspond to those in each of the boxes, using as many gap-filling spaces as needed to accomplish the vertical alignment. This procesε iε identical for light and heavy chainε. The three- dimenεional model of the acceptor antibody chain can now be fabricated in segments from the consensus SCRs derived above. For each SCRl in the linear sequence of the acceptor antibody chain, the column of SCRls in the array is inspected to find the SCRl with greatest sequence homology to the acceptor SCRl. The computer is used to construct the model of the acceptor SCRl by assigning to each residue in the acceptor SCRl coordinates corresponding to those of the selected sequence from the column of corresponding database SCRls. At thiε point, any reεidue in the εelected SCRl that doeε not match the correεponding reεidue in the donor SCRl is mutated to match the reεidue in the acceptor SCRl, while the coordinateε of all the atoms in the backbone and sidechains that correεpond to those in the acceptor residue are conserved. The remaining atoms are modeled under the constraintε of maintaining the εame bond lengthε, angles and dihedralε aε thoεe in the original databaεe residue, i.e., for the gamma and delta carbons. The process is repeated for each of the εubεequent SCRε, i.e., SCR2, SCR3 , etc.

Next, the length of each εegment of NSCR in the acceptor chain εequence, i.e., the εpanning εequence between each εucceεεive pair of boxeε, iε determined. Progreεεively from the amino terminuε of the chain, NSCR segments of the acceptor chain are modeled by selecting loops from the protein database to span between the endpoints of the SCRs of the acceptor chain model constructed above. The actual number of amino acid residues in each NSCR is counted (ignoring the space-filling gaps used to accompliεh vertical alignment) . For each span individually, the computer is instructed to εearch the protein databaεe, for instance using the Loop Search algorithm as is well known in the art, to discover from about eight to twelve candidate amino acid sequenceε having (1) the same number of amino acids as the actual acceptor NSCR and (2) flanking regions with the εame relative atomic coordinates as the flanking SCRs in the acceptor chain model as determined above. As one skilled in the art will appreciate, depending on local εtructural detailε, either all or εome subset of the residues adjacent to the loop in each SCR box can be identified as the flanking residueε. The candidate εequenceε whoεe flanking regionε are best fits with the relative atomic coordinates of the SCRε of the acceptor chain model, aε determined by computer algorithm, are εelected.

It has been discovered that in antibodies the general spatial conformation of the loops and NSCRε iε conεerved. Therefore, the beεt candidate for NSCR1,2 in the model should have a three dimensional spatial conformation generally similar to that of the corresponding NSCR1,2 segments in the antibody database structures. For each candidate that meets this general requirement, the backbone atoms of the flanking residues of the candidate NSCR are superimposed on the backbone atoms of the corresponding flanking reεidueε of the SCRε of the model that flank the NSCR under consideration. For example, to consider the candidates for the NSCR1,2 position in the model, the backbone atoms of the flanking residues of the candidate NSCR are superimposed on the backbone atoms of the corresponding flanking residues of the SCRl and SCR2 sequenceε from the model, and the candidate having (1) the beεt RMS fit of the backbone atomε of its flanking residueε with backbone atomε of the correεponding flanking residues from SCRl and SCR2 and (2) a spatial orientation most like that of NSCRls of the database antibodies displayed on the computer screen (to rule out interference with other loops) is selected. By repeating this procedure at each NSCR position, i.e. at NSCR1,2;

NSCR2,3; NSCR3,4, etc., the acceptor NSCRs are selected and then placed into the acceptor model as follows.

Once the best spatial orientation for an amino acid sequence of the given loop length for each NSCR iε εelected, the coordinateε of the backbone of the candidate εegment are aεεigned by the computer to the corresponding NSCR in the model. Now any residue in the selected candidate εequence NSCR dissimilar to the corresponding reεidue in the actual εequence of the acceptor NSCR is mutated to match the acceptor εequence while the computer algorithm iε uεed to (1) maintain the coordinates of all the atoms common between the two, and (2) model the dissimilar atoms while constraining the bond lengths, angles and dihedrals to thoεe of the candidate reεidue. Once all of the NSCRs making up the model are in turn selected from the database, fixed in space, and modeled to transform them into the coordinates of the corresponding acceptor NSCRε, the splice regions where the SCRs join the NSCRs are preferably refined to relieve any strain in the model that results from joining the SCRε and NSCRs. Thiε refinement can be accompliεhed uεing any εuitable computer algorithm, for instance the "Repair" algorithm in Insight II, to asεign the proper bond lengthε, bond angleε, and omega valueε to the reεidueε at the εplice junctions.

Now, the model as a whole is relaxed using a suitable computer algorithm to relieve any strain occasioned by the above procedures. Preferably the "Relax" algorithm of Inεight II iε applied in a εerieε of sequential steps to the model as a whole. Preferably, the order of the steps iε to apply the algorithm: (1) to the side chains of the NSCRs to aεεign proper geometries and remove any unfavorable non- bonded contacts between side chain atoms and other atoms in the molecule, (2) to all atoms of the NSCRs to remove any remaining unfavorable contacts between the NSCR and other atoms in the molecule, (3) to the mutated side chainε of the SCRs to remove any unfavorable non-bonded contacts between mutated SCR εide chain atoms and other atoms in the molecule, and (4) to all of the side chain atoms of the SCRε to remove the remaining unfavorable sidechain contacts.

Finally, an energy minimization procedure is performed uεing techniques well known in the art, for inεtance, uεing the "Discover" subprogram of Insight II, to allow the model to assume an energetically favorable conformation. In the preferred embodiment, however, the energy minimization is performed in a serieε of εequential εtepε. The entire model iε firεt subjected to energy minimization with backbone atoms tethered to their starting coordinateε with a force constant of 100 kcal/A². Then an energy minimization is performed for the entire model without the backbone atoms being tethered. The result of carrying out these stepε iε a model of the variable domain of each of the acceptor chains.

In the method of this invention, the model of the acceptor Fv is made by the following steps: (1) identify potential chain association residues by comparison of the sequence of the acceptor chain with the linear array of known structureε and εelect an appropriate known εtructure to uεe in modeling chain association of the acceptor molecule, (2) make a preliminary model by superimposing the backbone atoms of the potential chain association residues of the selected known structure, (3) subject the entire molecule to energy minimization, first, with the backbone atoms being tethered to their initial coordinates and, second, without the backbone atoms being tethered, (4) identify the chain association residues in the final acceptor Fv model, excluding all residues that are part of a CDR.

In the first step, in one chain of the structureε of each of the databaεe antibodies, each residue in the variable region of that chain having an atom within 4.5 Angstromε of an atom in a residue in the other chain is identified. If the residues so identified in each database antibody are not part of a CDR and are likely to have a significant interaction with residues in the other chain, they are earmarked in the linear sequence of the antibody as chain association residues. The procesε is repeated for the other chain of each database antibody.

Since the residues involved in chain association are generally conserved among antibodies, it can be assumed that there will be homology between chain asεociation residues in the V_L and VH of the database antibodies and those in the VL and VH of the acceptor antibody or immunoglobulin. Therefore, when a reεidue in the acceptor sequence is found to be identical to one earmarked in the array, it is earmarked as a chain association residue in the acceptor model. On the other hand, when an amino acid iε found in the acceptor that differε from the correεponding one in the database antibody in any of the positionε in the database sequences earmarked as chain association residues, it is designated as potentially disruptive to chain asεociation. Each database antibody is compared with the acceptor molecule. The database antibody with the greatest exceεε of favorable reεidueε over diεruptive reεidueε is choεen.

In the second step, superimpoεition iε accompliεhed uεing a program such as the "Superimpose" command in Insight II.

In the third step, a program such as "Discover" in Insight II is used to carry out the energy minimization, with the back bone atoms being tethered to their initial coordinateε with a force conεtant (uεually 100 kcal/A²) for the initial minimization and with no tethering for the final minimization.

In the fourth εtep, chain aεεociation reεidueε in the light chain are identified aε all reεidues from the light chain that contain an atom within a specific distance of any atom of any residue in the heavy chain selected as indicating possibility of significant interaction there between (usually about 4.5A). Similarly, chain association residues in the heavy chain are identified as all reεidueε from the heavy chain that contain an atom that is within a specific distance of any atom of any residue in the light chain selected as indicating possibility of significant interaction there between (usually about 4.5A) .

SteP T ree - _£k≤ Three-dimensional Modeling of Donor Fv and Identification of CJELR

Associated Residues

Models of donor Fv are arrived at in a manner identical to that described above for the acceptor Fv. CDR-asεociated residues are identified after minimization by determining those reεidueε containing an atom within a specific distance of any atom of any residue found within a CDR selected aε indicating the possibility of interaction there between (usually about 4.5A) . Theεe residues are defined as CDR-associated residues and are treated in a step in the humanization procesε deεcribed in Step 4 below. Step E£u Three-dimensional Modeling of

CPR-qraftefl £_a.

The CDR-associated residueε determined above are now identified in the primary amino acid sequence of the donor molecule, and the primary sequence for the altered light and heavy chain CDR grafted moleculeε are pieced together in segments.

First, the primary amino acid sequenceε of the donor and acceptor moleculeε are aligned with reference to the εequenceε of the known databaεe εtructureε. Second, on the donor linear array: (1) the CDR-associated residueε determined above are identified, (2) for SCRs or NSCRs that do not contain a CDR residue or a CDR-associated residue, the sequence of the entire segment is replaced with the sequence from the corresponding segment of the acceptor molecule, (3) for SCRs that contain one or more CDR residueε or CDR- associated residues, all residues that are neither CDR nor CDR-asεociated in the segment are replaced with those of the acceptor molecule, but the CDR residues and CDR-associated residueε are conserved as the donor residues, (4) in NSCRs that contain one or more CDR residues or CDR-asεociated reεidues, if the total number of residueε in the NSCR differε between the donor and acceptor, the entire NSCR is conserved as the donor sequence, (5) in NSCRs that contain one or more CDR residues or CDR-associated residues, if the total number of residueε in the NSCR is the same between the donor and acceptor, those residues that are neither CDR nor CDR- associated are replaced with those of the acceptor molecule, while the CDR residues and CDR-asεociated reεidueε are conεerved aε the donor residues. Thus, in all cases, CDR residues and CDR-associated residues in SCRs or NSCRs are conserved as the donor residues.

Third, the donor and acceptor models are superimposed. Once the two models are brought up on the computer screen, SCRs are determined. In this step SCRs are derived in a way distinct from that used in construction of the acceptor and donor modelε. In the latter case the SCRs were asεigned to the donor and acceptor based on the consensus SCRs determined from the known structures. In thiε εtep, SCRε are determined anew from the two modelε alone in a manner analogouε to that uεed to determine the SCRε between each of the known εtructureε, aε deεcribed in Step 1 above (wherein the acceptor waε deεignated to be held conεtant and the donor was superimpoεed upon it) .

Uεing the modeled three dimenεional εtructureε and εequenceε for the acceptor and donor Fvε, the operator uεeε the computer program to align the εequenceε for the Fvs and to superimpose the corresponding structures so that SCRs can be identified. For instance, the sequenceε are aligned in a linear array with each sequence constituting one row of the array, i.e. seqA (for acceptor) and εeqD (for donor). To facilitate alignment uεing the Insight II software, certain landmark amino acids known to be conserved among antibodies, such as the cyεteineε that form the intrachain disulfide bridge (i.e., the light chain cysteines at L23), are identified in each sequence and are aligned in vertical columns, as described in Step 1 above.

Three dimensional alignment of the two structureε iε further refined by identifying SCRs and superimposing them. Using the already superimposed structureε, the putative SCR1AD iε diεcovered by visual inεpection.

Preferably, εucceεsive SCRε are identified by working from amino to carboxy terminus of the molecules. The RMS deviation of the backbone atoms in the corresponding εegmentε of amino acidε in the two structures is calculated. The exact location of SCR1AD, and hence of the amino acids contained within the segments corresponding to SCR1AD, are adjusted by a procedure of trial and error whereby the amino acids in the linear εequenceε of the array that correεpond to thoεe in the putative SCR1AD are boxed and the RMS deviation iε calculated. The width of the box is maximized and the location of the box is adjusted until the RMS deviation reaches an acceptable maximum, for instance no more than about 0.75A.

To ensure that spatial alignment of SCR1AD at the amino terminus of the two structures is not destroyed by establishment of subsequent SCRs along the sequenceε (i.e., SCR2AD, SCR3AD, etc.), after the procesε haε been carried out to define SCR2AD, the two εtructures are superimpoεed again uεing the reεidueε for the back bone atoms in SCR1AD as well as SCR2AD. This proceεs is repeated for each subsequent SCR. Gaps, for example empty space holders, can be inserted within NSCRs as needed to accomplish vertical alignment of the SCRs For example, where any sequence has fewer amino acids between the SCRs than does the other, gapε can be uεed to make the two of equal length. Each εegment in the altered CDR grafted chain iε assigned spatial coordinates that correspond to those of the donor or acceptor residue to which it corresponds. Preferably this is done working from the amino to the carboxy terminus of the chain. Now the light and heavy chain minimized models constructed above are displayed on the computer screen together as an Fv. An energy minimization iε performed to allow thiε Fv model to assume an energetically favorable conformation using the steps described above. As a final check, the model is examined to determine whether any new CDR-associated residueε appear in the altered, CDR-grafted model using the techniques described above. If any new CDR-associated residue is seen in the altered CDR-grafted (and humanized) model, the amino acid at that position is replaced by the one found in the donor molecule. After the CDR-associated residues are modified as necesεary, the model iε analyzed to determine whether all the chain association sites identified in the acceptor model have been conserved in the altered CDR-grafted model. If differences are observed, they should be noted as possible future sites for mutagenesis if a significant decrease in secretion of the altered CDR-grafted protein iε obεerved aε compared to that of the acceptor molecule.

Step Five - Ji≤ Three-dimensional Modeling

The acceptor light and donor heavy chain primary amino acid εequenceε had already been aligned with reference to different εequences. Therefore, it was necesεary to bridge theεe alignmentε through realignment uεing a common sequence. In addition, the acceptor heavy chain provided information on chain asεociation residues. Donor heavy chain sequence was added to a linear array containing light chain donor and light and heavy chain acceptor sequenceε and aligned. Once aligned in thiε manner, SCRε were defined there between aε deεcribed in Step One, the Kabat defined CDRε and CDR-aεεociated reεidueε determined in Step Three were identified on the donor heavy chain linear array. For SCR or NSCR regionε which do not contain a CDR or CDR- associated residue, the entire region was replaced with the acceptor light chain sequence (and εtructure, i.e., coordinateε) . For SCRε which contain one or more CDR or CDR- aεεociated regions, the non-CDR-associated residueε were replaced with acceptor sequence (and structure, i.e., coordinates) , but donor heavy chain εequence (and εtructure, i.e., coordinateε) was conserved for the CDR-asεociated reεidueε. For NSCRε that contain one or more CDR or CDR- aεεociated reεidueε, the donor heavy chain εequence (and εtructure, i.e., coordinateε) waε conεerved for the entire region. In thiε way the primary sequence for the heavy chain CDR-grafted molecule was determined, and a compoεite εtructure waε developed.

Now, the reεultant model waε modified to aεεure that chain aεεociation reεidueε, derived from the acceptor model were conεerved. In all non-CDR or non-CDR-aεεociated regions, when the amino acid in the position occupied by the chain association residue was different than the correεponding acceptor reεidue, it was replaced with the corresponding acceptor heavy chain chain-association residue. In thiε example no chain-aεεociation reεidues were found to lie in the CDR or CDR-asεociated regionε. In the unlikely event that thiε should occur, the residue should be noted, but no change should be made.

Alternatively, humanized light chain can be used as acceptor and humanized heavy chain can be used as donor. In this caεe, chain association residueε used for the preliminary Fv model are those identified for humanized FV. Now that coordinates had been assigned for both light and heavy/light hybrid chains, these were displayed on the screen together. An energy minimization was performed uεing the "Discover" subprogram to allow the model to asεume an energetically favorable configuration. Firεt the entire model waε εubjected to energy minimization with backbone atomε tethered to their εtarting coordinateε with a force conεtant of 100 Kcal/A². Then the energy minimization algorithm was applied to the entire model without the backbone atoms being tethered.

CDR-Associated residues were determined for the modeled humanized light chain dimer as for the original donor Fv (Step Three) . Again, this was done by first identifying all residueε on the light or heavy/light hybrid chain that are within 4.5 A of any light chain CDR residue, and that also have a significant likelihood of interaction, based on orientation of the residue, charge, hydrophobicity, etc. Next, all residues on the light or heavy/light hybrid chain that were within 4.5 A of any heavy/light hybrid chain CDR residue were identified. Again, the set was limited to those with a high likelihood of significant interaction with the CDR residue of interest. In this way, the entire set of light and heavy chain CDR-associated residues was determined. The set of CDR-asεociated residues determined for the humanized light chain dimer was compared to that determined for the donor Fv. In any case where an additional CDR-associated residue is seen for the humanized, the amino acid at that position was replaced by the amino acid found in the murine donor.

After the CDR-associated residueε were modified aε necessary, the model was analyzed to determine if the chain asεociation reεidues identified for acceptor were conserved. In this example, they were conserved. If, however, differences are observed, these are noted, but no changes are made at thiε time. If, in addition, there iε a εignificant decreaεe in expreεsion observed for the humanized molecule, these are potential εiteε for modification.

As can be εeen from the results presented in the Examples below, these modeling methodε yield high affinity CSV_L recombinant antibodieε from an initial deεign without a requirement for iteration. Thiε method of modeling CDR εwitched antibodieε uεing εtructurally conεerved regionε can readily be modified by one skilled in the art to produce the CSV_L recombinant antibodies of this invention, εuch aε heavybodieε or CSV_L fragmentε. The acceptor amino acids identified as candidates for switching to donor amino acids by molecular modeling can be switched by oligonucleotide directed or site-directed mutageneεis of the DNA sequences encoding the CDR grafted heavy and light variable regions, for instance, as taught by T. Kunkel, Proc. Natl. Acad. Sri . USA. 82:488-492 (1985) or by codon-based mutagenesis whereby an amino acid alteration iε obtained for each in vitro substitution of a three nucleotide codon (Huse, et al . , Science. 246:1275 (1989)) . Preferably however, the DNA of the entire variable region of the heavy and light chainε iε prepared by oligonucleotide εyntheεiε aε deεcribed hereafter.

Once the DNA encoding a CSV_L recombinant antibody haε been prepared, it iε then incorporated into a vector and operably linked to nucleic acid εequenceε encoding tranεcriptional and translational regulatory sequenceε. Any εuitable expreεεion vector may be uεed in thiε invention and exemplary vectorε are provided in the Exampleε below. Thoεe with skill in the art will appreciate that the choice of vector is limited to those vectors capable of directing expreεεion of the nucleic acid sequence encoding protein and to those vectors that can incorporate and support the function of the regulatory regions used. Further, the choice of vector is limited by the cell type selected. Not all vectors and not all regulatory elements necessary for recombinant protein expression function in all cell types. As a general rule eukaryotic expression vectors are suitable for protein expresεion in eukaryotes and prokaryotic expresεion vectorε are εuitable for prokaryotes. Both types of vectors are commercially available and those with skill in the art of molecular biology will be able to εelect appropriate vectorε εuitable for recombinant protein expreεεion within a given cell type.

Methods for incorporating a particular region of nucleic acid into a nucleic acid vector are well known in the art of molecular biology (See Sambrook, et al . , Molecular Cloninσ A Laboratory Manual. Second Edition, Cold Spring Harbor Laboratory Press, 1989). For example, short regions of nucleic acid (lesε than 400 bp) can be prepared by generating sense and antisense oligonucleotides complementary to the desired gene sequence that overlap. These oligonucleotides hybridize to one another, and can be amplified in a PCR reaction, ligated and incorporated into an appropriate expression vector (see generally H.A. Erlich, PCR Technology: Principles and Applications for DNA Amplification. W.H. Freeman and Co . , New York, 1992).

In general, the recombinant antibodies of this invention can be prepared by recombinant methodε known in the art (εee generally, Sambrook, et al . , supra) from the amino acid and DNA sequences of the donor and acceptor antibodieε. For instance, if a monoclonal antibody is used as the donor antibody, hybridoma or polydoma technology using conventional procedures for immunization of mammals with an immunogenic antigen preparation, fusion of immune lymph or spleen cells 96/06625 PCI7US95/10791

- 46 -

with an immortal myeloma cell line, and isolation of specific hybridoma clones can be employed to obtain the monoclonal antibody.

Alternatively, the geneε encoding the donor and acceptor antibodieε can be obtained by methodε known in the art, for inεtance by chemical εyntheεiε, aε deεcribed above, if the εequenceε of the geneε are known. If the εequenceε are not known, or if the geneε have not previouεly been iεolated, they may be cloned from a cDNA library (made from RNA obtained from a εuitable tiεsue or batch of cellε in which the deεired gene is expressed, such as a hybridoma or polydoma) or from a suitable genomic DNA library. The mRNA iε extracted and cDNA for the coding regionε is derived using the enzyme reverse transcriptase and methodε well known in the art. The gene iε then identified uεing an appropriate molecular probe. For cDNA libraries, suitable probes include monoclonal or polyclonal antibodies (provided that the cDNA library iε an expreεεion library) , oligonucleotideε, and cDNAS or fragmentε thereof. The probeε that may be uεed to iεolate the gene of intereεt from genomic DNA librarieε include cDNAS or fragmentε thereof that encode the εame or a εimilar gene, homologous genomic DNAs or DNA fragmentε, and oligonucleotideε. Screening the cDNA or genomic library with the εelected probe is conducted using standard procedures as deεcribed in chapters 10-12 of Sambrook, et al . , supra. From the sequence of the cDNA or that of the genomic DNA, the corresponding amino acid sequenceε to be used in molecular modeling are deduced, usually by a computer software program, such as iε commercially available from DNAStar (Madiεon, WI) . Once the amino acid εequenceε of the donor and acceptor antibodieε are known, their CDRs are identified using the procedure of Kabat and Wu, supra. For modeling and conεtruction of a CSV_L domain, the amino acidε corresponding to at least one and preferably all three CDRs of the acceptor VL are replaced with CDRs of the donor VH.

Additional donor residueε identified by molecular modeling aε useful for retaining binding affinity and/or chain association are determined as described above.

When a nucleotide sequence capable of encoding the CDR grafted CSVL domain has been determined from the protein sequence, it is fabricated and ligated into a suitable replicable expression vector, optionally along with the desired constant region genes from the acceptor antibody. A similar procedure is then followed to construct the vector containing the geneε encoding the aεεociated CDR grafted light chain or heavy chain if applicable using methods well known in the art.

It is preferred that the DNA encoding the entire CDR-grafted variable regions, including the CSV_L domain(ε) be inserted into an appropriate sequencing vector (e.g. a TA vector) and sequenced employing, for instance, the

Sequenasell kit (United States Biochemical, Cleveland, OH) used with a Genesis® 2000 automated DNA sequencer (Dupont, Wilmington, DE) according to the manufacturer's instructions. The spliced and sequenced exon is then excised from the sequencing vector and ligated into a vector that may optionally contain one or more exons encoding constant regions for the CDR-grafted chain. If it is deεired to produce an recombinant antibody having a light and a heavy chain, the DNA encoding the light chain can be εpliced into one vector and the DNA encoding the heavy chain can be εpliced into another vector. Alternatively, the DNA encoding both chainε can be spliced into the same vector.

To obtain the recombinant antibodies of the invention, the DNA encoding one or more immunoglobulin chains prepared as described above is ligated into a replicable expression vector so as.to be operably linked to transcription regulatory element(ε); suitable host cells are transfected with the vectorε; and the tranεformed hoεt cellε are cultured under conditionε favorable for forming the deεired recombinant antibodieε.

