Heterodimeric antibodies that bind to claudin 18.2 and CD3

Abstract

The present invention is directed to antibodies, including novel antigen-binding domains and heterodimeric antibodies, that bind to claudin 18.2 (CLDN18.2).

Description

[Technical Field]

[0001] Priority This application claims priority to U.S. Provisional Patent Application No. 63 / 210,787, filed on 15 June 2021, which is incorporated herein by reference in its entirety for any purpose.

[0002] Sequence List This application includes a sequence listing filed electronically in ASCII format and incorporated herein by reference in its entirety. The ASCII copy, created on 13 June 2022, is named 067461-5232-WO_SL.txt and has a size of 1,371,138 bytes. [Background technology]

[0003] Antibody-based therapies have proven successful in treating a variety of diseases, including cancer. An increasingly popular avenue being explored is the engineering of a single immunoglobulin molecule that simultaneously binds to two different antigens. Such alternative antibody formats that bind to two different antigens are often referred to as bispecific antibodies. A typical approach to producing bispecific antibodies is to introduce a novel variable region into the antibody, as the considerable diversity of antibody variable regions (Fv) allows for the production of Fvs that recognize virtually any molecule.

[0004] A particularly useful approach with bispecific antibodies is to manipulate the first binding domain, which links to CD3, and the second binding domain, which links to an antigen that associates with cancer cells or is upregulated by cancer cells, thereby allowing the bispecific antibody to redirect CD3+ T cells and destroy cancer cells.

[0005] Claudins are a family of endogenous tight junction membrane proteins. Claudin 18 (or CLDN18) is one such protein in the claudin family. Claudin 18 can be expressed as two different splice variants. The claudin 18 isoform A2 splice variant (or CLDN18.2) is expressed on gastric cells, particularly gastric epithelial cells. Notably, CLDN18.2 has also been previously reported to be highly expressed in several cancers, including gastric, esophageal, and pancreatic cancers. Given this, anti-CLDN18.2 antibodies are useful, for example, for localizing antitumor agents (e.g., chemotherapeutic agents and T cells) to such CLDN18.2-expressing tumors.

[0006] The present invention provides novel bispecific antibodies against CD3 and CLDN18.2 that can localize CD3+ effector T cells to CLDN18.2-expressing tumors. [Overview of the project]

[0007] Accordingly, claudin 18.2 (CLDN18.2) antigen-binding domains and anti-CLDN18.2 antibodies (e.g., bispecific antibodies) are provided herein.

[0008] In one embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a common light chain is provided herein. The first monomer comprises VH-CH1-domain linker-scFv-domain linker-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises VH-CH1-hinge-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the second variant Fc domain. The common light chain comprises VL-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, one of the first and second variant Fc domains each contains the amino acid substitution E233P / L234V / L235A / G236del / S267K, one of the variant Fc domains contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. The VH domain contains SEQ ID NO: 81 or SEQ ID NO: 82, and the VL domain contains SEQ ID NO: 84. In some embodiments, scFv has sequences selected from the group consisting of SEQ ID NOs: 87, 88, 98, 99, 109, 110, 120, 121, 131, 132, 142, and 143.

[0009] In some embodiments, the antibodies are XENP24647, XENP31729, XENP24649, XENP31723, XENP31725, XENP31727, XENP29476, XENP29478, XENP31724, XENP31726, XENP31728, XENP29477, XENP29479, XENC10101, XENC10102, XENC10103, XENC 10104, XENC10105, XENC10106, XENC10107, XENC10108, XENC10109, XENC10110, XENC10111, XENC10112, XENC 10113, XENC10114, XENC10115, XENC10116, XENC10117, XENC10118, XENC10119, XENC10120, XENC10121, XENC 10122, XENC10123, XENC10124, XENC10125, XENC10126, XENC10127, XENC10128, XENC10129, XENC10130, XENC 10131, XENC10132, XENC10133, XENC10134, XENC10135, XENC10136, XENC10137, XENC10138, XENC10139, XENC Selected from the group consisting of 10140, XENC10141, XENC10142, XENC10143, XENC10144, XENC10145, XENC10146, XENC10147, XENC10148, XENC10149, XENC10150, XENC10151, XENC10152, XENC10153, XENC10154, XENC10155, and XENC10156.

[0010] In another embodiment, a composition comprising an anti-CLDN18.2 antigen-binding domain (ABD) is provided herein, the ABD comprising a) a variable heavy chain domain comprising SEQ ID NO: 81, and b) a variable light chain domain comprising SEQ ID NO: 84.

[0011] In another embodiment, a composition comprising an anti-CLDN18.2 antigen-binding domain (ABD) is provided herein, the ABD comprising a) a variable heavy chain domain comprising SEQ ID NO: 82, and b) a variable light chain domain comprising SEQ ID NO: 84.