Various types of vectors may be used such as plasmids and viruses, including animal viruses and bacteriophages. In the embodiment, a vector is employed which iε capable of integrating the deεired gene εequenceε into the hoεt cell chromoεome. The cellε which have stably integrated the introduced DNA into their chromosomes can be selected by also introducing one or more marker genes which allow for selection (i.e., growth of the cellε in the presence of a toxic drug) of hoεt cellε which contain the expreεsion vector. The introduced marker gene sequence will be incorporated into the plasmid or viral vector containing the gene(ε) encoding the construct containing a CSV_L domain.

Factors of importance in selecting a plasmid or viral expresεion vector include the eaεe with which recipient cells that contain the vector may be recognized and selected; the number of copies of the vector which can be introduced or desired in a particular host; and whether it iε desirable to "shuttle" the vector between host cells of different specieε.

Eukaryotic expression vectors for yeast or mammalian cells, as well as prokaryotic expression vectors, may be used to express the recombinant antibodies of this invention.

Although, either eukaryotes or prokaryotes can be used aε hoεt cells for this invention, the modeling methods uεed are exceptionally appropriate for eukaryotic cellε, and more specifically for mammalian B lymphocytes. Alternatively, however expression can be obtained in a multitude of species, using suitable vectorε and hoεtε. Suitable prokaryotic hoεt cells include E. coli strain JM 101, E. coli K12 strain 294 (ATCC No. 31,336), E. coli strain W3110 (ATCC No. 27,325), E. coli X1776 (ATCC No. 31,537), E. coli XL-1-Blue (Stratagene) , and E. coli B; however, many other strains of E. coli , such as HB101, NM522, NM538, MN539, and many other species and genera of prokaryotes may be used as well. In addition to the E. coli strains liεted above, bacilli εuch aε Bacilluε εubtiliε. other enterobacteriaceae such as Salmonella tvphimurium or Serratia marcesans. and various Pseudomonas species may all be used as hosts. As is well known to one skilled in the art, it is necessary to remove any introns from eukaryotic genes which are to be expresεed in prokaryotic hosts.

When the vector is designated for expression in baculovirus, suitable promoters and enhancer εequenceε include, but are not limited to AcMNPV polyhedrin, AcMNPV ETL and AcMNPV plO sequences. One particularly suitable polyadenylation signal is the polyhedrin AcMNPV. Ig Kappa, Ig Heavy and AcMNPV are examples of suitable signal sequenceε. These vectors are useful in the following insect cell lines, among otherε: SF9, SF21 and High 5.

Alternatively, the polypeptideε can be expressed in yeaεt εtrainε such as PS23-6A, W301-18A, LL20, D234-3, INVSC1, INVSC2, YJJ337. Promoter and enhancer sequenceε εuch as gal 1 and PEFT-1 are useful. Vra-4 also provides a εuitable enhancer sequence. Sequences useful as functional "origins or replication" include arsl and 2μ circular plasmid.

Following procedures outlined above, mammalian cell lines such as myeloma (P3-653) or hybridoma (SP2/0), Chinese Hamster Ovary (CHO), Green monkey kidney (COSl) and murine fibroblasts (L492) are suitable host cells for expresεion. Theεe "mammalian" vectorε can include a promoter, an enhancer, a polyadenylation signal, εignal εequenceε and genes encoding selectable markers including, but not limited to, geneticin (neomycin resistance) , mycophenolic acid

(xanthine guanine phosphoribosyl transferase) or histidinol (histidinol dehydrogenase) .

Suitable promoterε for use in mammalian host cells include, but are not limited to, Ig Kappa, Ig heavy, Cytomegalovirus (CMV) immediate early, Rous Sarcoma Virus (RSV) , Simian virus 40 (SV40) early, mouse mammary tumor (MMTV) virus and metallothionein. Suitable enhancers include, but are not limited to Ig Kappa, Ig Heavy, CMV early and SV40. Suitable polyadenylation sequences include Ig Kappa, Ig Gamma or SV40 large T antigen. Suitable signal sequences include, but are not limited to, Ig Kappa, Ig Heavy and human growth hormone (HGH) . For expresεion in mammalian cellε the vectorε containing the DNA encoding the heavy and light chain geneε of the antibody conεtruct can be placed into separate bacterial amplification vectors, such as E. coli DH 10 B Electromax (BRL, Gaitherεburg, MD. ) , cultured, and screened for antibiotic resistance to amplify the plasmid. Generally, the DNA of the εelected cloneε iε verified by reεtriction digeεtion and DNA εequencing. Double εtranded dideoxy εequencing iε performed, for example on a DuPont Geneεiε® 2000 inεtrument, uεing the DuPont Geneεiε® 2000 εequencing kit according to the manufacturer' ε inεtructions. Post gel processing can be done with the Base Caller 5.0 program (DuPont, Boεton, MA) . One εkilled in the art can readily provide alternative methodε of performing theεe εteps in the cloning process.

Particularly useful vectors for expresεion of the CSV_L recombinant antibodies of this invention in mammalian cells are pGIM9kappa and pNIM9k/hCEM-gamma deposited with the ATCC under the requirementε of the Budapeεt Treaty under Accession Nos. 75512 and 75511, respectively. These vectorε comprise human immunoglobulin regulatory elements and contain casεette εiteε for insertion of DNA encoding CDR grafted light and heavy chain sequences. These vectors, which are especially designed for expressing CDR grafted antibodies and fragments wherein the acceptor antibody is human, are preferably transfected into host cells of the B-cell lineage for production of optimal levels of immunoglobulin. Uεe of theεe vectorε iε exemplified in the exampleε below. The principal advantage of expressing the CSVL domain in the above described vectorε in host cells of the B-cell lineage, iε that thiε allowε for maximal conservation of aεεembly and εecretory components to assure reproducible high level expresεion and secretion of the molecules of interest.

After εelection of the tranεformed cellε, theεe cells are grown in culture media and screened for expresεion of the appropriate antibody conεtruct uεing techniques well known in the art for enzyme or radio assay, or by the methods exemplified in Example 15 below. Expression of the sequence results in the production of the fusion protein of the present invention.

A chelator may also be bound to the CSVL recombinant antibody through a short or long chain linker moiety, through one or more functional groups on the antibody, e.g., amine, carboxyl, phenyl, thiol or hydroxyl groups. See for example Schlo , "Monoclonal Antibody-based Therapy of a Human Tumor Xenograft with a l^Lutetium-labeled Immunocon ugate, " Cancer Research. 31:2889-2896 (1991); U.S. Patent 4,994,560 to Kraper, et al . ; and Sigel, et al . , "Coordinating Propertieε of the Amide Bond. Stability and Structure of Metal Ion Complexeε of Peptides and Related Ligandε," Chemical Review. 82:385-426 (1982). Various conventional linkers can be used, e.g., diisocyanates, diisothiocyanates, carbodiimideε, biε-hydroxyxuccinimide esters, maleimide-hydroxysuccinimide esterε, glutaraldehyde and the like, for instance, a selective sequential linker such as the anhydride-isothiocyanate linker disclosed in U.S. Patent 4,680,338.

This invention also contemplates fusing at least one of the genes encoding the CSVL recombinant antibodies to a second gene encoding a chelating peptide for binding a radiometal ion, a toxin, or an enzyme such that a fusion protein is generated during transcription and translation.

Fusion of two genes may be accomplished by inserting the gene encoding the chelating peptide into a particular site on a plasmid that contains an antibody gene, preferably a constant region gene, or by inserting an antibody gene into a particular site on a plasmid that contains a gene encoding the chelating peptide.

The plasmid is cut at the precise location that the gene is to be inserted using a restriction endonuclease site (preferably a unique site) . The plasmid is digested, phosphataεed, and purified aε deεcribed above. The gene encoding the second protein or protein segment is then inserted into this linearized plasmid by ligating the two DNA'ε together such that the reading frames of the gene already in the plasmid and of the gene to be inserted are preserved. If the two pieces of DNA to be ligated have blunt ends or sticky ends, ligation can be direct using a ligase such as bacteriophage T4 DNA ligase and incubating the mixture at 16"C for 1-4 hours or overnight in the presence of ATP and ligase buffer as described in Section 1.68 of Sambrook, et al . , supra. If the ends are not compatible, they must first be made blunt by using the Klenow fragment of DNA polymerase I or bacteriophage T4 DNA polymerase, both of which require the four deoxyribonucleotide triphosphates to fill in overhanging single-εtranded endε of the digeεted DNA.

When conεtructing a replicable expreεεion vector containing the DNA, encoding one or more of the chainε of the inεtant CSV_L recombinant antibodies, all subunitε can be regulated by the same promoter, typically located 5' to the DNA encoding the subunits, or each can be regulated by a separate promoter suitably oriented in the vector so that each promoter is operably linked to the DNA it iε intended to regulate. When the CSV_L DNA is composed of subunits, for example, the DNA for the heavy and light chains of an intact kappabody, generally one of the subunits is fused or operably linked to the gene for the chelating peptide, if one iε included. Thiε fused gene will contain a functional εignal εequence. A separate gene encodes the other subunit or subunits, and each subunit generally has its own εignal εequence. Alternatively, to increaεe the εpecific activity of the gene fuεion product, more than one gene for the chelating peptide can be fused to a subunit. For example, the gene for the chelating peptide can be fused to the genes encoding both the heavy and light chains of any antibody or antibody fragment, such as an intact kappabody or a heavybody or Fab-like fragment. A εingle promoter can regulate the expression of both subunits, or each subunit can be independently regulated by a different promoter. Thus, generally the complementary chain needed to provide the binding domain of the protein ligand may be provided by expreεεing the complementary chain as a single polypeptide in the host cell or such a εingle polypeptide can be added separately. For example, to produce a fusion protein composed of a chelating peptide and an kappabody fragment, a gene encoding a light chain (or portion thereof) is functionally linked to the chelating peptide gene and this hybrid gene iε expreεεed in a host cell. To allow formation of the binding domain or double chain fragment (e.g., kappabody fragment or ScFv (CSVL) , the same host cell can be engineered to expresε the other chain and excrete the aεsembled fragment having the chelating peptide attached to the corresponding light chain. In another embodiment, the chelating peptide can be attached to the light chain and expressed alone aε a fusion protein, (such as with a CSV or heavybody fragment) or both chains can be attached to chelating peptides as fusion proteins and the dimer construct can be expressed from a single host cell.

The molecules of this invention can be used in all in vitro diagnostic, in vivo diagnostic, and therapeutic applications for which antibodieε have been used or their use proposed. These include naked antibody therapy (both those requiring effector function and those only requiring binding function) , radioimmunotherapy, in vivo radioimmunodiagnostics, in vitro radioimmunometric assays, ELISA assays, quantitative ELISA assays, and immunohistochemical applications.

The scintigraphic imaging method of the invention is practiced by injecting a warm-blooded animal preferably a mammal, and more preferably a human, parenterally with an effective amount for scintigraphic imaging of the radiolabeled monospecific or multispecific antibody agent conjugate. By parenterally is meant, e.g. intravenously, intraarterially, intrathecally, interstitially or intracavitarily. For imaging cardiovascular lesions, intraveneous or intraarterial administration is preferred. Labeling with either Iodine-131 or Iodine-123 is readily effected using an oxidative procedure wherein a mixture of radioactive potassium or sodium iodide and the antibody is treated with chloramine-T, e.g., as reported by Greenwood, et al . , Bioche . J.. 89:114 (1963) and modified by McConahey, et al . , Int. Arch. Allerαv APPI. Immunol.. 29:185 (1969) . This resultε in direct εubεtitution of iodine atoms for hydrogen atomε on the antibody molecule. Alternatively, lactoperoxidaεe iodination may be used, as described by Feteanu, "Labeled Antibodies in Bioloσv and Medicine. " page 302 (McGraw-Hill Int. Bk. Co., New York, 1978) , and references cited therein.

Feteanu also discloεes a wide range of more advanced labeling techniques, supra, pages 214-309. Introduction of variouε metal radio-iεotopeε may be accompliεhed according to the procedureε of Wagner, et al . , J. Mucrl . Med.. 20:428 (1979); Sundberg, et al . , J. Med.

Chem.. 17:1304 (1974); and Saha et al . , J. Nucl. Med.. 6:542 (1976) , for inεtance.

As used in the methods of the present invention, the compounds taught herein can be administered to the subject animal such as a laboratory animal, a mammal or more preferably a human, by any meanε known to thoεe εkilled in the art, including parenteral injection or topical application. Injection can be done intravaεcularly, intraperitoneally, subcutaneously or intramuscularly. For parenteral administration, the compounds can be administered in admixture with a suitable pharmaceutically acceptable carrier. As used herein the term "pharmaceutically acceptable carrier" encompasses any of the standard pharmaceutical carriers, such as a phosphate buffered saline εolution, water, and emulsions, such as an oil/water or water/oil emulsion, and various types of wetting agents.

This invention also provides pharmaceutical compoεitionε containing any of the CSV_L recombinant antibodies fused to the metal chelating peptides described herein linked to protein ligands, with or without the radioion having been incorporated into the chelating peptide. 96/06625 PC17US95/10791

- 55 -

Therapeutic formulations of the compositions of this invention are prepared for storage by mixing the metal chelate-protein complex with optional physiologically acceptable buffers and carriers, excipients, or εtabilizerε, (Remington's Pharmaceutical Sciences. 16th edition, Osol, A., Ed.. (1980)), in the form of lyophilized cake or aqueouε εolutionε. Acceptable carrierε, excipients or stabilizerε are nontoxic to recipients at the dosageε and concentrationε employed, and include bufferε such as phosphate, citrate and other organic acidε; antioxidents including ascorbic acid; low molecular weight (less than about 10 residueε) polypeptideε, proteinε, εuch aε εeru albumin, gelatin, or immunoglobulinε; hydrophilic polymers such as polyvinylpyrrolidone; and the like. Theεe pharmaceutical compositions are used for in vivo diagnostic or therapeutic purposeε.

The recombinant antibodieε of thiε invention are present in the pharmaceutical composition in an effective amount. Methods of determining effective amountε are known to thoεe of skill in the art and depend upon a variety of factors, including the type of disorder, age, weight, sex and medical condition of the animal or human patient, the severity of the condition, the route of administration, and the type of diagnostic or therapeutic treatment deεired. A skilled veterinarian or physician can readily determine and prescribe the effective amount of the compound or pharmaceutical composition required to diagnose or treat the animal or patient, respectively. Therefore, the dose of the diagnostic compound would be selected to accommodate this requirement. For diagnostic applications a typical radiodose is between 20 and 30mCi. For instance if the CSVL recombinant antibody is an Fab' kappabody fragment the dosage is generally in the range between about 1 and 3.O Ci per nmol of fragment. As one skilled in the art will appreciate, the amount and type of CSV_L recombinant antibodies used will affect the pharmacokinetics of the compound and one skilled in the art would take these considerations into account in selecting the proper compound and dosage in uεe. Conventionally, for therapeutic application, one εkilled in the art would employ relatively low doεes initially and subεequently increase the dose until a maximum safe response iε obtained. The εpecific activity of the compound will determine the amount of the compound adminiεtered and hence, the doεage of the compound containing the radioion adminiεtered.

For human therapeutic regimenε the typical doεage of the radioion per injection iε in the range from about 10 to 30mCi per injection and the typical corresponding antibody dose is in the range from about 2 to lOmg. Although in certain instanceε a εingle therapeutic doεe can be effective, more typically the patient to be treated will be adminiεtered a series of gradually increasing doseε at intervalε εpaced appropriately to accommodate the needε of the patient. For inεtance, when CSV_L recombinant antibody is a kappabody fragment, is tumor-specific, and is fuεed to a chelating peptide incorporating Yttrium-90 aε the therapeutic radioion, a typical doεage regimen would conεiεt of repeated adminiεtration of the therapeutic compound over appropriately spaced intervals, for instance of two weeks duration, beginning with a doεage of 10mCi/2mg of antibody and increaεing to a doεage of about 30mCi/10mg of antibody. If the CΞV_L recombinant antibody is incorporated into a compound containing a separate chelating peptide, the weight of the chelating peptide is negligible in comparison to the weight of the antibody so that its weight can be ignored in calculating the proper ratio of radionuclide to delivery agent (i.e., chelating peptide plus antibody) .

Alternatively, paramagnetic compoundε useful for MRI image enhancement can be conjugated to a substrate bearing paramagnetic ion chelators or expoεed chelating functional groups, e.g., SH, NH2, COOH, for the ions, or linkers for the radical addends. The foregoing are merely illustrative of the many methods of radiolabeling proteins known to the art. The MRI enhancing agent must be preεent in O 96/06625 PCI7US95/10791

- 57 -

εufficient amountε to enable detection by an external camera, using magnetic field strengths which are reasonably attainable and compatible with patient safety and instrumental design. The requirements for such agents are well known in the art for those agents which have their effect upon water molecules in the medium, and are disclosed inter alia, in, e.g., Pykett, Scientific American, 246:78 (1982); and Runge, et al . , Am. J. Radiol.. 141:1209 (1987). The following examples illuεtrate the manner in which the invention can be practiced. It iε underεtood, however, that the exampleε are for the purpoεe of illustration and the invention is not to be regarded as limited to any of the εpecific materials or conditions therein.

Example 1

PCR Cloning of ZCE 025 Variable Regions

The initial cDNA cloning of the ZCE 025 Variable regions using the method of Okayama, H. and Berg, P. (Mol. and Cell. Biol., 2:161-170 (1982); Mol. and Cell. Biol., 3:280-289 (1983)) gave 3* sequences for both the heavy and light chains. In order to obtain the 5' sequenceε, the variable regions were isolated using a method termed "anchor PCR" (Loh, E.Y. et al.,Science, 243, 217-220 (1989)). Anchor PCR allows the use of a specific heavy or light chain primer (in our case, a sequence in the CK or CHI regions) and a second poly-C-containing primer that recognizeε a poly-G εequence added to all the mRNA-derived cDNAε, as is shown in Table 1 below. Another advantage of this technique is that the upstream primer recognizes an added synthetic segment of DNA, making it possible to obtain the native sequence of the entire signal region. Table 1

poly C primer C region primer poly A tail CCCCCCC xxxxxxx GGGGGGG xxxxxxx _ fτιrτifτifτϊmrτif^,

.region amplified.

a . Cloning of ZCE Kappa Light Chain cDNA

(1) ZCE 025 mRNA was obtained uεing the Guanidinium HCl procedure, as described in Sambrook, et al . ( supra , 7.18-

7.22) .

(2) The first and second strand cDNA syntheεeε were performed uεing the Stratagene (San Diego, CA) LambdaZap® cDNA cloning kit according to the manufacturer ' ε directions without the incorporation of a radioactive nucleotide. The reεulting cDNA waε ethanol precipitated.

(3) A poly G tail waε added to the 3' endε of the cDNA by reεuspending the precipitated cDNA in 23 μl water and adding 10 μL 5X tailing salts (0.9M Sodium Cacodylate, 150mM Tris-HCl (pH 6.8)), 5 μL ImM Dithiothreitol, 5 μL 10 mM dGTP, 5 μl lOmM Cobalt Chloride, 2 μl (40 Units) terminal deoxynucleotide transferase (Boehringer Mannheim, Indianapolis, IN) and incubating for 1 hour at 37' .

(4) The poly G tailed cDNA was digested with Xho I. This enzyme cleaves the cDNA at an Xho I site within the

Stratagene primer specific to the poly A region of the mRNA used for cDNA synthesiε and removeε the downεtream poly G tail on the εecond εtrand of the cDNA.

(5) The ZCE 025 Kappa V region was isolated from the cDNA using the Geneamp® PCR kit from Perkin Elmer Cetuε

(Norwalk, CT) according to the manufacturer' ε inεtructionε. The poly G-tailed, Xho I-cut cDNA waε used as template with the following poly C upstream primer:

5'GAC TAG CGG CCG CAT CGA TCC CCC CCC CCC CCC C (SEQ. I.D. No. 3) and a murine Kappa-specific downstream primer: 5 'CAG ACG TCG ACG ATG GAT ACA GTT GGT GCA GCA TC (SEQ. I.D. No. 4) The amplification conditions were 94' for 1 min, 45^" for 1 min, 72' for 3 min for 25 cycles.

(6) The amplified DNA was digested with Sal I and Not I and ligated into the pBluescript® cloning vector from

Stratagene which vector had been previously digested with Sal I and Not I. The ligated mixture was used to transform freshly prepared competent cells of the E. coli strain MC1061 (Clonetech, Palo Alto, CA) . The bacterial cells thuε transformed were identified by ampicillin resiεtance.

(7) Poεitive colonieε were confirmed by restriction enzyme analysiε and these had insertε of approximately 400 bp, the expected εize for the kappa V region.

(8) The poεitive cloneε were verified by εequence analyεiε on the Geneεiε® 2000 automated DNA εequencer from DuPont (Wilmington, DE) . The cDNA εeguence (SEQ. I.D. NO. 5) of the light chain variable region of ZCE 025 obtained and the corresponding amino acid sequence (Sequence I.D. No. 6)

SEQ. I.D. NO. 5

ZCE-025 Light Chain Variable cDNA GAC ATT GTG ATG ACC CAG TCT CAA AAA TTT ATG TCC ACA TCA GTT GGA GAC AGG GTC AAC ATC ACC TGC AAG GCC AGT CAG AAT GTT CGT ACT GCT GTA GCC TGG TAT CAA CAG AAA CCA GGG CAG TCT CCT AAA GCA CTG ATT TAC TTG GCA TCC AAC CGG TAC ACT GGA GTC CCT GAT CGC TTC ACA GGC

ATT GGA TCT GGG ACA GAT TTC ACG CTC ATC ATT AGC AAT GTG CAA TCT GAA GAC CTG GCA GAT TAT TTC TGT CTG CAA CAT TGG AAT TAT CCT CTC ACG TTC GGT GCT GGG ACC AAG CTG GAG CTG AAA C 381 SEQ. I.D. No. 6

Murine ZCE-025 Light Chain Variable Region Amino Acid Sequence:

DIVMTQSQKFMSTSVGDRVNITCKASQNVRTAVAWYQQKPGQSPKALIYLASNRYTGVPDR FTGIGSGTDFTLIISNVQSEDLADYFCLQHWNYPLTFGAGTKLELK b . Cloning of ZCE Gamma cDNA

(1) ZCE 025 mRNA was obtained using the Guanidinium HCl procedure, as described in Sambrook, et al . , ( supra , 7.18- 7.22) . (2) cDNA was prepared uεing the method deεcribed in Example l.a., above, for the ZCE kappa light chain.

(3) A poly G tail waε added to the 3' endε of the cDNA as deεcribed in Example l.a., above.

(4) The poly G tailed cDNA waε digeεted with Xho I and the ZCE 025 Gamma variable region waε iεolated from the cDNA using the Geneamp® PCR kit from Perkin Elmer Cetus according to the manufacturer' ε inεtructionε . The poly G-tailed, Xho-I cut cDNA waε used as template with the following poly C upstream primer: 5 'GACTAGCGGCCGCATCGATCCCCCCCCCCCCCCC (SEQ. I.D. NO. 3) and a murine Gamma 1 specific downstream primer: 5' CAG ACG TCG ACG TTC CAG GTC ACT GTC ACT GGC TC (SEQ. I.D. NO. 7) The amplification conditions were 94' for 1 min, 45" for 1 min, 72' for 3 min for 40 cycles. (6) The amplified DNA was digested with Sal I and Not I and ligated into the pBluescript® cloning vector (Stratagene, San Diego, CA) which had been previouεly digeεted with Sal I and Not I. The ligated mixture waε used to tranεform freεhly prepared competent cellε of the E. coli εtrain MC1061. The bacterial cellε thus transformed were identified by ampicillin resistance.

(7) Positive colonies were confirmed by restriction enzyme analyεiε and these had inserts of approximately 450 bp, the expected size for the Gamma chain variable region. (8) The positive clones were verified by sequence analysis on the Genesis® 2000 automated DNA sequencer from DuPont (Wilmington, DE) , according to the manufacturer' ε inεtructionε. The cDNA εequence (SEQ. I.D. NO. 8) of the ZCE heavy chain variable region obtained and the correεponding amino acid εequence (SEQ. I.D. NO. 9) SEQ . I . D . NO . 8

ZCE-025 Heavy Chain Variable cDNA Sequence:

GAT GTG CAG CTG GTG GAG TCT GGG GGA GGC TTA GTG CCG CCT GGA GGG TCC CGG AAA CTC TCC TGT GCA GCC TCT GGA TTC ACT TTC AGT AAC TTT GGA ATG CAC TGG ATT CGT CAG GCT CCA GAG AAG GGA CTG GAG TGG GTC GCA TAC ATT AGT GGT GGC AGT AGT ACC GTC CAC TAT GCA GAC TCC TTG AAG GGC CGA TTC ACC ATC TCC AGA GAC AAT CCC AAG AAC ACC CTG TTC CTA CAA ATG ACC AGT CTA AGG TCT GAA GAC ACG GCC ATG TAT TAC TGT GCA AGA GAT TAC TAC GTT AAT AAC TAC TGG TAC TTC GAT GTC TGG GGC GCA GGG ACC ACG GTC ACC GTC TCC TCA G 420

SEQ. I.D. NO. 9 Murine ZCE-025 Heavy Chain Variable Region Amino Acid Sequence:

DVQLVESGGGLVPPGGSRKLSCAASGFTFSNFGMHWIRQAPEKGLEWVAYISGGSSTVHYA DSLKGRFTISRDNPKNTLFLQMTSLRSEDTAMYYCARDYYVNNYWYFDVWGAGTTVTVSS

Example 2

Cloning and Sequencing IM9 light and heaw chain cDNAs

The human plasmacytoma cell line IM9 (ATCC #159) expreεεes an IgG(Yι,K) immunoglobulin.

a. Extraction of IM9 mRNA.

A total of 8X10⁷ IM9 cells were used for mRNA purification by the Fast-Trak™ kit from Invitrogen (San Diego, California) using an enzyme mix to digest the cells and oligo dT resin to adsorb the polyadenylated mRNA from the cell lysate according to manufacturer's directions. The resulting mRNA was redissolved in lOOμl of sterile water and split into lOμl aliquots. Each aliquot was stored at -20° in ammonium acetate and ethanol. b. Synthesis of cDNA.

The syntheεiε of a cDNA library was performed using a Librarian kit (Invitrogen) . The pooled mRNA from four of the tubes in a. was quantitated by measuring absorbance at 260nm. The firεt strand cDNA synthesiε waε performed according to manufacturer's directions using an oligo-dT primer and reverse transcriptase in the preεence of deoxyribonucleotideε and RNAaεe inhibitorε. Second εtrand εyntheεiε waε begun immediately by addition of ribonucleaεe H, E. coli ligaεe, and DNA polymeraεe in the preεence of the appropriate buffer. The reaction waε extracted once with phenol/chloroform and precipitated. The pellet waε reεuεpended in εterile water and ligated with BεtXI linkerε εupplied with the kit.

c. Purification of cDNAs.

The products of cDNA syntheεis and linker ligation were separated by size on an agarose gel in TAE (tris acetate EDTA) buffer (see Sambrook, et al . , supra) . The cDNA molecules over 700bp were cut out of the gel and separated from the agarose by electroelution into a small volume of TAE buffer (0.04M Tris-acetate, 0.001M EDTA). The cDNA waε extracted once with phenol/chloroform and precipitated. The εample waε centrifuged, and the pellet waε rinεed with ethanol, then air-dried.

d. Vector construction and transformation of bacterial cells.

The purified cDNA waε ligated to the vector provided in the kit, pCDNAII, which is already cut with an enzyme that leaves the appropriate εticky ends for the linker used on the cDNA and not for relegation to itself. The ligation mixture was electroporated into the E. coli strain DH10B (ElectroMAX) (BRL, Gaithersburg, Maryland) using the Cell-Porator (BRL) at 330uF, 2.5kV. The total number of colonies obtained in this library was 1.8X10^ clones. O 96/06625 PC17US95/10791

- 63 -

e . Preparation of Filter lifts of the IM9 cDNA library.

The library waε inoculated onto LB agar media (950ml deionized water, bacto-tryptone lOg, bacto-yeaεt extract 5g, NaCl lOg) with ampicillin at 7500cfu (colony forming unitε) per 15cm plate. A total of 12 plateε were made for a total of 9xl0⁴cfu of cDNA clones. The colonieε were blotted onto nylon filters by placing a dry filter on the colonies and removing the filter. The plates were returned to the incubator to allow the bacteria to grow back. The filters were placed on a layer of Whatman filter paper saturated with 5% SDS, 2 X SSC and put into the microwave oven on a high setting for 10 minutes. The filterε were air- dried and stored at 4°C.

f . Primary Screening of the IM9 cDNA library.

The filters were incubated at 45°C in prehybridization buffer (2XSSC, 1%SDS, 0.5% nonfat dry milk). These were then hybridized with human Ig mixed kappa and gamma conεtant region probeε using a method and probes described in C.B. Beidler, et al . , supra . The probeε were labeled uεing a Prime-It® kit (BRL) in 6 X SSC, 1% SDS, 0.5% nonfat dry milk, at 65°C overnight. The filterε were waεhed with 6 X SSC, 1% SDS, three timeε at 65°C, 5 minuteε each time, then with 1 X SSC , 0.1% SDS, three timeε at 65°C, 20 minuteε each time. The filters were put on Kodak XAR-5 X-ray film at room temperature overnight.

g. Secondary Screening of the IM9 cDNA library. Sixty-two poεitive colonieε were picked from the plateε and εtreaked onto LB agar media in duplicate, twelve to a plate, for two sets of six plates. These were blotted on nylon filters and hybridized using a method and probes described in C.B. Beidler, et al . , supra . One set was hybridized with a kappa constant region probe and one with a gamma constant region probe. h . Tertiary and Quaternary Screening of the IM9 cDNA library.

The streaks that were positive for the kappa probe were picked and plated out on LB media with ampicillin.

These plateε were blotted aε deεcribed before, the filters were hybridized, and the positiveε were picked. Theεe clones were subjected to one more round of blotting and hybridization to prove that the clone was pure. The sequence is provided below aε Sequence I.D. 10 and the amino acid εequence iε provided as Sequence I.D. 11.

SEQ. I.D. NO. 10

GAC ATC CAG ATG ACC CAG TTT CCT TCC ACC CTG TCT GCT TCT GTA GGA

GAC AGA GTC ACC 60

ATC ACT TGT CGG GCC AGT CAG AGT ATT AGT GCC TGG TTG GCC TGG TAT

CAG CAG AAA CCA 120

GGG AAA GCC CCT AAA CTC CTG ATC TAT AAG GCG TCT AGT TTA GAA AGT GGG GTC CCA TCA 180

AGG TTC AGC GGC AGT GGA TCT GGG ACA GAG TTC ACT CTC ACC ATC ACC

AGC CTG CAG CCT 240

GAT GAT TTT GCA ACT TAT TTC TGC CAA CAC TAT AAT CGA CCG TGG ACG

TTC GGC CAA GGG 300 ACC AAG GTG GAA ATC AAA GCA

IM9 Light Protein SEQ I.D. No. 11

DIQMTQFPSTLSASVGDRVTITCRASQSISAWLAWYQQKPGKAPKLLIY KASSLESGVPSRFSGSGSGTEFTLTITSLQPDDFATYFCQHYNRPWTFGQGTKVEIK

i . Southern blot and sequence analysis of light chain cDNA clones.

Ten putative kappa light chain clones were raised in LB broth with ampicillin. The plasmids were purified by the miniprep method of Holmes and Quigley (D.S. Holmeε and M. Quigley, Analytical Biochemistry, 114:193. 1981). The miniprep DNA waε characterized by reεtriction enzyme mapping and Southern blot analyεiε. The longeεt of the cDNA inserts obtained (clone kappa LI) waε 1.2kb. This clone was sequenced on the Genesis 2000 automated DNA sequencer (DuPont, Wilmington, Delaware) as described previously.

j . Rescreening, Southern blot, and sequence analysis of heavy chain cDNA clones.

None of the gamma clones was positive, so the library was rescreened as described in f. with the gamma constant region as a probe. The positives from this screening were picked and reεcreened aε deεcribed above in g. and h. until pure cultures were obtained. The putative clones were raised and characterized aε deεcribed in i. and two gamma cDNA cloneε were found. The cloneε were both 1.6kb in length. The cloneε were εequenced on the Geneεiε 2000 automated DNA sequencer (DuPont) as described previously. The Sequence is provided below as Sequence I.D. 12 and the correεponding amino acid Sequence iε provided as Sequence I.D. 13.

SEQ. I.D. No. 12

GAA ATG CAA CTG GTG GAA TTT GGG GGA GGC CTG CTA CAG CCT GGC AGG

GCC CTG AGA CTC 60 TCC TGT GCA GCC TCT GGA TTC AGG TTT GAT GAT TAT GCC ATG CAC TGG

GTC CGG CAA ACT 120

CCA GGG AAG GGC CTG GAG TGG GTC GCA GGT ATT AGT TGG AAT AGT GAC

ACC ATA GAC TAT 180

GCG GAC TCT GTG AAG GGC CGA TTC ACC ATC TCC AGA GAC AAC GCC AAG AAC TCC CTC TAT 240

TTG CAA ATG AAC AGT CTC AGA GCT GAC GAC ACG GCC TTG TAT TAC TGT

ACA AAA AGA AGG 300

GGG GTG ACA GAC ATT GAC CCT TTT GAT ATC TGG GGC CAA GGG ACA ATG

GTC ATC GTC TCT 360 TCA GAG 366 IM9 HEAVY PROTEIN SEQ I.D. No. 13

EMQLVEFGGGLLQPGRALRLSCAASGFRFDDYAMHWVRQTPGKGLEWVAGISWNSDTIDYA DSVKGRFTISRDNAKNSLYLQMNSLRAEDTALYYCTKRRGVTDIDPFDIWGQGTMVIVSS

Example 3.

Definition of Structurally Conserved Regions an _. identification of chain association residues usin . known three-dimensional structures Q_£ antibodies.

a . Definition of Light Chain SCRs

First, the linear amino acid εequences of the light chain variable regionε of a εet of antibodies with known three-dimensional structureε were compared. Eight εequenceε [Table 2] were compared in thiε example, but more or leεε may be used, by linear display of one εequence above the other on the computer screen [Figure 6] (SEQ. ID No. 14- 21)

Zafelέ 2

Antibody

Identifier PDB File Name Name_. Source Resoluti

MCP PDB1MCP.ENT MCPC60 Mouse 2.7A

FAB2 PDB4FAB.ENT 4-4-20 Mouse 2.7A

HFL PDB2HFL.ENT HYHEL-5 Mouse 2.54A

FDL PDB1FDL.ENT D1.3 Mouse 2.5A

FBJ PDB2FBJ.ENT J539 Mouse 1.95A

FABl PDB6FAB.ENT 36-71 Mouse 1.9A

FAB PDB3FAB.ENT NEW Human 2.0A

FB4 PDB2FB4.ENT KOL Human 1.9A SEQ I . D . No 14

DIVMTQSPSSLSVSAGERVTMSCKSSQSLLNSGNQKNFLAWYQQKPGQPPK LLIYGASTRESGVPDRFTGSGSGTDFTLTISSVQAEDLAVYYCQNDHSYPLTFGAGTKL

SEQ I.D. No 15

DWMTQTPLSLPVSLGDQASISCRSSQSLVHSQGNTYLRWYLQKPGQSPKV LIYKVSNRFSGVPDRFSGSGSGTDFTLKISRVEAEDLGVYFCSQSTHVPWTFGGGTKLE

SEQ I . D . No 16 DIVLTQSPAIMSASPGEKVTMTCSASSSVNYMYWYQQKSGTSPKRWIYDTS KLASGVPVRFSGSGSGTSYSLTISSMETEDAAEYYCQQWGRNPTFGGGTKLEIK

SEQ I . D . NO 17

DIQMTQSPASLSASVGETVTITCRASGNIHNYLAWYQQKQGKSPQLLVYYT TTLADGVPSRFSGSGSGTQYSLKINSLQPEDFGSYYCQHFWSTPRTFGGGTKLEIK

SEQ I.D. No 18 EIVLTQSPAITAASLGQKVTITCSASSSVSSLHWYQQKSGTSPKPWIYEIS

KLASGVPARFSGSGSGTSYSLTINTMEAEDAAIYYCQQWTYPLITFGAGTKLELK

SEQ I.D. No 19

DIQMTQIPSSLSASLGDRVSISCRASQDINNFLNWYQQKPDGTIKLLIYFT SRSQSGVPSRFSGSGSGTDYSLTISNLEQEDIATYFCQQGNALPRTFGGGTKLEIK

SEQ I.D. No 20

SVLTQPPSVSGAPGQRVTISCTGSSSNIGAGNHVKWYQQLPGTAPKLLIFH NNARFSVSKSGSSATLAITGLQAEDEADYYCQSYDRSLRVFGGGTKLTVL

SEQ I.D. NO 21 QSVLTQPPSASGTPGQRVTISCSGTSSNIGSSTVNWYQQLPGMAPKLLIYR DAMRPSGVPDRFSGSKSGASASLAIGGLQSEDETDYYCAAWDVSLNAYVFGTGTKVTVL

Using the Insight II Homology software to facilitate the three-dimensional alignment of these structures, a landmark amino acid, known to be universally conεerved among antibodieε, εuch aε the cysteine at L23 (Rabat, E. A., et al.. Sequences of Proteins of Immunoloσical Interest. Vol 1 edition, U.S. Department of Health and Human Services, PHS, NIH, Bethesda, Maryland (1991) ) was identified in each sequence. The sequenceε were vertically aligned on the computer εcreen.

Now, taking the firεt two of the linearly aligned εequenceε, one waε deεignated to be held conεtant and the other to be superimposed onto the first (In practice, the bottom εequence on the diεplay waε held constant due to the program deεign. ) A one reεidue box was drawn around the aligned cysteines. Then, using the commands for manual alignment of structureε, the program determined the minimum RMS (Root Mean Square) deviation, after applying the optimum rotation and tranεlation, of any boxed region. The minimum number of reεidues required in a box by this program before RMS deviation can be calculated in this way iε three. Aε an integral part of thiε process a visual representation of the superimpoεed εtructureε iε displayed on the screen. A three residue box was made, using the program, centered on the residue of interest (here, the cysteine L23) . The meaning of the box within this program iε to mathematically εuperimpoεe the εtructureε uεing the backbone atomε of the amino acidε within the box. The box was moved horizontally one residue in each direction, sequentially. The position giving the lowest RMS deviation for the superpoεition of backbone atomε of the three amino acids from the linearly aligned sequenceε waε εelected.

The object of thiε preliminary step was to approximately superimpose the two structureε, allowing structurally conserved regions (SCRs) to be discerned visually. Having achieved this objective, the box was now deleted. Using the already superimpoεed structureε, SCRs [uεually found in the regionε of the beta εheetε, but alεo in the other portionε of the framework regionε] were diεcovered by viεual inεpection. Uεing the Homology program, aε described previously in this εection, a box was made around the amino acid sequences that the SCR comprises . Gaps were introduced in the structurally non-conserved (NSCR) regions to align the SCR sequenceε.

For each SCR defined in thiε way, the structures of each of the other six sequences were superimposed, sequentially, in each case holding the same sequence (first sequence) constant, and the appropriate boxes were determined. Then, the second SCR was identified for the initial two structures and the procesε waε repeated working through each of the SCRs for all of the sequenceε (for example working from amino to carboxy terminuε) . Once the second set of SCRs waε εuperimpoεed, the program waε directed to εuperimpoεe the two structureε based on all of the backbone atoms of the residueε of both of the εets of SCRs. This procesε was also repeated for each of the subεequent εetε of SCRε.

Now that SCR boxeε had been determined for each of the εequences, consenεus boxeε were determined for each SCR. Consenεus boxes represent the maximum number of amino acid positionε (e.g. L60-L65 in Figure 6) contained in all of the SCR boxeε at a particular site. In this example εeven concenεus SCR boxes were formed as shown in Figure 6.

b. Definition of heavy chain SCRs using known three- dimensional structures of antibodies.

First, the linear amino acid εequenceε of the heavy chain variable regionε of a εet of antibodies with known three-dimensional structureε (we used eight sequences in this example, but more or less may be used) [Table 2] were compared by linear display of one sequence above the other on the computer screen [Figure 7] (SEQ I.D. No. 22-29)

SEQ I.D. No 22

EVKLVESGGGLVQPGGSLRLSCATSGFTFSDFYMEWVRQPPGKRLEWIAAS

RNKGNKYTTEYSASVKGRFIVSRDTSQSILYLQMNALRAEDTAIYYCARNYYGSTWYFDVW GAGTTVTVSS 96/06625 PC17US95/10791

- 70 -

SEQ I . D . No 23

EVKLDETGGGLVQPGRPMKLSCVASGFTFSDYWMNWVRQSPEKGLEWVAQIRNKPYNYETY YSDSVKGRFTISRDDSKSSVYLQMNNLRVEDMGIYYCTGSYYGMDYWGQGTSVTVSS

SEQ I.D. No 24

VQLQQSGAELMKPGASVKISCKASGYTFSDYWIEWVKQRPGHGLEWIGEILPGSGSTNYHE RFKGKATFTADTSSSTAYMQLNSLTSEDSGVYYCLHGNYDFDGWGQGTTLTVSS

SEQ I.D. No 25

QVQLKESGPGLVAPSQSLSITCTVSGFSLTGYGVNWVRQPPGKGLEWLGMIWGDGNTDYNS ALKSRLSISKDNSKSQVFLKMNSLHTDDTARYYCARERDYRLDYWGQGTTLTVSS

SEQ I.D. No 26

EVKLLESGGGLVQPGGSLKLSCAASGFDFSKYWMSWVRQAPGKGLEWIGEIHPDΞGTINYT PSLKDKFIISRDNAKNSLYLQMSQVRSEDTALYYCARLHYYGYNAYWGQGTLVTVSA

SEQ I.D. No. 27

EVQLQQSGVELVRAGSSVKMSCKASGYTFTSNGINWVKQRPGQGLEWIGYNNPGNGYIAYN EKFKGKTTLTVDKSSSTAYMQLRSLTSEDSAVYFCARSEYYGGSYKFDYWGQGTTLTVSS

SEQ I.D. No 28

VKLEQSGPGLVRPSQTLSLTCTVSGTSFDDYYSTWVRQPPGRGLEWIGYVFYHGTSDTDTP LRSRVTMLVNTSKNQFSLRLSSVTAADTAVYYCARNLIAGCIDVWGQGSLVTVSS

SEQ I.D. No 29

EVQLVQSGGGWQPGRSLRLSCSSSGFIFSSYAMYWVRQAPGKGLEWVAIIWDDGSDQHYA DSVKGRFTISRNDSKNTLFLQMDSLRPEDTGVYFCARDGGHGFCSSASCFGPDYWGQGTPV TVSS Uεing the Insight II homology software to facilitate the three-dimensional alignment of these structureε, a landmark amino acid, known to be univerεally conserved among antibodies, such as the cysteine at H22 (Kabat, E. A., et al..supra) was identified in each sequence. The sequenceε were vertically aligned on the computer εcreen.

Now, taking the first two of the linearly aligned εequenceε one waε designated to be held constant and the other to be superimpoεed onto the first. (In practice, the bottom sequence on the display was held constant due to the program design.) A one residue box was drawn around the aligned cysteineε. Then, using the commands for manual alignment of structures, the program determined the minimum RMS deviation, after applying the optimum rotation and translation, of any boxed region. The minimum number of residues required in a box by this program before RMS deviation can be calculated in this way iε three. Aε an integral part of thiε proceεs, a visual repreεentation of the εuperimposed structureε is displayed on the screen. A three residue box was made, using the program, centered on the residue of interest (here, the cysteine H22). The meaning of the box within this program is to mathematically superimpoεe the structures using the backbone atoms of the amino acids within the box. The box waε moved horizontally one residue in each direction, sequentially, and the position giving the loweεt RMS deviation for the superposition of backbone atoms of the three amino acids from the linearly aligned sequences was selected.

The object of this preliminary step was to approximately superimpose the two structureε, allowing SCRs to be discerned visually. Having achieved this objective, the box was now deleted. Using the already superimposed structureε, SCRs [usually found in the regions of the beta sheets, but also in other portions of the framework regions] are discovered by visual inspection and put within boxes including appropriate amino acids, guided by the RMS deviations. Once vertically aligned, the box waε expanded in both directionε to include more amino acids, until the RMS deviation became unacceptable (usually >0.75 A). Then the size of the box was reduced to the size which had the last acceptable RMS deviation. Gaps are introduced in the εtructurally non-conεerved (non-homologous) regions to help align the SCRε vertically.

For each SCR defined in thiε way, the structures of each of the other six sequenceε were superimposed, sequentially, in each case holding the same sequence (first sequence) conεtant, and the appropriate boxeε were determined. Then, the next SCR was identified for the initial two structures and the process was repeated working through each of the SCRs for all of the εequenceε (for example working from amino to carboxy terminuε) . Once the εecond εet of SCRε waε superimposed, the program was directed to superimpose the two structures based on all of the backbone atoms of the residues of both of the sets of SCRs. This proceεε waε also repeated for each of the subsequent sets of SCRs. Now that SCR boxes had been determined for each of the sequenceε, conεenεuε boxeε were determined for each SCR. Conεenεus boxes repreεent the maximum number of amino acid poεitionε (e.g. H3-H6 of antibody FB4 in Figure 7) contained in all of the SCR boxes at a particular site. Thuε, the amino acidε contained in each conεenεus SCR box are structurally conserved among all of the database antibodies under conεideration. In thiε example ten concensus SCR boxes were formed as shown in Figure 7.

c . Identification of Chain Association residues in lnown structures.

For each of the known εtructureε uεed in defining the SCRs, described above, chain-asεociation residues were identified. First, all residues from the light chain which contain any atom which is within about 4.5 A of any atom of any heavy chain residue, except those of the Kabat-defined CDRs, were identified. This set was then limited to those which by orientation have a significant likelihood of interaction with any atom of that heavy chain residue. All reεidueε from the heavy chain which contain any atom which iε within about 4.5A of any atom of any light chain reεidue, except thoεe of the Kabat-defined CDRε, were alεo identified, again limited to thoεe which have a εignificant likelihood of interaction. This process was carried out for each of the antibodies of known structure shown in Table 2.

Example 4

Three-dimensional Modeling of ZCE Fv

a . Three-dimensional modeling of ZCE Light chain variable domain.

The three dimensional coordinateε had not been determined for ZCE Fv. For thiε reaεon, homology modeling waε used to approximate the actual structure. The following four εtepε were used: (1) alignment of the ZCE light chain variable region εequence with the aligned sequences of the set of light chain variable regionε of known structure described in Example 3.a.; (2) homology modeling of SCRs using SCRs from the known light chain variable region structureε; (3) homology modeling of NSCRs using the full range of known structures available in the Brookhaven database, and (4) a series of energy minimizations carried out to obtain an energetically favorable structure.

(1) Alignment of ZCE Q25 light ςfaύn amjno aςjd seσuenςe wit amino acid sequences of known liσht chain structureε. The linear εequence of the ZCE 025 light chain variable region, determined from a cDNA clone aε described in Example l.a., above, waε displayed and aligned with the databaεe εequenceε deεcribed in εtep 1 above, using the Insight II εoftware. Aε deεcribed for the database sequences, the first step was to align the ZCE 025 sequence with the database sequenceε uεing the first consensus SCR box. This was accomplished by first identifying which residues (one or more) within the box were most highly conεerved between the known εtructures, identifying theεe reεidueε in the ZCE εequenceε, and aligning them. In some caseε only a εubεet of the reεidueε identified aε conεerved appeared in the ZCE εequence. In theεe inεtanceε, the εubεet waε aligned. Working from one concenεuε SCR box to the next (in thiε example, we worked from amino to carboxy terminuε) the process was repeated. Where necesεary, gapε were introduced either into regionε other than thoεe correεponding to SCRε (i.e. NSCRs) from ZCE or into identical positionε within the SCRs of each of the aligned known εtructureε.

The object of thiε preliminary εtep waε to align the ZCE εequence with the sequences of the other light chain variable regions of known structure. In each case great effort was made to identify the potential locations of ZCE SCRs by linear sequence homology to the consenεuε regionε alone. The reεult of thiε alignment iε shown in Figure 8.

(2) - Three dimensional modeling of SCRε. For each SCR, the actual known εtructure whose sequence has the greatest homology to the corresponding ZCE light chain SCR was εelected aε template for that segment, and its coordinates were assigned to the ZCE SCR. If there were a residue in a template SCR that did not match the corresponding residue in the ZCE SCR residue, the residue in the template waε mutated to match the ZCE SCR reεidue, while maintaining the coordinateε of all the atomε in the backbone and side chainε of the template residue that correspond to those in the ZCE residue and modeling the remaining atomε under the constraints of maintaining the same bond lengths, angles and dihedrals as thoεe in the original database residue, e.g., for gamma and delta carbons. Thiε waε done for each SCR (we worked from amino to carboxy terminuε) . After all of the SCRε were assigned coordinates in this manner a partial three-dimenεional structure comprising the modeled SCRε waε displayed, absent the NSCRs. (3) - Three dimensional modeling of NSCRs. For each ZCE light chain NSCR, the flanking SCRε of assigned coordinateε were used along with the length of the NSCR to identify a known structure with the greatest likelihood of being homologouε to the SCR/NSCR/SCR array. This was accomplished by using the "Loop Search" subprogram in Inεight II to εearch the databaεe for εtructures (1) containing proper lengths of flanking and spanning sequences and (2) having coordinates for the flanking sequences with the least RMS deviation from those of the assigned SCRs. In practice, approximately ten structures (more or lesε can be uεed) were ranked by the program on the baεiε of RMS deviation of the flanking εequences. These were sequentially displayed on the screen superimpoεed on the flanking SCRs. The structure best approximating the flanking sequences and having the same general orientation as NSCRs from light chain variable regions of known structure was chosen as template for that particular NSCR and its coordinateε were aεsigned to the ZCE NSCR. Thiε process was then repeated for each NSCR, and the NSCRs were added to the computer model by inserting each in itε appropriate place, for instance flanked by the adjoining SCRs, until the entire variable region had been modeled.

(4) - Energy Minimizations of modeled structure. Energy minimizations were carried out in stageε to aεsure that no major structural disruptions would occur. Once all of the NSCRs making up the model had in turn been selected from the database, fixed in space, and modeled to transform them into the corresponding ZCE NSCRs, the splice regions where the

SCRs join the NSCRs were refined to relieve any strain in the model that would result from joining the SCRs and NSCRs, using the "Repair" algorithm to asεign the proper bond lengths, bond angles, and omega values to the structures. Now, the "Relax" algorithm was applied in a series of sequential steps to the model as a whole: (1) to the side chains of the NSCRs to assign proper geometries, and remove any unfavorable non-bonded contactε between εide chain atoms and other atoms in the molecule, (2) to all atomε of the NSCRs, and (3) to the mutated side chains of the SCRε. In each of these steps, all regions other than those that are being relaxed remain fixed to their asεigned coordinates. Finally, an energy minimization analysis waε performed using the "Discover" εubprogram to allow the model to aεεume an energetically favorable structure. First the entire model was subjected to energy minimization with backbone atomε tethered to their εtarting coordinateε with a defined force conεtant (uεually 100 Kcal/A²). Then energy minimization waε performed on the entire molecule without the backbone atomε being tethered.

The reεult of carrying out theεe εteps waε a model of the ZCE light chain.

b. Three-dimensional modeling of ZCE 025 Heavy chain variable domain.

Like the light chain, coordinates had not been determined for ZCE heavy chain. For this reason, homology modeling was again used to approximate the actual structure. The same steps were used for heavy chain as for light.

(1) - Alignment of ZCE 025 heavy chain amino acid sequence with amino acid sequences of known heavy chain structures. The linear sequence of the ZCE heavy chain variable region, determined from a cDNA clone as described in Example l.b., above, was displayed and aligned with the database sequenceε deεcribed in Example 3.b. above, using the Insight II software. As described for the light chain, the first εtep waε to align the ZCE sequence with the database sequences uεing the first consenεuε SCR box. The remainder of the proceεε waε precisely as described for the light chain, with the final alignment displayed in Figure 9.

(2) - Three dimensional modeling of SCRs. For each SCR, the actual known structure whose sequence has the greatest homology to the corresponding ZCE SCR was selected as template for that particular SCR (these are shown in bold in Figure 9 and its coordinates were assigned to the ZCE SCR. This was done for each SCR (we worked from amino to carboxy terminus) . After all of the SCRs were asεigned coordinates in this manner a partial three-dimensional structure was displayed, absent the NSCRs.

(2) - Three dimensional modeling of NSCRs. AS for light chain, for each ZCE heavy chain NSCR, the flanking SCRs which had been assigned coordinates were used along with the length of the NSCR to identify a known structure with the greatest likelihood of being homologouε to the SCR/NSCR/SCR array, using the "Loop Search" subprogram in Insight II. This process was then repeated for each NSCR, until the entire variable region had been modeled.

(4) - Energy Minimizations of modeled structure. Energy minimizations were carried out in stages, as for the light chain, to assure that no major structural disruptionε would occur. The proceεε waε identical to that deεcribed for the light chain, including (1) use of the "Repair" algorithm to asεign the proper bond lengths, bond angles, and omega values to the structures; (2) use of the "Relax" algorithm to assign proper geometries and remove any unfavorable non-bonded contacts; (3) use of the "Discover" subprogram to allow the model to asεume an energetically favorable εtructure. Once again, the entire molecule was first subjected to energy minimization with backbone atoms tethered to their starting coordinates with a defined force constant (usually 100 Kcal/A²). Then energy minimization waε performed on the entire molecule without the backbone atomε being tethered.

The result of carrying out these stepε waε a model of the ZCE heavy chain. c. Three-dimensional modeling of ZCE 025 heavy and light chains together to form Fv.

Now that coordinates had been determined for both light and heavy chain, these were displayed on the εcreen together. Since energy minimization has been carried out on each chain separately, the first εtep waε to carry out the εame procedure on the chainε aε a εet. Finally, an energy minimization was performed uεing the "Discover" subprogram to allow the model to asεume an energetically favorable configuration. Firεt, potential chain association residues for ZCE light and heavy chains were identified by comparison with the chain association residueε of the known structures (determined in Example 3.c). Chain asεociation reεidueε in the aligned εequenceε were compared with reεidueε in the correεponding poεition in ZCE. When an identical residue was present, it waε designated aε favorable to chain asεociation; if a different residue was found, it was designated as potentially disrupting to chain association. Totals of favorable and disrupting residueε were determined for the compariεon of ZCE light and heavy chainε to each of the known structures. The known structure providing the comparison having the greatest excesε of favorable reεidues over disruptive residueε waε choεen as template for ZCE heavy/light association. If two or more known εtructures had the same excess of favorable over disruptive residueε, the εtructure having the greateεt number of favorable residues over disruptive residueε waε chosen. In thiε example 2HFL was chosen.

Next, the light chain structure determined for ZCE in Example 4.a. was superimposed on that of the light chain structure of 2HFL, using the backbone coordinateε of the favorable reεidues described above. This was carried out using the "superimpose" command in the Insight II software. The same was done for the ZCE heavy chain using the 2HFL heavy chain. Next the entire molecule was subjected to energy minimization with backbone atoms tethered to their starting coordinates with a defined force constant (uεually lOOKcal/A²). Then energy minimization was performed on the entire light/heavy association model without the backbone atoms being tethered.

d. Identification of CDR-Associated Residues in ZCE 025 Fv.

For ZCE, the aim of modeling is to identify regions that must be conserved to conserve the function of the CDRε. To do this it is necesεary to (1) identify all potential CDR- asεociated reεidues and (2) identify the subεet of these which have a reasonable likelihood of a significant interaction with the CDR residue involved.

To determine which amino acids lying outside of the Kabat-defined CDRs may influence binding of the CDRε, all the amino acid residues outside of the Kabat-defined CDR regions that have an atom located within about 4.5 Angstromε of any atom in an amino acid located in the Kabat-defined CDR regionε of the ZCE conεtruct were identified aε CDR- aεεociated residues. As these were predicted to be important for maintaining the binding specificity of the ZCE antibody, they were earmarked for preservation aε donor amino acidε in the CDR grafted antibody conεtruct in addition to thoεe in the defined CDR regionε.

We first identified all residues on the light or heavy chain that have atoms that are within 4.5 A of any atoms of any light chain CDR residue. The set was limited to those with a significant likelihood of interaction, baεed on orientation of the residue, charge, hydrophobicity, etc. Next, all residueε on the light or heavy chain that contain atomε which are within 4.5 A of any atom of any heavy chain CDR reεidue were identified. Again, the εet waε limited to those with a high likelihood of significant interaction with the CDR residue of intereεt. In this way, the entire set of light and heavy chain CDR-associated residues was determined. Example 5 Three dimensional modeling of IM9 Fv.

a . Three-dimensional modeling of IM9 Light chain variable domain.

Homology modeling waε uεed to approximate the structure of the IM9 antibody. Thiε proceεε had four εteps: (1) alignment of the IM9 light chain variable region sequence with the aligned sequences of the set of light chain variable regions of known structure. (See Example 3.a. above); (2) homology modeling of the IM9 light chain SCRε uεing SCRε from the known light chain variable region εtructureε; (3) homology modeling of NSCRε (non-structurally conserved regionε) uεing the full range of known εtructureε available in the Brookhaven databaεe (other known structureε could alεo be uεed); and (4) a series of energy minimization routines to determine the energetically preferred structure.

(1 ) Alignment of IM9 light chain amino acid sequence with amino acid sequences of known light chains.

The linear DNA sequence [SEQ. I.D. No. 10] of the IM9 light chain variable region was determined from a cDNA clone aε deεcribed in Example 2.i.:

The linear amino acid sequence [SEQ. I.D. No. 11] of the IM9 light chain variable domain was displayed on the computer screen and aligned with the sequences of the eight light chain variable regions of known structure described in Example 3.a. above, using the Insight II software. The IM9 sequence was aligned with the database εequenceε uεing the firεt consensuε SCR box. The reεidueε (one or more) within the box which were moεt highly conserved between the known structures were identified, after which the corresponding reεidues in the IM9 sequences were identified and the structures were aligned. When a subset of the residues identified as conserved in the known structures appeared in the IM9 sequence the subset was aligned. The alignment proceeded from one consensus SCR box to the next as described above for the known sequences. The alignment proceeded from amino to carboxy terminus, but would work as well if reversed. Gaps were introduced either into regions other than those corresponding to SCRs (i.e. NSCRs) from IM9 or into identical positions within the SCRs of each of the aligned known structures when necessary for alignment.

This preliminary step allowed alignment of the IM9 sequence with the sequences of the other light chain variable regions of known structure. The potential locations of IM9 SCRs were identified by linear sequence homology to the consensus regions. This alignment is shown in Figure 10.

(2) Three-dimensional modeling of IM9 light chain SCRs . For each SCR, the actual known light chain structure whose sequence had the greatest homology to the corresponding IM9 light chain SCR was selected as the template for that segment (these are shown in bold in Figure 10 and its coordinates assigned to the IM9 SCR. In instances where a residue in a template SCR did not match the corresponding residue in the IM9 SCR, the coordinates of all the atoms in the backbone and sidechains of the template residue that correspond to those in the IM9 residue were maintained. The remaining atoms (e.g., for gamma and delta carbons and the atoms bonded to them) were modeled under the conεtraintε of maintaining the εame bond lengths, angles and dihedrals as those in the original database residue. This was done for each SCR (we worked from amino to carboxy terminus) . After all of the SCRs were assigned coordinates in thiε manner a partial three-dimensional εtructure comprising the modeled SCRs was displayed, absent the NSCRs.

(3) Three-dimensional modeling of IM9 light chain NSCRs. For each IM9 light chain NSCR, the flanking SCRs which had been assigned coordinates were uεed along with the length of the NSCR to identify a known structure with the greateεt likelihood of being εtructurally homologouε to the SCR componentε of the SCR/NSCR/SCR array. In addition, the known εtructure containing a region corresponding to the NSCR component of the aforementioned SCR/NSCR/SCR array, is identified which has an orientation most like that of the corresponding region of the antibodies of known εtructure. Thiε waε accompliεhed by using the "Loop Search" subprogram in Insight II to search the database for εtructures (1) containing proper lengthε of flanking and εpanning εequences and (2) having backbone coordinateε for the flanking εequenceε with the leaεt RMS deviation from thoεe of the aεεigned SCRε. In practice, a maximum of ten εtructureε (more or leεε can be used depending on the limitations of the program used) were ranked by the program on the basis of RMS deviation of the coordinates of the backbone atoms of the flanking sequenceε. Theεe were εequentially diεplayed on the εcreen superimposed on the flanking SCRs. The structure beεt approximating that of the flanking sequences, having the same general orientation as NSCRs from light chain variable regions of known structure, and having a minimum of εtructurally εignificant mutationε waε choεen aε template for that particular NSCR and its coordinates were asεigned to the NSCR. This procesε waε then repeated for each NSCR, until the entire variable region had been modeled.

(4) Energy Minimizations of modeled IM9 light chain structure.

Energy minimizations were carried out in stageε to aεεure that no major εtructural diεruptionε would occur. Firεt the εplice regions where the SCRs join the NSCRs were refined to relieve any strain in the model that would result from joining the SCRs and NSCRε, using the "Repair" algorithm to assign the proper bond lengths, bond angles, and omega valueε to the residues in the splice region. Then, the "Relax" algorithm was sequentially applied to the regions as follows: (1) to the sidechainε of the NSCRs to asεign proper geometrieε, and remove any unfavorable non-bonded contacts between NSCR sidechain atoms and other atoms in the molecule; (2) to all atomε of the NSCRε to remove remaining unfavorable contactε between the

NSCR and other atomε in the molecule; (3) to the altered εide chainε of the SCRε to remove any unfavorable non-bonded contactε between mutated SCR εide chain atomε and other atoms in the molecule, and (4) to all the sidechain atoms of the SCR to remove remaining unfavorable side chain contacts. In each of the above described stepε, all regions other than those which are being "relaxed" remain fixed to their asεigned coordinateε.

Finally, an energy minimization waε performed uεing the "Discover" program to allow the model to assume an energetically favorable structure. First the entire model was subjected to energy minimization with backbone atoms tethered to their starting coordinates with a defined force constant (usually 100 Kcal/A²). Then energy minimization was performed on the entire molecule without the backbone atoms being tethered.

The result of carrying out theεe εtepε waε the homology model of the IM9 light chain. b. Three-dimensional modeling of IM9 Heavy chain variable domain.

The steps used to model the IM9 heavy chain are similar to those used in modeling the IM9 light chain.

( 1 ) Alignment of IM9 heavy chain amino acid sequence with amino acid sequences of known heavy chain structures.

The linear DNA εequence [SEQ. I.D. NO. 12] of the IM9 heavy chain variable region waε determined from a cDNA clone aε deεcribed in Example 2.j.

The linear amino acid εequence [SEQ. I.D. NO. 13] of the IM9 heavy chain variable domain waε diεplayed and aligned with the databaεe sequenceε, deεcribed in

Example 3.b. above, uεing the Inεight II εoftware. Aε described for the light chain, the first step was to align the IM9 sequence with the database sequences using the first conεensuε SCR box. The remainder of the process was precisely as described for the light chain, with the final alignment displayed in Figure 11.

(2) Three-dimensional modeling of IM9 heavy chain SCRs . For each SCR, the actual known εtructure whose sequence haε the greatest homology to the corresponding IM9 SCR was selected aε the template (εhown in bold in Figure 11) and itε coordinateε aεεigned to the correεponding IM9 SCR. The process - working from amino to carboxy terminus - was repeated for each SCR. After all of the SCRs were assigned coordinates a partial three-dimensional structure was displayed, absent the NSCRs.

(3) Three-dimensional modeling of IM9 heavy chain NSCRs.

As for the light chain, for each IM9 heavy chain NSCR, the flanking SCRs which had been assigned coordinates were uεed along with the length of the NSCR to identify a known structure with the greatest likelihood of being structurally homologous to the SCR components of the SCR/NSCR/SCR array. In addition, the known structure containing a region correεponding to the NSCR component of the aforementioned SCR/NSCR/SCR array, iε identified which has an orientation moεt like that of the corresponding region of the antibodieε of known structure. This was accomplished by using the "Loop Search" εubprogram in Inεight II to εearch the databaεe. Thiε process was then repeated for each NSCR, until the entire variable region had been modeled.

(4) Energy Minimizations of modeled IM9 heavy chain structure.

Energy minimizations were carried out in stages, as for the light chain, to assure that no major structural disruptionε would occur. The procesε uεed waε in εubstantial accordance with that described for the light chain. The procesε comprised the following steps: (1) use of the

"Repair" algorithm to assign the proper bond lengths, bond angles, and omega values to the splice regions; (2) use of the "Relax" algorithm to assign proper geometries and remove any unfavorable non-bonded contacts from the mathematical model; (3) uεe of the "Discover" subprogram to allow the model to assume an energetically favorable structure. As described for the light chain, the entire molecule waε firεt εubjected to energy minimization with backbone atomε tethered to their εtarting coordinateε with a defined force constant (usually 100 Kcal/A²) . Then energy minimization was performed on the entire molecule without the backbone atoms being tethered.

The resultant structure was used as the model of the IM9 heavy chain. c . Three-dimensional modeling of IM9 heavy and light chains together to form Fv.

The coordinates determined for the light and heavy chain, were used to generate a model of the Fv. As an initial step, potential chain asεociation reεidueε for IM9 light and heavy chainε were identified by compariεon with the chain aεεociation reεidues of the known structures (determined in Example 3.c). Chain association residues in the aligned sequences were compared with residueε in the corresponding position in IM9. When an identical reεidue was present, it was designated aε favorable to chain association; if a different residue waε found, it was deεignated aε potentially diεrupting to chain aεsociation. Totals of favorable and disrupting residues were determined for the comparison of IM9 light and heavy chains to each of the known structures. The known structure providing the comparison having the greatest excess of favorable residues over disruptive residues was chosen as template for IM9 heavy/light association. If two or more known structureε had the εame exceεs of favorable over disruptive residueε, the εtructure having the greatest number of favorable residueε waε choεen aε template. In thiε example, FDL waε choεen.

Next, the light chain structure determined for IM9 in Example 5.a. was superimposed on the template light chain structure of FDL, using the backbone coordinates of the favorable residues described above. This was carried out using the "superimpose" command in the Insight II software. The same was done for the IM9 heavy chain using the FDL heavy chain. Next the entire molecule was subjected to an energy minimization with the backbone atoms tethered to their starting coordinates with a defined force constant (uεually 100 Kcal/A²). Then an energy minimization was performed on the entire light/heavy associated (Fv) model without the backbone atoms being tethered. d. Identification of Chain-Association Residues in IM9 Fv.

The regionε of IM9 that should be conserved to allow for optimal associations between the chains in regionε other than thoεe that will be replaced (the CDRs and CDR associated regionε) was determined by (1) identification of all chain asεociation residueε; (2) identification of all CDR asεociated reεidueε; and (3) delineation of the not CDR- associated subεet of chain aεεociation reεidueε. The individual εtepε are described in detail below.

Reεidueε from the light chain that contain an atom that iε within about 4.5A of any atom of any heavy chain reεidue were identified. Thiε εet waε then limited to thoεe reεidueε that have a εignificant likelihood of interacting with that heavy chain reεidue (or any other) . All residues from the heavy chain containing an atom that is within about 4.5 A of any atom of any light chain residue were identified, again limited to those that have a significant likelihood of interaction. Next, all residues on the light or heavy chain that contain an atom that is within about 4.5A of any atom of any light chain CDR residue were identified. Again the εet iε limited to those with a significant likelihood of interaction. Next, all residues on the light or heavy chain that contain an atom that is within about 4.5 A of any atom of any heavy chain CDR residue with a high likelihood of significant interaction with the CDR residue of intereεt were identified. Finally, the εubεet of chain aεεociation reεidueε not contained within either set of CDR-associated residueε was determined and claεεed aε IM9 chain-aεεociation residues. Example 6 Three-dimensional modeling g_f Humanized, Z£I 0ΣL5 Fy_t.

a. Modeling of CDR-grafted ZCE/IM9 light chain variable region.

The IM9 and ZCE light chain amino acid εequenceε were aligned with reference to the εequences of the eight known structureε. On the ZCE linear array, the Kabat-defined CDRε and the CDR-aεεociated reεidueε determined in Example 4 were identified. For SCR or NSCR regions which do not contain a CDR or CDR-associated residue, the entire region was replaced with the IM9 sequence. For SCRs which contain one or more CDR or CDR-associated residues, the non-CDR and non-CDR-asεociated reεidueε were replaced with IM9 εequence, but the ZCE sequence was conserved for the CDR or CDR- associated residues. For NSCRs which contain one or more CDR or CDR-asεociated reεidueε, the replacement iε dependent upon the relative lengths of the region of interest in acceptor and donor molecules. If the NSCR has the same number of residueε in both the acceptor (IM9) and the donor (ZCE) molecules, the non-CDR asεociated reεidueε were replaced with acceptor (IM9) εequence. If however, the NSCR differε in number of reεidues between the acceptor and donor, the donor (ZCE) εequence was conserved for the entire segment. In this way the primary sequence for the light chain CDR-grafted molecule waε determined. The residues of the CDR-grafted primary sequence were assigned coordinates to match those of the residueε in the light chain εequenceε of the superimposed models of ZCE and IM9 from which they were derived. This was done working from amino to carboxy terminus.

b. Modeling of CDR-grafted ZCE/IM9 heavy chain variable region.

The IM9 and ZCE heavy chain amino acid sequences were aligned with reference to the sequences of the eight known heavy chain structureε. On the ZCE linear array, the Kabat-defined CDRs and the CDR-associated residueε determined in Example 4 were identified. For SCR or NSCR regions which do not contain a CDR or CDR-associated residue, the entire region waε replaced with the IM9 sequence. For SCRs which contain one or more CDR or CDR-associated residueε, the non- CDR and non-CDR-associated residueε were replaced with IM9 εequence, but ZCE εequence waε conεerved for the CDR or CDR- aεεociated reεidues. For NSCRs which contain one or more CDR or CDR-associated residues, the ZCE sequence was conserved for the entire region. In this way the amino acid sequence for the heavy chain CDR-grafted molecule was determined. The coordinates of the residues of the CDR-grafted primary sequence were obtained from those of the residues in the heavy chain sequences of the superimposed models of ZCE and IM9 from which they were derived. This was done working from amino to carboxy terminus.

c . Modeling of Humanized ZCE Fv.

Now that coordinates had been assigned for both light and heavy chain, these were displayed on the screen together. An energy minimization was performed using the "Discover" subprogram to allow the model to asεume an energetically favorable structure. First the entire model was subjected to energy minimization with backbone atoms tethered to their starting coordinates with a defined force constant (usually 100 Kcal/A²) . Then the energy minimization was performed on the entire model without the backbone atomε being tethered.

d. Modification of the humanized ZCE 025 model so that only CDR-associated residues found in the murine ZCE 025 model meet the definition of CDR- associated residues.

CDR-aεεociated residues were determined for the modeled humanized ZCE Fv in substantial accordance with the methodology taught for the original ZCE Fv. First, all residues on the light or heavy chain which contain atoms which are within about 4.5 A of any atoms of any light chain - 90 -

CDR reεidue, which alεo haε a εignificant likelihood of interaction, baεed on orientation, charge, hydrophobicity, etc. were identified. Next, all atomε of all reεidueε on the light or heavy chain which are within about 4.5 A of any atomε of any heavy chain CDR reεidue were identified. The εet waε then limited to thoεe with a high likelihood of εignificant interaction with any atomε of the CDR reεidue of intereεt. The entire εet of light and heavy chain CDR- aεεociated reεidueε was thusly determined. The εet of CDR-aεεociated reεidues determined for the humanized Fv was compared to that determined for the ZCE Fv. In any caεe where an additional CDR-aεsociated residue waε present in the humanized, the amino acid at that position was replaced by the amino acid found in the murine ZCE. In the case where a CDR-asεociated reεidue in ZCE waε not identified aε CDR-aεsociated in the humanized ZCE and is found in a NSCR, the entire NSCR was changed to the donor (ZCE) sequence.

e. Confirmation of Chain-association residues.

After the CDR-aεεociated residues were modified if neceεεary as described above, the model waε analyzed to determine if the chain aεεociation reεidueε identified for IM9 were conεerved. In thiε example, they were conεerved. If, however, differences are observed, they are noted, but no changes are made at this time. If, in addition, a significant decreaεe in εecreted protein iε obεerved for the humanized molecule, theεe are potential εiteε for modification. The amino acid εequenceε for light and heavy chain hZCE, determined above, are εhown in Figure 7 and Figure 8, reεpectively. Exam le 7

Modeling of hZCE-CSVL and hZCE-kb Fv.

a. Modeling of hZCE-CSVL.

The IM9 light and ZCE heavy chain primary amino acid sequenceε had already been aligned with reference to different εequences. Therefore, it was necessary to bridge these alignments through realignment using a common sequence. The IM9 heavy chain sequence was used for this purpose as shown in Figure 12. In addition, the IM9 heavy chain provided information on chain association residueε. ZCE heavy chain εequence waε added and aligned with the linear array containing light chain ZCE and light and heavy chain IM9 εequenceε. Once aligned in thiε manner, SCRε were defined there between aε deεcribed in Example 3, the Kabat defined CDRs and CDR-associated residueε determined in Example 4, were identified on the ZCE heavy chain linear array. For SCR or NSCR regions which do not contain a CDR or CDR-aεεociated residue, the entire region was replaced with the IM9 light chain εequence (and structure, i.e., coordinates) . For SCRε which contain one or more CDR or CDR- aεsociated regions, the non-CDR-associated residueε were replaced with IM9 εequence (and structure, i.e., coordinates) , but ZCE heavy chain sequence (and structure, i.e., coordinates) was conserved for the CDR-aεsociated residues. For NSCRs that contain one or more CDR or CDR- asεociated residues, the ZCE heavy chain εequence (and structure, i.e., coordinates) was conserved for the entire region. In thiε way the primary sequence for the heavy chain CDR-grafted molecule waε determined, and a composite structure was developed.

Now, the resultant model was modified to asεure that chain association residueε, derived from the IM9 model were conεerved. In all non-CDR or non-CDR-associated regions, when the amino acid in the position occupied by the chain association residue was different than the corresponding IM9 heavy chain residue, it was replaced with the corresponding IM9 heavy chain chain-association residue. In thiε example no chain-association residueε were found to lie in the CDR or CDR-aεεociated regionε. In the unlikely event that thiε should occur, the residue should be noted, but no change should be made. In addition one residue leucine at position 94, of the mature CSVL was changed to a methionine. The final amino acid sequence of the mature CSVL is shown in Figure 13 [SEQ I.D. No. 30]

SEQ I.D. No. 30

DIQMTQFPST LSASVGDRVN ITCRASGFTF SNFGMHWIRQ KPGKGLKWVA YISGGSSTVH YADSLKGRFT ISRDNPKNEL FLTITSLQPD DFAMYYCARD YYVNNYWYFD VWGQGTKVEI KR (122 residueε)

Alternatively, hZCE light chain can be used as acceptor and hZCE heavy chain can be used as donor. In thiε caεe, chain aεsociation residueε uεed for the preliminary Fv model are thoεe identified for hZCE FV.

b. Model hZCE-kb Fv.

Now that coordinates had been assigned for both light and heavy/light hybrid chains, these were displayed on the screen together. An energy minimization was performed using the "Discover" subprogram to allow the model to assume an energetically favorable configuration. First the entire model was subjected to energy minimization with backbone atoms tethered to their starting coordinates with a force constant of 100 Kcal/A². Then the energy minimization algorithm was applied to the entire model without the backbone atoms being tethered.

c . Modify to assure no added CDR-associated residues . CDR-Associated residues were determined for the modeled humanized ZCE light chain dimer as for the original ZCE Fv of Example 4.d. Again, this was done by first identifying all reεidues on the light or heavy/light hybrid chain that are within 4.5 A of any light chain CDR residue, and that also have a εignificant likelihood of interaction, baεed on orientation of the residue, charge, hydrophobicity, etc. Next, all residues on the light or heavy/light hybrid chain that were within 4.5 A of any heavy/light hybrid chain CDR residue were identified. Again, the set was limited to those with a high likelihood of significant interaction with the CDR residue of interest. In this way, the entire set of light and heavy/light hybrid chain CDR-associated residues was determined.

The set of CDR-asεociated residueε determined for the humanized light chain dimer was compared to that determined for the ZCE Fv. In any case where an additional CDR-associated residue is seen for the humanized, the amino acid at that position waε replaced by the amino acid found in the murine ZCE. Care should be taken in this step as these replacements would be dependent upon whether that residue lies in an SCR or NSCR segment as explained in Example 6 above.

d. Confirm Chain-association residues.

After the CDR-associated residueε were modified as neceεεary, the model was analyzed to determine if the chain association residueε identified for IM9 were conεerved. In thiε example, they were conεerved. If, however, differenceε are obεerved, these are noted, but no changes are made at this time; If there is a εignificant decrease in expression observed for the humanized molecule, these are potential siteε for modification. Example 8.

Construction of expression vectors PGIM9K and pGIM9K/hZCE-kappa.

a . Construction and Screening the IM9 Genomic library in E. coli Bacteriophage Lambda for the Ig Kappa Gene.

IM9 genomic DNA waε extracted and purified uεing methodε deεcribed in Sambrook (supra, pp. 9.4-9.30). The DNA waε partially digeεted with Mbol and εeparated by εucrose density gradient ultra-centrifugation. The gradients were fractionated and the aliquots were analyzed for size by agarose gel electrophoreεiε, aε deεcribed in Sambrook (supra, pp. 6.3-6.19). The fractions between 8-20 Kb were pooled, and dialyzed against TE Buffer (10 mM Tris HCl; 1 mM EDTA, pH 7.4). "Tris" is [Tris (hydroxymethyl)amino methane].

The IM9 DNA was ligated to Lambda EMBL3 arms (commercially available from Stratagene, San Diego,

California) and packaged with the lambda bacteriophage packaging kit, Gigapack® Gold (Stratagene). The recombinant bacteriophage particles were used to tranεfect E. coli εtrain P2/392, which waε inoculated onto 1% NZY agar medium in 140 mm diameter plateε. The lambda library contained 6.55 X 10⁵ individual cloneε, and waε amplified by plating at 3.3 X 10⁴ plaques per plate on twenty plateε and εuεpending the bacteriophage in 200 ml total of SM buffer (5.8 g NaCl, 2 g MgSθ₄-6H₂0, 50 ml 1 M TriεHCl, pH 7.5, and 5 ml 2% gelatin per liter) .

The library was plated as deεcribed in Sambrook (≤iiEta., pp. 2.61-2.63), on twenty, 140 mm agaroεe plates at 2.5 x 10^ plaques per plate. The lambda phage plaques were blotted onto nitrocelluloεe and treated with denaturing and neutralizing solutions followed by baking at 80 C in a vacuum oven. Filters were then pre-hybridized in 50% formamide, 5 X SSC (75 mM Na citrate; 750 mM NaCl), 0.1% SDS, 5 X Denhard 's εolution (0.1% bovine serum albumin (BSA) , 0.1% ficoll, 0.1% polyvinylpyrrolidone) , 200 μg/ml yeast tRNA, 100 μg/ml salmon sperm DNA at 42°C for 2 hours. Fragments of human immunoglobulin kappa chain DNA were labeled with a Prime-It® kit (commercially availble from Stratagene) in subεtantial accordance with the directionε provided by the manufacturer, and hybridized with the blotε overnight in hybridization εolution (50% forma ide, 5 X SSC , 0.1% SDS , 1 X Denhard 'ε solution (0.02% BSA, 0.02% ficoll, 0.002% polyvinylpyrrolidone), 100 μg/ml salmon sperm DNA) at 42°C. The blots were washed twice at 42°C in 2 X SSC and 0.1% SDS for 20 minutes, then at 65°C in 0.2 X SSC, 0.1% SDS for 20 minutes and exposed to XAR-5 X-ray film (commercially available from Eastman Kodak Corp.) overnight at -70°C between two intensifying screens.

The positive plaques were picked and subjected to two rounds of phage DNA purification as deεcribed in Sambrook (supra, pp. 2.73-2.76). The purified phage DNA was analyzed by restriction enzyme mapping and Southern blot, as deεcribed in Sambrook (supra, pp. 9.31-9.57). Figure 14 provides a restriction map of the IM9 kappa gene in bacteriophage lambda EMBL3.

b. Subcloning the Intact Kappa Gene into pBluescript ®

Southern Blot analysis was used to map the intact kappa chain gene to an 8.8 Kb BamHI fragment. This fragment was isolated from the lambda phage DNA by digestion with

BamHI followed by agarose gel electrophoresiε. The 8.8 Kb BamHI fragment waε ligated uεing T4 DNA ligase (commercially available from Life Technologies, Inc.) following manufacturers instructions, with pBluescript®SK^" (commercially available from Stratagene, San Diego, CA) which had been previously digested with BamHI. Restriction endonuclease mapping revealed the 5' end of the gene was adjacent to the Sacl end of the polylinker. In order to facilitate modification of the gene, the 5' end of the gene waε then εub-cloned aε a BamHI to BεtEII fragment containing the IgK promoter, the variable exonε and a portion of the major intron. The B_s_L.EII reεtriction endonuclease leaves a 5 ' overhang that is not compatible with any of the siteε in the pBlueεcript®SK^~ polylinker, εo it waε necessary to modify the overhanging sequence to make it blunt ended. Thiε waε carried out by digeεting the pBlueεcript®SK^" clone deεcribed above with BεtEII and filling in the 5' overhang with Klenow fragment and a εolution of all four deoxyribonucleotides, using the method described in Sambrook (supra, pp. 5.40-5.43). This was followed by BamHI digestion and isolation of the 2.2 Kb fragment by agarose gel electrophoresiε. Thiε fragment waε ligated with pBlueεcript®SK", previouεly digeεted with EcoRV (which leaveε a blunt end) and BamHI. The reεulting plaεmid iε εhown in Figure 15. The DNA εequence of the clone waε determined aε deεcribed in Example 1, above.

c. Engineering the 5' End of the Gene to Create unique Sfil Sites Flanking the Variable Exon and Removing an Mstll Site to Make the Sites Flanking the Constant Exon unique.

Two oligonucleotide primers were syntheεized on a Millipore DNA εyntheεizer (Bedford, MA) , following manufacturerε instructions, for mutagenesis of the 5' end of the IM9 kappa gene: primer B239 (SEQ. I.D. NO. 31) TAGTGGATCCAACTGATTTCTCCAT upstream for the BamHI site at the

5' end of the kappa gene and primer B240 (SEQ. I.D. NO. 32) TTATTTACTTCTGGGTCACCAGGTTTATTC downstream for the BstEII site in the major intron. The downεtream primer recreates the BstEII site that had been altered in the previous step for insertion into pBluescript®SK-.

Two Sfil siteε were deεigned to flank the variable region exon, each having a unique εticky end so as not to re- ligate to each other in cloning but to allow for forced orientation cloning of synthetic variable region cassettes for CDR grafted antibodies. Each S_fil site, the upstream Sfil site and the downstream Sfil site, involved the design of a pair of oligonucleotide primers and a round of overlap mutageneεis (see Figure 16) as described in D.H. Jones and B.H. Howard, "A rapid method for recombination and site- εpecific mutageneεiε by placing homologouε ends on DNA using polymerase chain reaction", Biotechniαues, 1^:62-66 (1991). Two PCR reactions were performed, each using the variable exon clone as the template. The first used the 5¹ flanking primer B239 as the 5' primer and the upεtream S_fil primer B435 (SEQ. I.D. NO. 33)

AAGAGGCCGAGCTGGCCCTTCCCTGAATAACCAGGCAGT aε the 3 ' primer. The second used the 3' flanking primer B240 as the 3' primer and the upstream Sfil primer B434 (SEQ. I.D. NO. 34) GGGAAGGGCCAGCTCGGCGTGTTCCTATAATATGATCAA as the 5' primer. The products of these reactions were purified and used together as templates in an overlap PCR reaction with primers B239 and B240 as shown in Figure 16. The product of the overlap reaction was the full BamHI to BstEII fragment and contained an Sfi site in the appropriate upstream location. Thiε product waε uεed as the template in a new pair of reactions to install the downstream Sfi site in a similar manner, uεing primerε B379 (SEQ. I.D. NO. 35) TTCCTGGCCCTGCAGGCCCAGTTGTCTGTGTCTTCTGTT and B380 (SEQ. I.D. NO. 36) AACTGGGCCTGCAGGGCCAGGAAGCAAAGTT-TAAATTCTA . The PCR waε performed according to the inεtructionε in the GeneAmp® PCR kit (commercially available from Perkin Elmer-Cetuε, Norwalk, CT) on a Thermal Cycler® (commercially available from Perkin Elmer Cetuε) . The reaction waε performed for 30 cycleε of one minute at 94°C, one minute at 55°C, and two minutes at 72°C in a buffer that contained a 1.5 mM final concentration of MgCl2-

The product of the PCR reaction was cloned into pCR™II vector uεing a TA Cloning™ Kit (both commercially available from Invitrogen) in substantial accordance with the manufacturer's protocol. The identity of the clone waε verified by reεtriction mapping to be the IM9 kappa BamHI to BstEII fragment with two engineered Sfil siteε of the appropriate εize and location.

The Mεtll εite upεtream of the kappa promoter, εhown in Figure 16, waε deεtroyed by linearizing the clone deεcribed above with Mεtll and filling in the 5' overhang to make a blunt end. Re-ligation of the modified endε yielded a sequence that no longer contained an MS_tII site.

The clone described above was characterized by DNA sequencing analysiε aε deεcribed in Example 1.

The engineered BamHI to BεtEII fragment was isolated from pCR™II by PCR using two primerε, B495 and B496 (SEQ. I.D. No. 37 CATGTCTGGATCCAACTGATTT and SEQ. I.D. No. 38 CTGATTTACTTCTGGGTGACCAGGTTTATTCAA reεpectively) .

d. Ligation of the Kappa Gene Fragments with the pSV2gpt (Enhancer minus).

The mutated BamHI to BstEII fragment from the Sfil mutagenesis, described in Example 8.c. still contained the native IM9 kappa variable region sequence. It was then ligated with the BstEII to Clal fragment taken from the pBluescript®SK- clone and the pSV2gpt (enhancer minuε) Clal to BamHI fragment (Beidler, e_£ _&1, εupra) .

The reεulting clone waε analyzed by reεtriction enzyme mapping, Southern blot analyεiε, and DNA εequence analyεiε. The confirmed sequence is provided as a restriction map in Figure 17.

e . Insertion of an hZCE Kappa Variable Exon into the pGlM9kappa Vector Using the Engineered Sfil Sites.

The hZCE kappa variable region was taken from a pCRlOOO™ clone using PCR mutagenesis according to the manufacturer's instructions to add the Sfil siteε at the 5' and 3' endε. The oligonucleotide B510 (SEQ. I.D. NO. 39) 5'-AAGGGCCAGCTCGGCCT- CTTCCTATAATATGATCAATAGTATAAATATTTGTGTTTCTATTTCCAATCTCAGGTGCCA AATGTGACATCCAGATGACCCA-3 ' waε uεed aε the 5 ' end primer and B511 (SEQ. I.D. NO. 40) 5'- TGGGCCTGCAGGGCCAGGAAGCAAAGTTTAAATTCTAC- TCACGTTTGATTTCCACCTTGGTT-3 ' aε the 3' end primer. The resulting PCR fragment was digested with Sfil. The plasmid pGIM9kappa, deposited with the ATCC with accesεion number 75512, waε alεo digeεted with Sfil reεulting in three fragments. The hZCE kappa variable region containing fragment deεcribed above waε ligated with the largeεt two of the three fragmentε reεulting from the pGIM9kappa digestion. This was carried out as a three fragment ligation reaction. The three Sfil εiteε have different overhanging εequenceε due to the nature of the Sfil recognition εequence and εo oriented cloning of the three fragmentε into pGIM9kappa was achieved. The reεulting clone pGIM9k/hZCE-kappa waε verified by DNA εequence analyεiε aε having the correct Variable exon sequence.

E mple 9_

Construction &&£ Subcloning of hZCE-CSVL gene.

a. Construction of hZCE-CSVL gene.

The amino acid sequence derived above for the hZCE CDR-grafted CDR switched variable light region was converted into DNA sequence using software from DNA STAR (Madison. WI) . Six oligonucleotideε with overlapping endε and spanning the sequence of the hZCE-CSVL gene were syntheεized on a

Millipore DNA synthesizer (Bedford, MA) . The sequences of the six oligonucleotides compriεing the template are provided aε

SEQ. I.D. Nos. 41-46:

B695 = 5' -GGG-AAG-GGC-CAG-CTC-GGC-CTC-TTC-CTA-TAA-TAT-GAT- CAA-TAG-TAT-AAA-TAT-TTG-TGT-TTC-TAT-TTC-CAA-TCT-CAG-GTG-CCA- AAT-GTG-ACA-TCC-AGA-TGA-CCC-AGT-TTC-CT- 3 (SEQ. I.D. NO. 41) B696 = 5' -GCA-TGC-CGA-AGT-TGG-AGA-AGG-TGA-AGC-CGG-AGG-CGC- GGC-AGG-TGA-TGT-TCA-CGC-GGT-CGC-CCA-CGG-AGG-CGG-ACA-GGG-TGG- AAG-GAA-ACT-GGG-TCA-TCT-GGA-TGT- 3 (SEQ. I.D. NO. 42) '

B549 = 5' -GGC-TTC-ACC-TTC-TCC-AAC-TTC-GGC-ATG-CAC-TGG-ATC- CGC-CAG-AAG-CCC-GGC-AAG-GGC-CTG-AAG-TGG-GTG-GCC-TAC-ATC-TCC- GGC-GGC-TCC-TCC-ACC-GTG-CAC-TA- 3 (SEQ. I.D. NO. 43) '

B550 = 5' -GGT-GAT-GGT-CAG-GAA-CAG-CTC-GTT-CTT-GGG-GTT-GTC- GCG-GGA-GAT-GGT-GAA-GCG-GCC-CTT-CAG-GGA-GTC-GGC-GTA-GTG-CAC- GGT-GGA-GGA-GCC-GCC-GGA-GAT-GTA- -3 (SEQ. I.D. NO. 44) '

B697 = 5' -CCC-CAA-GAA-CGA-GCT-GTT-CCT-GAC-CAT-CAC-CTC-CCT- GCA-GCC-CGA-CGA-CTT-CGC-CAT-GTA-CTA-CTG-CGC-CCG-CGA-CTA-CTA- CGT-GAA-CAA-CTA-CTG-GTA-CTT-CGA-CGT-GT (SEQ. I.D. NO. 45)

B698 = 5' -CAC-AGA-CAA-CTG-GGC-CTG-CAG-GGC-CAG-GAA-GCA-AAG- TTT-AAA-TTC-TAC-TCA-CGT-TTTG-ATC-TCC-ACC-TTG-GTG-CCC-TGG-CCC- CAC-ACG-TCG-AAG-TAC-CAG-TAG-TT (SEQ. I. D. No. 46)

The six oligonucleotideε were used in a PCR reaction using Taq polymeraεe and two additional oligonucleotide primerε, B553 (SEQ. I.D. No. 47) 5' -GGG-AAG-GGC-CAG-CTC-GGC-CTC-TT -3' and B554 (SEQ. I.D. No. 48) 5' -CAC-AGA-CAA-CTG-GGC-CTG- CA- 3' for amplification. The oligonucleotide templateε, primerε, PCR reagents and buffers were used at concentrations described by the manufacturer. Twenty five cycles of amplification were carried out, aε followε: (1) Denature at 94 C for one minute, anneal at 55 C for one minute, and extend at 72 C for one minute.

b. Subcloning of hZCE-CSVL gene into TA Vector.

Following PCR εyntheεis of the CDR-grafted variable region containing ZCE-025 heavy chain CDRs, the approximately 500 base pair DNA fragment was ligated into a TA holding vector as per the manufacturer's protocol (In Vitrogen, San Diego) . TA vectors are provided by the manufacturer as linear molecules containing a single deoxythymidylate as an overhang on each of the vector's 3' endε. Thiε iε complementary to the deoxyadenylate overhangε found on the 3 ' ends of PCR products due to the terminal transferaεe activity of Taq polymerase.

TA clones containing inserts of the correct size (about 500 base pairs) were identified by EcoRI restriction digests of DNA minipreps using methodε known in the art. Up to ten cloneε with appropriate insert sizes were sequenced on a Genesis® DNA sequencer (DuPont, Delaware, MD) . A clone with the appropriate sequence waε digeεted to completion with Sfil restriction endonuclease. This restriction site was present at the 5 ' and 3 ' ends of the hZCE-CSV_L gene for cloning into the final expresεion vector as described in Example 10, below. The hZCE-CSVL fragment was isolated following electrophoresis using the gel purification method described above. After ethanol precipitation, the fragment waε resuεpended in εterile distilled H2O and the concentration was determined by running a εmall aliquot on a gel, aε deεcribed previouεly.

Example 10

&a_TEa. and_. Expression of hZCE(CSVL) -kappabody

a. Construction of pGIM9k/hZCE(CSV_L) -kappa.

The 484 bp DNA Sfil to Sfil fragment containing the hZCE-CSV_L region was combined with a 9 kb Sfil to Sfil fragment isolated from the pGIM9 kappa expresεion vector by standard ligation (Sambrook, et al.). As shown in Figure 18, the resulting expression vector, pGIM9k/hZCE(CSV_L)-kappa contained the following components:

(1) Human IM-9 kappa promoter, signal exon 1 and signal intron (up to added Sfil site) . (2) The hZCE(CSV_L) gene beginning with an Sfil εite in the εignal intron and including the pGIM9 kappa εignal exon II hZCE(CSV_L) region and extending to an Sfil εite at beginning of the major intron. (3) Human IM-9 kappa major intron (from Sfil εite) , kappa conεtant exon and 3' flanking εequenceε (containing native polyadenylation εite) .

(4) XGPRT gene under the control of an enhancerleεs SV40 early promoter.

(5) Bacterial plaεmid origin of replication, derived from pBR 322.

(6) Bacterial β-lactamaεe, driven off its native promoter.

b. Transfection of SP 2/0 and hZCEK.

Vector pGIM9k/hZCE(CSV_L) -kappa, on deposit with ATCC under the proviεionε of the Budapest Treaty Deposit No. 75530, was electroporated into two different host cell lines, SP 2/0 and hZCEk. hZCEk is a transfectoma derived from SP 2/0 by transfection with the vector pGIM9k/hZCE- kappa, which expreεεeε CDR grafted ZCE/IM-9 light chain (hZCEK-homodimer) [Example 8.e.] . For SP 2/0, pGIM9k/hZCE(CSV_L) -kappa waε electroporated together with the drug εelectable gene neo in the vector pSV2Neo, and transfectants were selected by growth in HH4 medium containing 1.5 mg/ml geneticin (Betheεda Reεearch Labε/Gibco, Gaitherεberg, MD) . For hZCEk, pGIM9k/hZCE (CSV_L) -kappa waε alεo co-electroporated with pSV2neo to allow selection of transfectants in medium with geneticin 1.5 mg/ml.

Electroporation conditions and selection media recipes were aε described by Chu, et al . (Nucleic Acids Research. 15:1311- 1325 (1987)) . Briefly, the SP2/0 cells were grown in media containing 10% FBS and were maintained in log phase growth for the three days preceding electroporation. Fifty micrograms of the plasmid vector waε linearized uεing the reεtriction enzyme Pvul (1 unit/μg) and the Reaction Buffer #7 from GIBCO-BRL (Gaithersburg, MD) . At the time of tranεfection the SP2/0 cellε were collected by centrifugation in an IEC clinical centrifuge (800 rpm, 10 min, room temperature) . Cellε were waεhed in Hanks Buffered Saline Solution from Gibco Laboratories (Grand Island, NY) containing an additional 6 mM dextrose and resuεpended at a final concentration of 1.0 x 10⁷ cellε/ml. 0.5 ml of cells were aliquoted into cuvetteε and the linearized DNA waε added. Electroporation was done using the Cell-Porator® (GIBCO-BRL) with settings of 300 μF and 350 volts.

c. Selection and characterization of hZCE-kb expressing clones.

Resistant clones of each host cell line were identified by growth on appropriate selective media and aεsayed for hZCE(CSV_L) chain production (SP 2/0 host) and CEA binding (hZCEk host) activity as described in Example 15, shown below. The resultant clones were called hZCEhb (SP 2/0 host) and hZCEkb (hZCEk host). hZCEhb produces only the human kappa light chain with ZCE heavy chain CDRs secreted as a homodimer, while hZCEkb produceε a human light chain dimer with one kappa chain containing ZCE heavy chain CDR's and the other containing ZCE light chain CDRs. A conventional human kappa ELISA can be used to quantitate production levels of the homodimer from hZCEhb, but a CEA-binding ELISA is required to quantitate the antigen binding heterodimer hZCEkb. The hZCEkb chain or hZCEhb chain were secreted aε dimers. The hZCEhb homodimer did not bind CEA, while the hZCEkb had affinity for CEA. Example 11

Construction of hZCE(CSV_j Expression Vector and Expression of hZCEf CSV_h)

a. Preparation of gene for chelating peptide.

A CDR εwitched variable region iεolate waε constructed as a variation of the hZCE(CSV_L) kappa chain where the human kappa constant region would be deleted εo aε to expreεε the hZCE(CSV_L) light chain domain only. To εcreen for the CDR εwitched iεolate construct, it was desirable to expresε it aε a fuεion protein containing a metal chelating peptide for purification. The gene encoding the chelating peptide waε prepared by creating a DNA fragment which would ultimately replace the human kappa conεtant exon in the pGIM9k/hZCE(CSV ) -kappa vector. Uεing PCR techniqueε, an approximately 330 baεe pair Mstll/Mstll modified fragment ⁽Fragment A) was prepared using the pGIM9k/hZCE(CSV ) -kappa expresεion vector as template. The upstream primer in this PCR reaction was B1000 (SEQ. I.D. No. 49) 5'-CAC-CAT CCT GTT TGC TTC TTT CCT CAG GAA CTG TGC ACT GGC ACC ACC ACC CAT AGA GGG AGA AGT GCC CCC ACC TGC TCC TCA GTT -3 ' , which included the codons for a 6-amino acid chelating peptide, and the downstream primer waε B441 (SEQ. I.D. No. 50) 5'- GGGTAAAAATAGAATGAAGGATGAT-TTTTATAAAT-3 ' . Fragment A conεiεted 5' to 3 ' of (1) an MSTII reεtriction εite and the εplice acceptor εite from the IM9 kappa conεtant region; (2) the codonε for the firεt three amino acidε of the kappa conεtant region; (3) the codonε for a εix amino acid chelating peptide sequence (HWHHHP) and a termination codon; and (4) 3' untranslated sequence including the polyadenylation site and native MSTII-restriction εite.

b. Construction of pGIM9k/hZCE (CSV_L) expression vectior.

Fragment A and pGIM9k/hZCE ( CSV_L ) - kappa were digeεted wi th either Mεt ll or Bεu3 6 - 1 ( Stratagene , 10X Universal buffer, 37 ' C for a minimum of 3 hours) to produce ligatable ends. Fragments (-330 bp of Fragment A and -12.8 kb pGIM9k/hZCE(CSV_L) -kappa were thus isolated and purified using Milligen's Ultrafree-MC (Yonezawa, Japan) method. Ligation waε carried out uεing componentε and ligation conditionε from a TA Cloning Kit (Invitrogen, San Diego, CA) following the manufacturer' ε protocol. Electroporation into Electromax DH10B cellε (BRL, Gaitherεburg, MD) was performed. Transformed cells were plated onto agar, incubated overnight, and colonies were grown-up for plasmid mini prepε uεing Qiagen'ε (Chatsworth, CA) "Mini Plasmid" protocol. Construct size waε verified by reεtriction digeεt analyεiε uεing EcoRI, Mεtll or Bsu36-1, Sεtl, and BamHI enzymeε.

Large εcale plaεmid preparations were performed uεing Qiagen'ε "Maxi-plaεmid" prep procedure. DNA εequencing waε performed to verify the correct εequence, which iε called hZCE(CSV_L) (SEQ I.D. No. 51). The cloned plaεmid herein is called pGIM9k/hZCE(CSV_L) .

SEQ. I.D. No. 51

GAC ATC CAG ATG ACC CAG TTT CCT TCC ACC CTG TCC GCC TCC GTG

GGC GAC CGC GTG AAC ATC ACC TGC CGC GCC TCC GGC TCC ACC TTC

TCC AAC TTC GGC ATG CAC TGG ATC CGC CAG AAG CCC GGC AAG GGC

CTG AAG TGG GTG GCC TAC ATC TCC GGC GGC TCC TCC ACC GTG CAC TAC GCC AAC TCC CTG AAG GGC CGC TTC ACC ATC TCC CGC GAC AAC

CCC AAG AAC GAG CTG TTC CTG ACC ATC ACC TCC CTG CAG CCC GAC

GAC TTC GCC ATG TAC TAC TGC GCC CGC GAC TAC TAC GTG AAC AAC

TAC TGG TAC TTC GAC GTG TGG GGC CAA GGG ACC AAG GTG GAA ATC AAA

c. Expression of hZCE(CSV^) .

Linearization of pGIM9k/hZCE (CSV_L ) DNA was performed via Clal digestion. Electroporation into SP2/0 cells was performed aε previouεly deεcribed in Example 10.b. Cellε were εeeded in HH4 medium εupplemented with 10% FCS.

Three dayε later, cells were plated § 2 x 10 5/ml in 24- well cluεter plateε in the preεence of HH4, 10% FCS, MAX (MAX = 1.0 με/ml mycophenolic acid pluε 100 με/ml xanthme) . At day 14 after plating, colonieε were harveεted and tranεferred to 6-well plateε (Falcon) for expansion and εerum-free medium adaptation. Clones were succeεεfully expanded and adapted to εerum-free conditionε within 2 weekε.

Example 2

Construction of pNIM9k/hZCE-gamma

a. Construction of hZCE heavy chain variable exon.

The protein εequence of the heavy chain of hZCE waε converted to nucleic acid εequence in the following manner: (1) if the ammo acid waε derived from ZCE, the actual ZCE codon at the εite was used; (2) if the ammo acid was derived from IM9, the actual IM9 codon at the site was used; (3) if the ammo acid was derived from a conεenεuε εequence, any appropriate codon was used. The hZCE gamma variable exon (SEQ. I.D. NO. 58) εhown below waε obtained by PCR reactionε.

SEQ I.D. No. 58

GAA ATG CAA CTG GTG GAA TCT GGG GGA GGC CTG CTA CAG CCT GGC CGG GCC CTG CGG CTC TCC TGT GCA GCC TCT GGA TTC ACT TTT AGT AAC TTT GGA ATG CAC TGG ATT CGG CAA ACT CCA GGG AAG GGC CTG GAG TGG GTC GCA TAC ATT AGT GGT GGC AGT AGT ACC GTC CAC TAT GCA GAC TCC TTG AAG GGC CGA TTC ACC ATC TCC CGG GAC AAC GCC AAG AAC TCC CTC TAT TTG CAA ATG ACC AGT CTC CGG GCT GAG GAC ACG GCC TTG TAT TAC TGT GCA CGG GAT TAC TAC GTT AAT AAC TAC TGG TAC TTC GAT GTC TGG GGC CAA GGG ACA ATG GTC ATC GTC TCT TCA G

Five overlapping oligonucleotideε, B156. B159, B396. B397. and B398 (SEQ. I.D. NO. 60-64) SEQ I . D . No . 60

5 ' -GAT CCG AAA TGC AAC TGG TGG AAT CTG GGG GAG GCC TGC TAC AGC CTG GCC GGG CCC TGC GGC TCT CCT GTG CAG CCT CTG GAT TCA CCT TTA G-3 '

SEQ I.D. No. 61

5 ' -CAC CAC TAA TGT ATG CGA CCC ACT CCA GGC CCT TCC CTG GAG TTT GCC GAA TCC AGT GCA TTC CAA AGT TAC TAA AGG TGA ATC CAG AGG C-3 '

SEQ I.D. No. 62

5 ' -GGG TCG CAT ACA TTA GTG GTG GCA GTA GTA CCG TCC ACT ATG CAG ACT CCT TGA AGG GCC GAT TCA CCA TCT CCC GGG ACA ACG CCA AGA A 3 '

SEQ I.D. No. 63

5' -TAT TAC TGT GCA CGG GAT TAC TAC GTT AAT AAC TAC TGG TAC TTC GAT GTC TGG GGC CCA GGG ACA ATG GTC ATC GTC TCT TCA -3 '

SEQ I.D. No. 64 were εyntheεized on a DNA Synthesizer (Millipore) following manufacturer's inεtructionε. They were fused together by a PCR reaction using B161 (SEQ. I.D. NO. 65) 5' -AAG- GAT CCG AAA TGC AAC TGG TGG AAT CT -3' and B162 (SEQ. I.D. NO. 66) GAC GAA TTC TGA AGA GAC GAT GAC CAT TG as the end primers. The resulting fused fragment was cloned into pCR™II (Invitrogen) and the sequence was verified as described in Step 2.1.j .

b. Construction of the hZCE gamma expression vector, pNIM9k/hZCE-gamma (cDNA). The hZCE heavy variable exon and the entire IM9 gamma constant region (from 5' IM9 heavy CHI exon to the BstEII site 3 ' of the CH3 exon) were fused together by an overlap PCR reaction. Two PCR reactionε were performed: the first PCR reaction uεed the pCRII clone from 3.a. as template and primerε B611 and B612. The PCR product was reamplified with primers B467 and B567. The second PCR reaction used primerε B566 and B514. The productε of these two reactions were uεed together aε templateε in the overlap reaction with primerε B467 and B514. The resulting fusion fragment of hZCE heavy variable exon and 5 ' IM9 heavy CHI exon to BstEII was cloned into pCR™II (Invitrogen) and the sequence was verified aε described in Step 3.1. The resulting vector is phZCE/CHIBstEII.

A pair of oligonucleotideε, B743 and B744, were deεigned to add the εplice recognition εite and the Sfil site 3' of the variable region. The IM9 heavy chain cDNA vector waε digeεted with BamHI and Hindlll, extracted with phenol and chloroform mixture, precipitated with EtOH, and reεuεpended in TE. Primerε B743 and B744 were kinεed, annealed together, and ligated with the digested vector. The ligation reaction was uεed to tranεform E. coli DH10B by electroporation. The colonieε were picked for analyεiε by reεtriction enzyme mapping and the resulting vector is pIM9gammacDNASfii.

The phZCE/CHlBεtEII vector and pIM9gammacDNASfil were digested with Sfil and BstEII. The 740 bp fragment from phZCE/CHIBstEII and the 950 bp fragment were purified by agaroεe gel electrophoreεiε. The pGIM9kappa vector waε digeεted with Sfil and the 12 Kb fragment waε purified by agaroεe gel electrophoreεiε. The three purified fragmentε were ligated and uεed to transform E. coli DH10B by electroporation. The colonies were picked for analyεiε by reεtriction enzyme mapping. The reεulting vector iε pGIM9k/hZCE-gamma .

The Neomycin resistance gene was inεerted into pGIM9kappa vector to make pNIM9kappa. Both the pGIM9kF2 and the pSV2neo vectorε were digested by Apal and Pvul, the 5 Kb Neomycin resistance gene-containing fragment from the pSV2neo digest and the 9 Kb fragment from the pGIM9k digest were purified by gel electrophoreεiε. The two fragmentε were ligated and used to transform E. coli DH10B by electroporation. The colonies were analyzed by restriction enzyme mapping, the resulting plasmid is pNIM9kappa. Both the pNIM kappa and the pG(IM9k) /hZCE-gamma vectors were digested with Sfil, the 9 Kb and 5 Kb fragments from pNIM9kappa and the 1.6 Kb fragment from pG (IM9k) /hZCEgamma were purified by agarose gel electrophoresiε. The three purified fragmentε were ligated and used to transform E. coli DH10B. The colonies were picked and analyzed by restriction enzyme mapping, the resulting plasmid is pN(IM9k) /hZCE-gamma(cDNA) .

Exam le 11

Construction of hZCE ( CSV_j^) -gamma Expression Vector &S£ expression of hZCE < CSV_L) -intact kappabpfly

In another variation using the variable kappa region from IM-9 containing ZCE heavy chain CDRs (hZCE(CSV_L) region) , a human gamma heavy chain was constructed.

a. Construction of pGIM9k/hZCE(CSVj -gamma

Using the polymerase chain reaction (PCR) , a 2.1 kilobase DNA fragment was amplified using primers B922 (SEQ. I.D. No. 52) 5 ' -AAG-AGC-TCC-TGA-ACC-TCG-CGG-ACA-GTT-AA-3 ' ) and B923 (5 ' -AAA-TCG-ATC-TCA-GGC-CTC-AGA-CTC-GGC-CTG-ACC-CGT- GGA-AA-3 ' ) (SEQ. I.D. No. 53) from a fragment of pNIM9k/hZCE-gammaι• The 5' end of thiε fragment contained an Sst-1 restriction site and the 3' end contained a Cla-1 site. A second Cla-1 (5') to Sst-1 (3') fragment of 8.5 kilobases containing the neomycin gene, β lactamase gene and the hZCE- CSVL variable region gene was ligated together with the PCR generated 2.1 kilobase fragment. The 10.6 kilobase plasmid resulting from this ligation was reopened with Sst-1 restriction endonuclease and ligated together with a 2.2 kilobaεe Sst-1 fragment from pGIM9kappa containing a portion of the human kappa major intron with enhancer. The final expreεεion vector is 12.8 kilobaseε and called pNIM9k/hZCE(CSV_L) -gamma.

b. Expression of hZCE(CSVL) -intact kappabody.

The pNIM9k/hZCE(CSV_L) -gamma expreεεion vector waε electroporated (aε deεcribed in Example 10.b.) into cellε expreεεing the pGIM9k/hZCE-kappa gene. Three dayε following electroporation the cells were put under drug selection (geneticin 1.5 mg/ml) and colonieε which grew up under this εelection were analyzed for εecretion of hZCE(CSV_L) -intact antibody.

Protocol for Subcloning Transfectomas

Individual wellε from the initial εcreening for cells εecreting the highest levels of immunoglobulin were further subcloned to inεure a εingle clone had been εelected. Briefly, the cellε were diluted to 10, 5 or 0.3 cellε per 200μl and plated into two 96-well tiεεue culture plates at each dilution. The medium is HH4 with 10% fetal calf serum, 100 ug/ml xanthine and the appropriate selection drug. After fourteen days individual wells were viεually εcreened for εingle colonieε, then harveεted and cultured further so aε to obtain a quantitative ELISA value aε deεcribed in Example 15, below.

Example I

Cloning and Expression of a, Single Chain Fv Containing a CSV_L Fragment

a. Construction of pGIM9k/hZCE (CSV_L) -ScFv expression vector.

To conεtruct an expreεεion vector for a CSV-^ , the earlier expreεsion vector pGIM9k/hZCE-kappa was reconεtructed to contain the cdr-grafted kappa variable region in place of the human kappa conεtant region. In addition, the vector contained a 5' extension to the kappa variable region to - Ill -

serve as a linker (L) between variable regions as well as a 3' chelating peptide (CP) sequence. Therefore, diagramatically, the linear construct is as follows:

The kappa variable region with 5' linker and 3' chelating peptide was synthesized in three separate PCR reactions. The first DNA fragment (Fragment-1) of 457 baεe pairε waε amplified from vector pGIM9k/hZCE-kappa (Example 10.a.) with a 5' primer (ClOl) (SEQ I.D. No. 54) 5 ' -CTG TTT GCT TCT TTC CTC AGG AGG CGG TTC AGG AGG ATC AGG CGG TTC AGG TGG ATC AGG AGG CGA CAT CCA GAT GAC CCA GTC TCC T-3 ' containing the Mεtll reεtriction site, the linker (G-G- S)4GG and the firεt 24 baεeε of the conεtant kappa gene

(Example lO.a.); and a 3' primer C102 [SEQ I.D. No. 55] 5 ' -GTC AGG CTG GAA CTG AGG AGC AGG TGG GGG CAC TTC TCC CTC TAT GGG TGA TGG TGC CAA TGT TTG ATT TCC ACC TTG GTC CCT TGG CCG -AA-3 ' containing the baεeε of the 3' end of the kappa conεtant region, a H-W-H-H-H-P chelating peptide and stop codon. The second DNA fragment (Fragment-2) of 335 base pairε waε generated uεing 5' primer C103 [SEQ I.D. No. 56] 5 ' -GAG AAG TGC CCC CAC CTG CTC CTC AGT TCC AGC CTG ACC CCC TCC CAT CCT -3' and 3' primer B441 [SEQ I.D. No. 50] and the same template as above, i.e. pGGhZCE-HB. Thiε Fragment contained the 3 ' human kappa constant region containing the polyadenylation εignal. The final DNA fragment (Fragment-3) waε amplified uεing Fragment-1 and Fragment-2 as template and 5' primer ClOl [SEQ I.D. No. 54] and 3' primer B441 [SEQ I.D. No. 50] to give the approximately 800 base pair Fragment-3. This Fragment-3 was cloned into a TA vector for confirmation of DNA sequence as described in Example 9.b. Following confirmation of sequence the Fragment-3 insert was re- isolated from the TA vector as an Mstll fragment and cloned into the vector pGIM9k/hZCE-hb (which had its Mstll fragment, containing the human kappa conεtant region, deleted) . All PCR amplificationε were carried out aε deεcribed in Example 9.a. b. Expression of hZCE(CSV_L) -ScFv.

After cloning and εcale up, the final expression vector, herein called pGhZCE-CSV_L-sFV, was electroporated into SP2/0 hybridoma cells as described in Example 10.b. Clones secreting the CSVL~SFV construct were identified as described in Example 15.f., below. Finally, the affinity of the construct was analyzed via a competitive inhibition assay as described in Example 15.e, below.

Example 15

Identification. guantitation and affinity determinationof engineered constructs produced

a. Identification and quantitation of secreted hZCE-CSVL-kappa homodimer and hZE(CSVL) /hZCE- kappa heterodimers

Identification and quantitation of secreted CDR grafted human kappa chains from transfected SP 2/0 cells expresεing hZCE kappa homodimer, and thoεe expreεεing hZCE- CSV_L homodimer were identified by a standard enzyme-linked immunoεorbent asεay ("ELISA", as deεcribed by Engvall, E. and Perlmann, P., Im unochemiεtrv, 8:871-874 (1971)) for human kappa. The purpoεe of thiε assay was to identify those cells secreting the highest levels of kappa chain polypeptide coded for by pGIM9k/hZCE-kappa or pGIM9k/hZCE(CSV_L) -kappa plasmid vector. A 5μg/ml solution of goat anti-human kappa chain (Tago #4106, Tago Inc., Burlingame, CA) in lO M sodium phosphate pH 7.4 was prepared. Each well of a 96 well plate was coated with 50μl of this solution. The plates were then incubated overnight at 37^*c Plates were then rinsed thoroughly in H2O, and then PBS with 1.0% Tween-20™ (w/v) .

Fifty μl of the supernatant fractions were added to each well, and incubated for two hours at room temperature.

Plateε were again rinsed as detailed above. A goat anti- human kappa chain alkaline phoεphataεe conjugate (Tago #2496, Tago, Inc.) was diluted 1:1000 in the same medium as the supernatant material. lOOμl were added per well and allowed to incubate for one hour at room temperature. Plates were rinsed as above. The alkaline phosphatase subεtrate (Hybritech, Inc., San Diego, CA Part #100103) was prepared aε per package inεtruction, one tablet per 3ml of distilled H2O and 150μl of this subεtrate was added to each well and allowed to incubate 30 minutes at 37'C. The reaction was quenched with 50μl of 300 mM EDTA and then the absorbance was read at 405 nM. Colonies, whoεe εupernatantε εhowed the higheεt levelε of kappa expression, were subcloned and cryo- preεerved. Expreεεion levelε are εhown in Table 3.

b. Identification and quantitation of hZCE(CSV_L)- intact kappabodies.

Detection of aεεembled hZCE(CSV_L) -intact kappabodieε waε carried out by coating the microtiter plate wellε with goat anti-human IgG heavy chain antibody reagent (Tago #3100, Tago, Inc., 887 Mitten Road, Burlingame, CA) at 5 μg/ml in 10 mM phoεphate pH 7 to 8. Plates were dried overnight at 37"C, then washed with PBS and 0.1% Tween-20™, then H2O. Fifty microliters of the cell supernatant were added to each well and incubated for 2 hours at room temperature. Plates were again rinsed as detailed above. A goat anti-human kappa chain alkaline phosphataεe conjugate

(Tago #2496 Tago, Inc., 887 Mitten Road, Burlingame, CA) waε diluted 1:1000 in the same medium as the supernatant material. 100 μl were added per well and allowed to incubate for 1 hour at room temperature. Plates were rinsed aε above. The alkaline phosphatase substrate (Hybritech) , one tablet per 3 ml of distilled H2O, and 150 μl of this substrate was added to each well and allowed to incubate 30 minutes at 37'C. Purified protein, IgG -kappa, from the human lymphoblastoid cell line IM9 waε used as a positive control. c. ELISA for detecting the presence of hZCE(CSV_L) constructs bound to Carcino-Embryonic Antigen

(CEA) .

To detect the hZCE(CSVL> constructε which can bind CEA, ELISAε were performed aε follows:

On the first day CEA stock εtandard (lmg/ml) (Hybritech Part # 211288) waε diluted to 10 μg/ml in PBS with lmg/ml BSA in a final volume of 6 mL for each ELISA plate. 96-well ELISA plateε (Titertek, McLean, VA) were coated at 50μl/well, tapped to enεure that all well bottomε were completely covered, and incubated overnight at 37 "C.

On the εecond day the plateε were waεhed twice with diεtilled, deionized water, twice with lXPBS+0.1% Tween-20™, and twice again with diεtilled, deionized water. Samples containing the hZCE(CSVL) -heterodimer, hZCE(CSV_L) -intact, and εtandardε were added to the plateε at 50μl/well. Plateε were then sealed and incubated at room temperature for 2 hours. Goat Anti- (Human Kappa) conjugated with alkaline phosphataεe (Tago # 2496 Burlingame, CA) waε diluted 1:1000 in RPMI medium (Gibco) with 10% horεe serum and 3% goat serum to a volume of lOml/plate. Plates were washed as before, and the anti-Kappa conjugate was added to the plates at 50μl/well. Then the plates were sealed and incubated at room temperature on a shaker at -100 rpm. for 1 hour. PNPP Alkaline Phosphatase tablets (Hybritech Part

#100103) were disεolved in diεtilled, deionized water at a ratio of 1 tablet per 3ml of water, and 150μl/well of the alkaline phoεphatase solution waε added to the plateε. The plates were incubated at 37 'C for half an hour and then absorbencieε were read at 405nm using a CERES900 ELISA reader (BioTek, Inc., Winoosk, Vermont) . The cultures correεponding to the wellε whose supernatantε yielded the higheεt optical denεitieε were εelected for further εcale up and ELISA quantitation. d. Competitive inhibition ELISA for quantifying CEA binding of hZCE CSV_L heterodimers.

The binding affinity of the hZCE(CSV_L) -heterodimers for carcinoembryonic antigen was quantified as follows: On the first day the substrate antibody was prepared. Briefly, CEV124.1, a murine monoclonal anti-CEA antibody obtained from Hybritech (San Diego, CA) was diluted 1:1000 in phosphate buffered saline (PBS) to a final volume of 6 mL. The PBS was prepared by mixing 1494 g NaCl, 36 g KCl, 36 g KH2PO4, and QS to 18L H₂0, then diluted 1:10 with diεtilled, deionized water. A 96 well plate waε coated with the antibody-containing εolution uεing about 50μl/well. The plate waε tapped to enεure that each entire wall bottom was covered. The plate was sealed and left at room temperature overnight. The next day the CEA antigen waε prepared aε deεcribed in Example 15.C.

The plates containing bound antibody were washed four timeε with diεtilled, deionized water, and 50μl of the CEA/BΞA antigen-containing solution was dispenεed into each well. The plateε were εealed and placed on a rotator εhaking at -300 rpm. for 2 hr. Finally the plates were washed as before.

A supernatant of hZCE(CSV_L) -heterodimer was loaded at 50μl/well. A standard curve was generated by diluting a lOμg/ml solution of XCEM F(ab)' or ZCE Fab' at 1:2 increments along the top row of the assay plate. The XCEM chimeric antibody was described in Beidler, C.B., et al . , "Cloning and High Level Expresεion of a Chimeric Antibody with Specificity for Human Carcinoembryonic Antigen, " J. of Immunol.. 141:4053-4060 (1988). Plates were sealed and incubated on a rotator as before for 45 minutes to allow the test antibody to bind to the antigen.

For use as the competition antibody, biotinylated F(ab)' fragmentε of XCEM chimeric monoclonal antibody or ZCE Fab' were prepared. Biotinylation waε conducted aε deεcribed by Enzotin Biochem, Inc., New York, NY. The biotinylated fragmentε were diluted to a final concentration of 0.4μg/ml (experimentally determined to give an OD 490 of about 0.6! and 50μL of the εolution waε added to each well without waεhing the plate. The plate waε εealed and incubated aε before for 45 minuteε and then waεhed aε before. Streptavidin/horse radish peroxidase (Fisher Biotech,

Pittsburgh, PA) conjugate was prepared as per manufacturer's directions and then diluted in IX PBS with 1% BSA. Fifty μl of the εtreptavidin labeled conjugate were added to each well, the plate waε εealed, and incubated aε before for 45 minuteε. Finally the plate waε waεhed aε before to remove the unbound conjugate.

Final εubεtrate waε made by completely dissolving one lOmg tablet of o-phenylenediamine dihydrochloride (Sigma #P8287, St. Louis, MO.) into 10 ml of PCB (18.45g Citric Acid (monohydrate), 25.86g Na2HPθ4, bring to 1.81 @ MilliQ

H2O, pH to 5.0, QS to 2L) , then adding 15μl of 30%H O2. lOOμl of εubεtrate was added to each well. When a standard curve could be visualized, the plates were quenched by adding 50μl of 4M H2SO4 to each well. The asεay waε read on the Biotek CERES900 assay reader at absorbance of 490 nM.

Concentration calculations were done using the built-in software "Kineticalc Jr." from BioTek Instruments (Winoosk, Vermont) . Reεultε of theεe experimentε are shown in Table 3 below.

e . Competitive inhibition assay for determination of affinity of anti-CEA antibodies and constructs.

Affinities of unlabeled recombinant antibodies were determined by a modification of the method described by H. Motulsky and L. Mahan, Molecular Pharmacology, .2_5_:l-9, 1983). Thiε method can measure the affinity of unlabeled antibodies by evaluating their ability to inhibit the binding of a labeled tracer antibody which reactε with the εame epitope of an antigen. Tandem® R CEA Beadε (Hybritech #600211), which contain the mouse an iCEA antibody CEV124, were put into 13cm x 75cm polystyrene tubeε (1 bead per tube) and incubated with lOOmg of CEA, diluted in 1% BSA/PBS solution to a final volume of lOOul, for 2-5 hours at room temperature. The source of the CEA used for these experimentε is CEA Stock Standard Solution (Hybritech, #200288) . The beadε were then waεhed twice with 2ml of 0.1% Tween20™ in phosphate buffered saline juεt prior to adding the antibodies for affinity testing.

The tracer antibody is a isothiobenzyl-DTPA conjugate of ZCE025 Fab' fragment labeled with 3uCi of ^¹In Citrate per microgra of Fab'. The tracer iε first titrated for binding to the above CEA beads to determine a 40-60% saturation point. This concentration of tracer (usually 1.5 x 10-9 M) is used for all the following inhibition reactionε. Varying concentrationε of unlabeled XCEM or supernatant containing hZCE-CSV_L heterodimer were added (lOOul) to the

CEA beads at 2X their final concentrations (final is 1 x 10^~7 M down to 1 X 10^"11 M, diluted in 1%BSA/PBS) together with an equal volume of the 2X tracer (100 μl) . The reaction waε then incubated overnight at room temperature on an Orbital Shaker (150-200 RPM) .

f . Identification and quantification of hZCE(CSV_L) isolate.

Cells putatively secreting hZCE(CSV_L)-isolate were seeded at 4 x 10^/ml in serum-free HH4 medium containing 100 μg/ml xanthine and 1.0 μg/ml mycophenolic acid. When cell numbers reached -1 x 10^/ml, 1.0 ml of their supernatants were collected and mixed with 100 μl of Ni+ -loaded nitrilo acetic acid agarose beads (Qiagen, Inc., Chatsworth, CA) . The beadε and conditioned cell supernatant from 24 individual cloneε were incubated for a minimum of four hourε on a rotating wheel at room temperature. The beadε were waεhed 3 timeε with 50 mM sodium phosphate, lOOmM sodium chloride buffer, pH 7.4. Bound protein was eluted from the beads by addition of 100 μl of SDS-PAGE reduced sample buffer. The elutate was electrophoresed on 15 - 20% SDS-PAGE gelε and the gelε were εilver εtained to visualize and quantitate the hZCE-CSVL~iεolate. The SDS-PAGE gelε, bufferε and εilver staining kit were carried out using reagents from Biorad, (Richmond, CA) according to the manufacturer's inεtructionε. Reεultε are εhown in Table 3 below.

TABLE 3

Secreted Protein Affinity M.W.

Construct μ^q/ml* M (Kd)

mZCE Fab EP Z X 109 52

hZCE-kappa homodimer 20 50

hZCE(CSV_L) -kappa homodimer 20 54

hZCE(CSV_L) -kappa heterodimer 20 2 x 10⁹ 52

hZCE(CSV ) intact ND 160

hZCE(CSV :

ND 18

= average

EP = enzymatically produced ND = not determined

The foregoing description of the invention iε exemplary for purpoεeε of illustration and explanation. It should be understood that various modifications can be made without departing from the spirit and scope of the invention. Accordingly, the following claimε are intended to be interpreted to embrace all εuch modificationε. GGSGGSGGSGGSGG (14 RESIDUES) (SEQ I.D. No. 1)

HWHHHP (6 RESIDUES) (SEQ I.D. No. 2)

5'GAC TAG CGG CCG CAT CGA TCC CCC CCC CCC CCC C (SEQ. I.D. No. 3)

5 'CAG ACG TCG ACG ATG GAT ACA GTT GGT GCA GCA TC (SEQ. I.D. No. 4)

SEQ. I.D. NO. 5

ZCE-025 Light Chain Variable cDNA

GAC ATT GTG ATG ACC CAG TCT CAA AAA TTT ATG TCC ACA TCA GTT

GGA GAC AGG GTC AAC ATC ACC TGC AAG GCC AGT CAG AAT GTT CGT ACT GCT GTA GCC TGG TAT CAA CAG AAA CCA GGG CAG TCT CCT AAA

GCA CTG ATT TAC TTG GCA TCC AAC CGG TAC ACT GGA GTC CCT GAT

CGC TTC ACA GGC ATT GGA TCT GGG ACA GAT TTC ACG CTC ATC ATT

AGC AAT GTG CAA TCT GAA GAC CTG GCA GAT TAT TTC TGT CTG CAA

CAT TGG AAT TAT CCT CTC ACG TTC GGT GCT GGG ACC AAG CTG GAG CTG AAA C 381

SEQ. I.D. NO. 6

DIVMTQSQKFMSTSVGDRVNITCKASQNVRTAVAWYQQKPGQSPKALIYLASNRYTGVPDR FTGIGΞGTDFTLIISNVQSEDLADYFCLQHWNYPLTFGAGTKLELK

5 'CAG ACG TCG ACG TTC CAG GTC ACT GTC ACT GGC TC (SEQ. I.D. NO. 7) SEQ . I . D . NO . 8

ZCE-025 Heavy Chain Variable cDNA Sequence:

GAT GTG CAG CTG GTG GAG TCT GGG GGA GGC TTA GTG CCG CCT GGA

GGG TCC CGG AAA CTC TCC TGT GCA GCC TCT GGA TTC ACT TTC AGT

AAC TTT GGA ATG CAC TGG ATT CGT CAG GCT CCA GAG AAG GGA CTG

GAG TGG GTC GCA TAC ATT AGT GGT GGC AGT AGT ACC GTC CAC TAT

GCA GAC TCC TTG AAG GGC CGA TTC ACC ATC TCC AGA GAC AAT CCC AAG AAC ACC CTG TTC CTA CAA ATG ACC AGT CTA AGG TCT GAA GAC

ACG GCC ATG TAT TAC TGT GCA AGA GAT TAC TAC GTT AAT AAC TAC

TGG TAC TTC GAT GTC TGG GGC GCA GGG ACC ACG GTC ACC GTC TCC TCA G 420

SEQ. I.D. NO. 9

DVQLVESGGGLVPPGGSRKLSCAASGFTFSNFGMHWIRQAPEKGLEWVAYISGGSSTVHYA DΞLKGRFTISRDNPKNTLFLQMTSLRSEDTAMYYCARDYYVNNYWYFDVWGAGTTVTVSS

SEQ. ID NO 10

GAC ATC CAG ATG ACC CAG TTT CCT TCC ACC CTG TCT GCT TCT GTA GGA GAC AGA GTC ACC 60

ATC ACT TGT CGG GCC AGT CAG AGT ATT AGT GCC TGG TTG GCC TGG

TAT CAG CAG AAA CCA 120

GGG AAA GCC CCT AAA CTC CTG ATC TAT AAG GCG TCT AGT TTA GAA

AGT GGG GTC CCA TCA 180 AGG TTC AGC GGC AGT GGA TCT GGG ACA GAG TTC ACT CTC ACC ATC

ACC AGC CTG CAG CCT 240

GAT GAT TTT GCA ACT TAT TTC TGC CAA CAC TAT AAT CGA CCG TGG

ACG TTC GGC CAA GGG 300

ACC AAG GTG GAA ATC AAA GCA - 121 -

IM9 Light Protein SEQ I.D. No. 11

DIQ MTQ FPSTLSASVGDRVTITCRASQSISAWLAWYQQKPGKAPKLLIY KAΞSLESGVPSRFSGSGSGTEFTLTITSLQPDDFATYFCQHYNRPWTFGQGTKVEIK

SEQ. ID NO. 12

GAA ATG CAA CTG GTG GAA TTT GGG GGA GGC CTG CTA CAG CCT GGC

AGG GCC CTG AGA CTC 60 TCC TGT GCA GCC TCT GGA TTC AGG TTT GAT GAT TAT GCC ATG CAC

TGG GTC CGG CAA ACT 120

CCA GGG AAG GGC CTG GAG TGG GTC GCA GGT ATT AGT TGG AAT AGT

GAC ACC ATA GAC TAT 180

TTG CAA ATG AAC AGT CTC AGA GCT GAG GAC ACG GCC TTG TAT TAC

TGT ACA AAA AGA AGG 300

GGG GTG ACA GAC ATT GAC CCT TTT GAT ATC TGG GGC CAA GGG ACA

ATG GTC ATC GTC TCT 360 TCA GAG 366

IM9 HEAVY PROTEIN SEQ I.D. No. 13

EMQLVEFGGGLLQPGRALRLSCAASGFRFDDYAMHWVRQTPGKGLEWVAGISWNSDTIDYA DSVKGRFTISRDNAKNSLYLQMNSLRAEDTALYYCTKRRGVTDIDPFDIWGQGTMVIVSS SEQ I.D. NO 14

DIVMTQSPSSLSVSAGERVTMSCKSSQSLLNSGNQKNFLAWYQQKPGQPPKLLIYGASTRE SGVPDRFTGSGSGTDFTLTISSVQAEDLAVYYCQNDHSYPLTFGAGTKL

SEQ I . D . No 15 DWMTQTPLSLPVSLGDQASISCRSSQSLVHSQGNTYLRWYLQKPGQSPKVLIYKVSNRFS GVPDRFSGSGSGTDFTLKISRVEAEDLGVYFCSQSTHVPWTFGGGTKLE

SEQ I.D. No 16

DIVLTQSPAIMSASPGEKVTMTCSASSSVNYMYWYQQKSGTSPKRWIYDTSKLASGVPVRF SGSGSGTSYSLTISSMETEDAAEYYCQQWGRNPTFGGGTKLEIK SEQ I . D . NO 17

DIQMTQSPASLSASVGETVTITCRASGNIHNYLAWYQQKQGKSPQLLVYYTTTLADGVPSR FSGSGSGTQYSLKINSLQPEDFGSYYCQHFWSTPRTFGGGTKLEIK

SEQ I.D. No 18 EIVLTQSPAITAASLGQKVTITCSASSSVSSLHWYQQKSGTSPKPWIYEISKLASGVPARF SGSGSGTSYSLTINTMEAEDAAIYYCQQWTYPLITFGAGTKLELK

SEQ I.D. No 19 DIQMTQIPSSLSASLGDRVSISCRASQDINNFLNWYQQKPDGTIKLLIYFTSRSQSGVPSR FSGSGSGTDYSLTISNLEQEDIATYFCQQGNALPRTFGGGTKLEIK SEQ I.D. No 20

SVLTQPPSVSGAPGQRVTISCTGSSSNIGAGNHVKWYQQLPGTAPKLLIFHNNARFSVSKS GSΞATLAITGLQAEDEADYYCQSYDRSLRVFGGGTKLTVL

SEQ I.D. No 21 QSVLTQPPSASGTPGQRVTISCSGTSSNIGSSTVNWYQQLPGMAPKLLIYRDAMRPSGVPD RFSGSKSGASASLAIGGLQSEDETDYYCAAWDVSLNAYVFGTGTKVTVL

SEQ I.D. No 22 EVKLVESGGGLVQPGGSLRLSCATSGFTFSDFYMEWVRQPPGKRLEWIAASRNKGNKYTTE YSASVKGRFIVSRDTSQSILYLQMNALRAEDTAIYYCARNYYGSTWYFDVWGAGTTVTVSS

SEQ I.D. No 23 EVKLDETGGGLVQPGRPMKLSCVASGFTFSDYWMNWVRQSPEKGLEWVAQIRNKPYNYETY

YSDSVKGRF ISRDDSKSSVYLQMNNLRVEDMGIYYCTGSYYGMDYWGQGTSVTVSS

SEQ I.D. No 24 VQLQQSGAELMKPGASVKISCKASGYTFSDYWIEWVKQRPGHGLEWIGEILPGSGSTNYHE RFKGKATFTADTSSSTAYMQLNSLTSEDSGVYYCLHGNYDFDGWGQGTTLTVSS SEQ I.D. NO 25

SEQ I.D. No 26 EVKLLESGGGLVQPGGSLKLSCAASGFDFSKYWMSWVRQAPGKGLEWIGEIHPDSGTINYT PSLKDKFIISRDNAKNSLYLQMSQVRSEDTALYYCARLHYYGYNAYWGQGTLVTVSA

SEQ I.D. NO. 27 EVQLQQSGVELVRAGSSVKMSCKASGYTFTSNGINWVKQRPGQGLEWIGYNNPGNGYIAYN EKFKGKTTLTVDKSSSTAYMQLRSLTSEDSAVYFCARSEYYGGSYKFDYWGQGTTLTVSS

SEQ I.D. No 28 VKLEQSGPGLVRPSQTLSLTCTVSGTSFDDYYSTWVRQPPGRGLEWIGYVFYHGTSDTDTP LRSRVTMLVNTSKNQFSLRLSSVTAADTAVYYCARNLIAGCIDVWGQGSLVTVSS SEQ I . D . No 29

EVQLVQSGGGWQPGRSLRLSCSSSGFIFSSYAMYWVRQAPGKGLEWVAIIWDDGSDQHYA DSVKGRFTISRNDSKNTLFLQMDSLRPEDTGVYFCARDGGHGFCSSASCFGPDYWGQGTPV TVSS

DIQMTQFPST LSASVGDRVN ITCRASGFTF SNFGMHWIRQ KPGKGLKWVA

YISGGSSTVH YADSLKGRFT ISRDNPKNEL FLTITSLQPD DFAMYYCARD

YYVNNYWYFD VWGQGTKVEI KR (122 reεidueε) (SEQ. I.D. NO. 30)

SEQ. I.D. Nos. 31 - TAGTGGATCCAACTGATTTCTCCAT

SEQ. I.D. No. 32 - TTATTTACTTCTGGGTCACCAGGTTTATTC

SEQ. I.D. No. 33 - AAGAGGCCGAGCTGGCCCTTCCCTGAATAACCAGGCAGT

SEQ. I.D. NO. 34 - GGGAAGGGCCAGCTCGGCGTGTTCCTATAATATGATCAA SEQ. I.D. No. 35 - TTCCTGGCCCTGCAGGCCCAGTTGTCTGTGTCTTCTGTT

SEQ. I.D. No. 36 - AACTGGGCCTGCAGGGCCAGGAAGCAAAGTTTAAATTCTA

SEQ. I.D. No. 37 - CATGTCTGGATCCAACTGATTT

SEQ. I.D. NO. 38 - CTGATTTACTTCTGGGTGACCAGGTTTATTCAA

SEQ. I.D. NO. 39

5 ' -AAGGGCCAGCTCGGCCTCTTCCTATAATATGATCAATAGTATAAATATTTGTGTTTC- TATTTCCAATCTCAGGTGCCAAATGTGACATCCAGATGACCCA-3 '

SEQ. I.D. No. 40

5'-

TGGGCCTGCAGGGCCAGGAAGCAAAGTTTAAATTCTACTCACGTTTGATTTCCACCTTGG- TT-3 '

#1 = B695 = 5' -GGG-AAG-GGC-CAG-CTC-GGC-CTC-TTC-CTA-TAA-TAT- GAT-CAA-TAG-TAT-AAA-TAT-TTG-TGT-TTC-TAT-TTC-CAA-TCT-CAG-GTG- CCA-AAT-GTG-ACA-TCC-AGA-TGA-CCC-AGT-TTC-CT- 3 (SEQ. I.D. NO. 41)

#2 = B696 = 5' -GCA-TGC-CGA-AGT-TGG-AGA-AGG-TGA-AGC-CGG-AGG- CGC-GGC-AGG-TGA-TGT-TCA-CGC-GGT-CGC-CCA-CGG-AGG-CGG-ACA-GGG- TGG-AAG-GAA-ACT-GGG-TCA-TCT-GGA-TGT- 3 (SEQ. I.D. NO. 42) ' B549 = 5' -GGC-TTC-ACC-TTC-TCC-AAC-TTC-GGC-ATG-CAC-TGG-ATC- CGC-CAG-AAG-CCC-GGC-AAG-GGC-CTG-AAG-TGG-GTG-GCC-TAC-ATC-TCC- GGC-GGC-TCC-TCC-ACC-GTG-CAC-TA- 3 (SEQ. I.D. NO. 43) '

SEQ. I. D. No. 46 #6 = B698 = 5' -CAC-AGA-CAA-CTG-GGC-CTG- CAG-GGC-CAG-GAA-GCA-AAG-TTT-AAA-TTC-TAC-TCA-CGT-TTTG-ATC-TCC- ACC-TTG-GTG-CCC-TGG-CCC-CAC-ACG-TCG-AAG-TAC-CAG-TAG-TT

SEQ. I.D. No. 47 - 5' -GGG-AAG-GGC-CAG-CTC-GGC-CTC-TT -3'

SEQ. I.D. No. 48 - 5 ' -CAC-AGA-CAA-CTG-GGC-CTG-CA- 3'

SEQ I.D. No. 49 - 5 ' -CAC-CAT CCT GTT TGC TTC TTT CCT CAG GAA CTG TGC ACT GGC ACC ACC ACC CAT AGA GGG AGA AGT GCC CCC ACC TGC TCC TCA GTT -3 '

SEQ. I.D. No. 50 5 ' -GGGTAAAAATAGAATGAAGGATGATTTTTATAAAT-3 '

SEQ. I.D. No. 51 GAC ATC CAG ATG ACC CAG TTT CCT TCC ACC

CTG TCC GCC TCC GTG GGC GAC CGC GTG AAC ATC ACC TGC CGC GCC

TCC GGC TCC ACC TTC TCC AAC TTC GGC ATG CAC TGG ATC CGC CAG AAG CCC GGC AAG GGC CTG AAG TGG GTG GCC TAC ATC TCC GGC GGC

TCC TCC ACC GTG CAC TAC GCC AAC TCC CTG AAG GGC CGC TTC ACC

ATC TCC CGC GAC AAC CCC AAG AAC GAG CTG TTC CTG ACC ATC ACC

TCC CTG CAG CCC GAC GAC TTC GCC ATG TAC TAC TGC GCC CGC GAC

TAC TAC GTG AAC AAC TAC TGG TAC TTC GAC GTG TGG GGC CAA GGG ACC AAG GTG GAA ATC AAA 5 ' -AAG-AGC-TCC-TGA-ACC-TCG-CGG-ACA-GTT-AA- ' SEQ . I . D. No . 5 2

5 ' -AAA-TCG-ATC-TCA-GGC-CTC-AGA-CTC-GGC-CTG-ACC-CGT-GGA-AA-3 ' SEQ . I . D . No . 53

SEQ. I.D. No. 54 5 ' -CTG TTT GCT TCT TTC CTC AGG AGG CGG

TTC AGG AGG ATC AGG CGG TTC AGG TGG ATC AGG AGG CGA CAT CCA GAT GAC CCA GTC TCC T-3 '

SEQ. I.D. No. 55 - 5 ' -GTC AGG CTG GAA CTG AGG AGC AGG TGG GGG CAC TTC TCC CTC TAT GGG TGA TGG TGC CAA TGT TTG ATT TCC ACC TTG GTC CCT TGG CCG -AA-3 '

SEQ. I.D. NO. 56 - 5 ' -GAG AAG TGC CCC CAC CTG CTC CTC AGT TCC AGC CTG ACC CCC TCC CAT CCT -3 '

SEQ I.D. NO. , 59

GAA ATG CAA CTG GTG GAA TCT GGG GGA GGC CTG CTA CAG CCT GGC

CGG GCC CTG CGG CTC TCC TGT GCA GCC TCT GGA TTC ACT TTT AGT

AAC TTT GGA ATG CAC TGG ATT CGG CAA ACT CCA GGG AAG GGC CTG

GAG TGG GTC GCA TAC ATT AGT GGT GGC AGT AGT ACC GTC CAC TAT

GCA GAC TCC TTG AAG GGC CGA TTC ACC ATC TCC CGG GAC AAC GCC

AAG AAC TCC CTC TAT TTG CAA ATG ACC AGT CTC CGG GCT GAG GAC

ACG GCC TTG TAT TAC TGT GCA CGG GAT TAC TAC GTT AAT AAC TAC

TGG TAC TTC GAT GTC TGG GGC CAA GGG ACA ATG GTC ATC GTC TCT

TCA G

SEQ I.D. NO. 60

5 ' -GAT CCG AAA TGC AAC TGG TGG AAT CTG GGG GAG GCC TGC TAC AGC CTG GCC GGG CCC TGC GGC TCT CCT GTG CAG CCT CTG GAT TCA CCT TTA G-3 ' SEQ I . D . No . 61

SEQ I.D. No. 62

5 * -GGG TCG CAT ACA TTA GTG GTG GCA GTA GTA CCG TCC ACT ATG CAG ACT CCT TGA AGG GCC GAT TCA CCA TCT CCC GGG ACA ACG CCA AGA A 3 '

SEQ I.D. No. 63

5 ' -TAT TAC TGT GCA CGG GAT TAC TAC GTT AAT AAC TAC TGG TAC TTC GAT GTC TGG GGC CCA GGG ACA ATG GTC ATC GTC TCT TCA -3 ' SEQ I.D. No. 64

5 ' -GTA ATC CCG TGC ACA GTA ATA CAA GGC CGT GTC CTC AGC CCG GAG ACT GTT CAT TTG CAA ATA GAG GGA GTT CTT GGC GTT GTC CCG GGA G -3 ' SEQ I.D. NO. 65

5 ' -AAG GAT CCG AAA TGC AAC TGG TGG AAT CT -3 '

SEQ I.D. No. 66 - GAC GAA TTC TGA AGA GAC GAT GAC CAT TG

Claims

We Claim :

1. A recombinant antibody or antigen binding fragment thereof, comprised of at least one light chain variable domain, which domain, in turn, compriεeε three CDRε wherein the amino acid εequence of one or more of the CDRε iε derived from the amino acid εequence of the correεponding CDR(ε) of a heavy chain variable domain of one (donor) antibody and further compriεes four framework regions wherein the amino acid sequence of one or more of the framework regions is derived from the amino acid sequence of the corresponding framework region(ε) from the light chain variable domain of the εame or a different (acceptor) antibody.

2. A recombinant antibody or antigen binding fragment thereof of claim 1, wherein the antibody or antigen binding fragment thereof iε εelected from the group conεiεting of: a) a CSV-^ fragment; b) a heavy body [CSV_L--C_L] ; c) a kappa body fragment [CDR-grafted V_L--C_L ! I CSV_L—C_L]; d) an intact kappa body {2X [CDR-grafted V_L-- C_L I I CSV_L- -C_H] } ; or e) an ScFv(CSV_L) fragment [either CDR-grafted

V_L-1inker--CSV_L or CSV_L --linker-- CDR- grafted V_L> ]

3. A recombinant fragment of claim 2, wherein the donor and acceptor antibodies are independently chosen from the group consiεting of murine, rabbit, and primate antibodieε.

4. A recombinant antibody fragment of claim 3, wherein the amino acid sequences of all three CDRs of the CSVL domain are derived from the amino acid εequenceε of the correεponding CDRε of the heavy chain variable domain of the donor antibody and the amino acid sequenceε of all four framework regionε of the CSVL domain are derived from the amino acid εequenceε of the corresponding framework regionε of the light chain variable domain of the acceptor antibody.

5. A recombinant antibody fragment of claim 4, wherein the acceptor antibody iε human.

6. A recombinant antibody fragment of claim 5, wherein the human acceptor antibody haε light chainε of the kappa claεε.

7. A recombinant antibody fragment of claim 6, wherein the acceptor antibody is of the IgG claεε.

8. A recombinant antibody fragment of claim 7, wherein the donor antibody iε murine.

9. A recombinant antibody fragment of claim 8, wherein the donor murine antibody haε affinity for tumor antigenε or antigens on thrombi.

10. A recombinant antibody fragment of claim 9, wherein the donor murine antibody fragment has affinity for tumor antigens.

11. A recombinant antibody fragment of claim 10, wherein the donor murine antibody haε affinity for the tumor arkerε choεen from the group conεisting of AFP, CA-125, CEA, Neuron Specific Enolase, C-erb2/Her-2/NEU protein, Cathepsin D, Chromagraninε A, B, and C, the Cytokeratinε, Epidermal Growth Factor Receptor, Epithelial Membrane Antigen, Eεtrogen Receptor, Progeεterone Receptor, Proεtatic Acid Phoεphataεe, Prostate Specific Antigen, Ki67, PGP-170 (MDR) , PGP-180 (MDR) , pl20, Proliferating Cell Nuclear Antigen, Vimentin, and the proteins expressed by the c-myc, N-myc, N-ras, Ki-ras and Ha-raε oncogenes.

12. A recombinant antibody fragment of claim 11, wherein the donor murine antibody has affinity for the tumor antigen CEA.

13. A recombinant antibody fragment of claim 12, wherein the donor antibody is ZCE 025 or CEM 231.

14. A recombinant antibody fragment of claim 13, wherein the acceptor antibody is IM9, and the framework regions are mostly the same as the corresponding IM9 framework regionε.

15. A recombinant antibody fragment of claim 14, wherein the donor antibody is ZCE 025.

16. A recombinant antibody fragment of claim 2, wherein the C-terminus or N-terminus of the fragment molecule iε fuεed to a metal chelating peptide.

17. A recombinant antibody fragment of claim 16, wherein the metal chelating peptide has the amino acid sequence HWHHHP (Sequence I.D.No. 2) and is fused to the C- terminuε of the fragment molecule through the N-terminal hiεtidine reεidue of the chelating peptide.

18. A recombinant antibody fragment of claim 4, wherein the fragment iε selected from the group consiεting of a) kappabody fragment and b) intact kappabody, the amino acid sequences of all three CDRs of the CDR-Grafted V_L domain are derived from the amino acid sequences of the corresponding CDRs of the VL domain of the CDR-Grafted donor antibody, the amino acid sequences of all four framework regions of the CDR-Grafted VL domain are derived from the amino acid sequences of the corresponding framework regions of the VL domain of the CDR-Grafted acceptor antibody, and the light chain constant domains are identical in sequence to the corresponding constant domains of the acceptor antibody or antibodies.

19. A recombinant antibody fragment of claim 18, wherein the amino acid sequence of the framework regions of the CSV_L and the CDR-Grafted V and the C domains are derived from the εame human acceptor antibody.

20. A recombinant antibody fragment of claim 19, wherein the framework regionε of the CSV_L and the CDR-grafted V_L/ as well aε the complete CL domainε are derived from the correεponding regions and domainε of a human acceptor antibody whoεe light chain iε of the kappa claεε.

21. A recombinant antibody fragment of claim 20, wherein the acceptor antibody iε of the IgG class.

22. A recombinant antibody fragment of claim 21, wherein the donor antibody for both the CSVL and the CDR- grafted V_L is the εame murine antibody.

23. A recombinant antibody fragment of claim 22, wherein the donor murine antibody haε affinity for a tumor antigen or an antigen on thrombi.

24. A recombinant antibody fragment of claim 23, wherein the donor murine antibody haε affinity for a tumor antigen.

25. A recombinant antibody fragment of claim 24, wherein both donor murine antibodies have affinity for tumor antigens chosen from the group consisting of AFP, CA-125, CEA, Neuron Specific Enolase, C-erb2/Her-2/NEU protein, Cathepsin D, Chromagranins A, B, and C, the Cytokeratins, Epidermal Growth Factor Receptor, Epithelial Membrane Antigen, Estrogen - 131 -

Receptor, Progesterone Receptor, Prostatic Acid Phoεphataεe, Proεtate Specific Antigen, Ki-67, PGP-170 (MDR), PGP-180 (MDR), pl20, Proliferating Cell Nuclear Antigen, Vi entin, and the proteinε expreεsed by the c-myc, N-myc, N-ras, Ki-ras and Ha-raε oncogeneε.

26. A recombinant antibody fragment of claim 25, wherein the donor murine antibody haε affinity for the tumor antigen CEA.

27. A recombinant antibody fragment of claim 26, wherein the donor murine antibody iε either ZCE 025 or CEM 231.1.

28. A recombinant antibody fragment of claim 27, wherein the human acceptor antibody iε IM9, and the amino acid εequenceε of both εets of framework regions are derived from the amino acid sequenceε of the correεponding IM9 light chain framework regionε.

29. A recombinant antibody fragment of claim 28, wherein the murine donor antibody iε ZCE 025.

30. A recombinant antibody fragment of claim 29, wherein the C-terminuε or the N-terminuε of either the CSV

.containing or the CDR-Grafted - containing chain of the fragment molecule iε fused to a metal chelating peptide.

31. A recombinant antibody fragment of claim 30, wherein the metal chelating peptide haε the amino acid sequence HWHHHP and is fused to the C-terminus of the CSV containing chain of the fragment molecule through the N- terminal histidine residue of the chelating peptide.

32. A recombinant antibody or fragment thereof of claim 2, wherein the antibody or fragment thereof is the fragment SCFV(CSVL), wherein a CDR-Grafted VL domain iε covalently bonded to a CSVL domain through a polypeptide linker.

33. A recombinant fragment of claim 32, wherein the donor and acceptor antibodies are independently chosen from the group consiεting of murine, rabbit, and primate antibodieε.

34. A recombinant antibody fragment of claim 33, wherein a) The amino acid εequence of all three CDRε of the CSVL domain derived from thoεe of the correεponding CDRε of the heavy chain variable domain of the donor antibody used;

b) The amino acid sequence of all four CSVL framework regions are derived from those of the correεponding framework regionε of the light chain variable domain of the acceptor antibody uεed;

c) The amino acid sequences of all three CDRs of the CDR-Grafted V domain are derived from those of the corresponding CDRs of the VL domain of the donor antibody uεed; and

d) The amino acid εequenceε of all four framework regionε of the CDR-Grafted VL domain are derived from those of the correεponding framework regions of the VL domain of the acceptor antibody.

35. A recombinant antibody fragment of claim 34, wherein the amino acid sequenceε of the framework regionε of both the CSV_L and the CDR-Grafted V are derived from those of the corresponding framework regions of the same human acceptor antibody.

36. A recombinant antibody fragment of claim 35, wherein the polypeptide linker is composed of about 12 to about 18 amino acids.

37. A recombinant antibody fragment of claim 36, wherein the C-terminus of the CDR-Grafted V_L domain is fused to the N-terminus of the polypeptide linker, and wherein the C-terminus of the polypeptide linker is bonded to the N- terminuε of the CSV-^ domain.

38. A recombinant antibody fragment of claim 37, wherein the donor antibody is murine.

39. A recombinant antibody fragment of claim 38, wherein the donor murine antibody has affinity for tumor antigens or antigens on thrombi.

40. A recombinant antibody fragment of claim 39, wherein the donor murine antibody haε affinity for a tumor antigen.

41. A recombinant antibody fragment of claim 40, wherein the donor murine antibody haε affinity for tumor antigenε choεen from the group consisting of AFP, CA-125, CEA, Neuron Specific Enolase, C-erb2/Her-2/NEU protein, Cathepsin D, Chromagranins A, B, and C, the Cytokeratinε, Epidermal Growth Factor Receptor, Epithelial Membrane Antigen, Eεtrogen Receptor, Progeεterone Receptor, Proεtatic Acid Phoεphataεe, Prostate Specific Antigen, Ki-67, PGP-170 (MDR), PGP-180 (MDR), pl20, Proliferating Cell Nuclear Antigen, Vimentin, and the proteins expresεed by the c-myc, N-myc, N-raε, Ki-raε and Ha-raε oncogenes

42. A recombinant antibody fragment of claim 41, wherein the donor murine antibody has affinity for the tumor antigen CEA.

43. A recombinant antibody fragment of claim 42, wherein the linker polypeptide is composed of εerine and glycine amino acid reεidueε.

44. A recombinant antibody fragment of claim 43, wherein the donor murine antibody iε either ZCE 025 or CEM 231.1.

45. A recombinant antibody fragment of claim 44, wherein the human acceptor antibody iε IM9, and both εetε of framework regionε are moεtly the εame aε the correεponding IM9 light chain framework regionε.

46. A recombinant antibody fragment of claim 45, wherein the murine donor antibody iε ZCE025.

47. A recombinant antibody fragment of claim 46, wherein the linker polypeptide iε of the formula -GGSGGSGGSGGSGG-.

48. A recombinant antibody fragment of claim 47, wherein the C-terminus or the N-terminus of the SCFV(C_SVL) is fused to a metal chelating peptide.

49. A recombinant antibody fragment of claim 48 wherein the metal chelating peptide has the amino acid sequence HWHHHP and is fused to the C-terminus of the CSVL domain of the fragment molecule through the N-terminal hiεtidine reεidue of the chelating peptide.

50. A DNA or RNA εequence coding for a recombinant antibody or fragment thereof, wherein the antibody or fragment thereof iε compriεed of at leaεt one light chain variable domain, which domain, in turn, comprises three CDRs wherein the amino acid sequence of one or more of the CDRs is derived from the amino acid sequence of the corresponding CDR(ε) of a heavy chain variable domain of one (donor) antibody and further comprises four framework regions wherein the amino acid sequence of one or more of the framework regions are derived from the amino acid sequence of the corresponding framework region(s) from the light chain variable domain of the same or a different (acceptor) antibody.

51. A DNA or RNA sequence coding for a recombinant antibody or antigen binding fragment thereof of claim 50, wherein the recombinant antibody or antigen binding fragment thereof is selected from the group consiεting of:

a) a CSVL fragment; b) a heavy body [CSV_L--C J ; c) a kappa body fragment [CDR-grafted VL~-CL I I CSV_L—C_L]; d) an intact kappa body {2X [CDR-grafted V_L--C_L I I CSV_L--C_H]}; or e) an ScFv(CSV_L) fragment [either CDR-grafted

V --linker--CSV_L or CSV_L --linker-- CDR- grafted V_L. ]

52. A DNA or RNA sequence of claim 51, wherein in the recombinant fragment the donor and acceptor antibodies that are coded for are independently chosen from the group consiεting of murine, rabbit, and primate antibodieε.

53. A DNA or RNA εequence of claim 52, wherein the amino acid εequenceε of all three CDRε of the CSVL domain that are coded for are derived from the amino acid sequences of the corresponding CDRs of the heavy chain variable domain of the donor antibody and the amino acid sequenceε of all four framework regions of the CSVL are derived from the amino acid sequences of the corresponding framework regions of the light chain variable domain of the acceptor antibody.

54. A DNA or RNA εequence of claim 53, wherein the acceptor antibody that iε coded for iε human.

55. A DNA or RNA εequence of claim 54, wherein the human acceptor antibody that iε coded for haε light chainε of the kappa claεε.

56. A DNA or RNA εequence of claim 55, wherein the acceptor antibody that iε coded for iε of the IgG class.

57. A DNA or RNA sequence of claim 56, wherein the donor antibody that iε coded for iε murine.

58. A DNA or RNA εequence of claim 57, wherein the donor murine antibody that iε coded for haε affinity for tumor antigenε or antigenε on thrombi.

59. A DNA or RNA εequence of claim 58, wherein the donor murine antibody fragment that iε coded for has affinity for tumor antigens.

60. A DNA or RNA sequence of claim 59, wherein the donor murine antibody that iε coded for haε affinity for the tumor markerε choεen from the group consisting of AFP, CA- 125, CEA, Neuron Specific Enolase, C-erb2/Her-2/NEU protein, Cathepsin D, Chromagranins A, B, and C, the Cytokeratinε, Epidermal Growth Factor Receptor, Epithelial Membrane Antigen, Eεtrogen Receptor, Progeεterone Receptor, Prostatic Acid Phosphatase, Prostate Specific Antigen, Ki-67, PGP-170 (MDR), PGP-180 (MDR), pl20, Proliferating Cell Nuclear

Antigen, Vimentin, and the proteins expressed by the c-myc, N-myc, N-ras, Ki-ras and Ha-raε oncogeneε.

61. A DNA or RNA sequence of claim 60, wherein the donor murine antibody that iε coded for haε affinity for the tumor antigen CEA.

62. A DNA or RNA sequence of claim 61, wherein the donor antibody that is coded for is ZCE 025 or CEM 231.

63. A DNA or RNA sequence of claim 62, wherein the acceptor antibody that is coded for is IM9, and the framework regions are mostly the same as the corresponding IM9 framework regions.

64. A DNA or RNA sequence of claim 63, wherein the donor antibody that is coded for is ZCE 025.

65. A DNA or RNA εequence of claim 51 wherein the 5' -terminus or 3 ' -terminus of the DNA or RNA coding for the fragment molecule is fused to a DNA or RNA sequence respectively, coding for metal chelating peptide.

66 . A DNA or RNA sequence of claim 65, wherein the metal chelating peptide that is coded for has the amino acid sequence HWHHHP (Sequence I.D.No.) and is fused to the C- terminuε of the fragment molecule through the N-terminal hiεtidine reεidue of the chelating peptide.

67. A DNA or RNA εequence of claim 51, whereinthe fragment encoded is selected from the group consisting of a) kappabody and b) intact kappabody and the amino acid sequenceε of all three CDRε that are coded for of the CDR- Grafted V_L domain are derived from the amino acid εequenceε of the corresponding CDRs of the VL domain of the CDR-Grafted donor antibody and the amino acid sequences of all four framework regions that are coded for of the CDR-Grafted VL domain are derived from the amino acid sequences of the corresponding framework regions of the VL domain of the CDR- Grafted acceptor antibody, and the light chain constant domains that are coded for are identical in sequence to the corresponding conεtant domainε of the acceptor antibody or antibodies.

68. A DNA or RNA sequence of claim 67, wherein the amino acid sequences of the framework regions of both the CSV_L and the CDR-Grafted V_L and the amino acid sequence of the C domains coded for acceptor antibodies are derived from the same human acceptor antibody.

69. A DNA or RNA sequence of claim 68, wherein the amino acid sequence of the framework regionε of the CSV_L and the CDR-Grafted VL and the amino acid εequence of the C_L domainε that are coded for are derived from the correεponding regions and domains of a human acceptor antibody whose light chain iε of the kappa claεε.

70. A DNA or RNA sequence of claim 69, wherein the acceptor antibody that iε uεed iε of the IgG claεε.

71. A DNA or RNA εequence of claim 70, wherein the donor antibody used for both is the same murine antibody.

72. A DNA or RNA sequence of claim 71, wherein the donor murine antibody used has affinity for a tumor antigen or an antigen on thrombi.

73. A DNA or RNA sequence of claim 72, wherein the donor murine antibody used has affinity for a tumor antigen.

74. A DNA or RNA sequence of claim 73, wherein the donor murine antibody that is used has affinity for tumor antigens chosen from the group consiεting of AFP, CA-125, CEA, Neuron Specific Enolaεe, C-erb2/Her-2/NEU protein, Cathepεin D, Chromagranins A, B, and C, the Cytokeratins, Epidermal Growth Factor Receptor, Epithelial Membrane Antigen, Eεtrogen Receptor, Progesterone Receptor, Prostatic Acid Phoεphatase, Prostate Specific Antigen, Ki-67, PGP-170 (MDR), PGP-180 (MDR), pl20, Proliferating Cell Nuclear Antigen, Vimentin, and the proteins expressed by the c-myc, N-myc, N-ras, Ki-ras and Ha-raε oncogeneε.

75. A DNA or RNA εequence of claim 74, wherein the donor murine antibody uεed has affinity for the tumor antigen

CEA.

76. A DNA or RNA sequence of claim 75, wherein the donor murine antibody used is either ZCE 025 or CEM 231.1.

77. A DNA or RNA sequence of claim 76, wherein the human acceptor antibody used is IM9, and both setε of framework regionε are derived from the correεponding IM9 light chain framework regionε.

78. A DNA or RNA εequence of claim 77, wherein the murine donor antibody used is ZCE 025.

79. A DNA or RNA sequence of claim 78, wherein the C-terminus or the N-terminus of either the CSV_L .containing or the CDR-Grafted - containing chain of the fragment molecule is fused to a metal chelating peptide.

80. A DNA or RNA sequence of claim 79, wherein the metal chelating peptide that is coded for has the amino acid sequence HWHHHP and is fused to the C-terminus of the CSV containing chain of the fragment molecule through the N- terminal histidine reεidue of the chelating peptide.

81. A DNA or RNA sequence coding for a recombinant antibody or fragment thereof of claim 51, wherein the antibody or fragment thereof that is coded for iε the fragment SCFV(CSVL) , wherein a CDR-Grafted V_L domain iε covalently bonded to a CSVj_^ domain through a polypeptide linker.

82. A DNA or RNA sequence of claim 81, wherein the donor and acceptor antibodies that are coded for are independently choεen from the group conεisting of murine, rabbit, and primate antibodieε.

83. A DNA or RNA εequence of claim 82, wherein a) The amino acid εequences of all three

CDRε of the CSV_L domain that are coded for are derived from those of the correεponding CDRε of the heavy chain variable domain of the donor antibody uεed; b) The amino acid εequenceε of all four CSVL framework regionε that are coded for are derived from thoεe of the correεponding framework regionε of the light chain variable domain of the acceptor antibody uεed; c) The amino acid εequenceε of all three CDRε of the CDR-Grafted V_L domain that are coded for are derived from thoεe of the correεponding CDRε of the V_L domain of the donor antibody used; and d) The amino acid sequences of all four framework regions of the CDR-Grafted V_L domain that are coded for are derived from those of the corresponding framework regions of the VL domain of the acceptor antibody used.

84. A DNA or RNA εequence of claim 83, wherein the amino acid sequenceε of the framework regionε of both the

CΞV_L and the CDR-Grafted V_L are derived from those of the corresponding framework regions of the same human acceptor antibody.

85. A DNA or RNA εequence of claim 84, wherein the polypeptide linker that iε coded for iε compoεed of about 12 to about 18 amino acidε.

86. A DNA or RNA εequence of claim 85, wherein the C-terminuε of the CDR-Grafted V_L domain that iε coded for is fused to the N-terminus of the polypeptide linker, and wherein the C-terminus of the polypeptide linker that is coded for is fused to the N-terminus of the CSVL domain.

87. A DNA or RNA sequence of claim 86, wherein the donor antibody that is uεed iε murine.

88. A DNA or RNA εequence of claim 87, wherein the donor murine antibody that is used haε affinity for tumor antigenε or antigenε on thrombi.

89. A DNA or RNA εequence of claim 88, wherein the donor murine antibody that iε used has affinity for a tumor antigen.

90. A DNA or RNA sequence of claim 89, wherein the donor murine antibody that is used has affinity for tumor antigenε choεen from the group consisting of AFP, CA-125, CEA, Neuron Specific Enolase, C-erb2/Her-2/NEU protein, Cathepsin D, Chromagranins A, B, and C, the Cytokeratinε, Epidermal Growth Factor Receptor, Epithelial Membrane Antigen, Eεtrogen Receptor, Progesterone Receptor, Prostatic Acid Phosphatase, Prostate Specific Antigen, Ki-67, PGP-170 (MDR), PGP-180 (MDR), pl20, Proliferating Cell Nuclear Antigen, Vimentin, and the proteins expressed by the c-myc, N-myc, N-ras, Ki-ras and Ha-ras oncogenes

91. A DNA or RNA εequence of claim 90, wherein the donor murine antibody used has affinity for the tumor antigen CEA.

92. A DNA or RNA sequence of claim 91, wherein the linker polypeptide that is coded for is composed of serine and glycine amino acid residues.

93. A DNA or RNA εequence of claim 92, wherein the donor murine antibody uεed iε either ZCE 025 or CEM 231.1.

94. A DNA or RNA εequence of claim 93, wherein the human acceptor antibody uεed is IM9, and the amino acid sequences of both the CSV_L and CDR-Grafted V framework regions that are coded for are derived from those of the corresponding IM9 light chain framework regions.

95. A DNA or RNA εequence of claim 94, wherein the murine donor antibody used is ZCE025.

96. A DNA or RNA εequence of claim 95, wherein the linker polypeptide that is coded for is of the formula

-GGSGGSGGSGGSGG-.

97. A DNA or RNA sequence of claim 96, wherein the C-terminus or the N-terminus of the SCFΓ(CSVL) that is coded for iε fuεed to a metal chelating peptide.

98. A DNA or RNA sequence of claim 97 wherein the metal chelating peptide that is coded for has the amino acid sequence HWHHHP and is fuεed to the C-terminuε of the CΞV_L domain of the fragment molecule through the N-terminal histidine residue of the chelating peptide.