[0012] In one embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a third monomer is provided herein. The first monomer comprises scFv-domain linker-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises VH-CH1-hinge-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the second variant Fc domain. The third monomer comprises VL-CL. In such antibodies, one of the variant Fc domains contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, one of the first and second variant Fc domains each contains the amino acid substitution E233P / L234V / L235A / G236del / S267K, one of the variant Fc domains contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. In some embodiments, scFv has a sequence selected from the group consisting of SEQ ID NOs: 87, 88, 98, 99, 109, 110, 120, 121, 131, 132, 142, and 143. In some embodiments, the variable heavy (VH) domain includes SEQ ID NO: 81 or SEQ ID NO: 82, and the variable light (Vl) domain includes SEQ ID NO: 84.

[0013] In some embodiments, the CH1-hinge-CH2-CH3 component of the second heavy chain comprises SEQ ID NO: 32, the first variant Fc domain comprises SEQ ID NO: 33, and the constant light chain domain comprises SEQ ID NO: 74. In some embodiments, the antibody is selected from the group consisting of XENP29472, XENP29473, XENP29474, and XENP29475.

[0014] In another embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a common light chain is provided herein. The first monomer comprises VH1-CH1-hinge-CH2-CH3-domain linker-VH2 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises VH1-CH1-hinge-CH2-CH3-domain linker-VL2 from the N-terminus to the C-terminus, wherein CH2-CH3 is the second variant Fc domain. The common light chain comprises VL1-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, and the first and second variant Fc domains each contain the amino acid substitution E233P / L234V / L235A / G236del / S267K, one variant Fc domain contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. VH2 and VL2 are selected from pairs consisting of SEQ ID NOs. 89 and 93, SEQ ID NOs. 100 and 104, SEQ ID NOs. 111 and 115, SEQ ID NOs. 122 and 126, SEQ ID NOs. 133 and 137, and SEQ ID NOs. 144 and 148. Furthermore, the VH1 domain includes sequence number 81 or sequence number 82, and the VL1 domain includes sequence number 84.

[0015] In another embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a common light chain is provided herein. The first monomer comprises VH1-CH1-hinge-CH2-CH3-domain linker-scFv from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises VH1-CH1-hinge-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the second variant Fc domain. The common light chain comprises VL1-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, and the first and second variant Fc domains each contain the amino acid substitution E233P / L234V / L235A / G236del / S267K, one variant Fc domain contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. scFv has a sequence selected from the group consisting of SEQ ID NOs: 87, 88, 98, 99, 109, 110, 120, 121, 131, 132, 142, and 143. Furthermore, the VH1 domain includes sequence number 81 or sequence number 82, and the VL1 domain includes sequence number 84.

[0016] In one embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a common light chain is provided herein. The first monomer comprises VH1-CH1-domain linker-VH2-domain linker-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises VH1-CH1-domain linker-VL2-domain linker-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the second variant Fc domain. The common light chain comprises VL1-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, and the first and second variant Fc domains each contain the amino acid substitution E233P / L234V / L235A / G236del / S267K, one variant Fc domain contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. VH2 and VL2 are selected from pairs consisting of SEQ ID NOs. 89 and 93, SEQ ID NOs. 100 and 104, SEQ ID NOs. 111 and 115, SEQ ID NOs. 122 and 126, SEQ ID NOs. 133 and 137, and SEQ ID NOs. 144 and 148. Furthermore, the VH1 domain includes sequence number 81 or sequence number 82, and the VL1 domain includes sequence number 84.

[0017] In another embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a light chain is provided herein. The first monomer comprises VH-CH1-domain linker-scFv-domain linker-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises a second variant Fc domain comprising CH2-CH3. The light chain comprises VL-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, and the first and second variant Fc domains each contain the amino acid substitution E233P / L234V / L235A / G236del / S267K, one variant Fc domain contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. scFv has a sequence selected from the group consisting of SEQ ID NOs: 87, 88, 98, 99, 109, 110, 120, 121, 131, 132, 142, and 143. Furthermore, the VH domain includes sequence number 81 or sequence number 82, and the VL domain includes sequence number 84.

[0018] In one embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a light chain is provided herein. The first monomer comprises scFv-domain linker-VH-CH1-hinge-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises a second variant Fc domain comprising CH2-CH3. The light chain comprises VL-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, and the first and second variant Fc domains each contain the amino acid substitution E233P / L234V / L235A / G236del / S267K, one variant Fc domain contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. scFv has a sequence selected from the group consisting of SEQ ID NOs: 87, 88, 98, 99, 109, 110, 120, 121, 131, 132, 142, and 143. Furthermore, the VH domain includes sequence number 81 or sequence number 82, and the VL domain includes sequence number 84.

[0019] In another embodiment, a heterodimer antibody comprising a first monomer, a second monomer, and a common light chain is provided herein. The first monomer comprises scFv-domain linker-VH-CH1-hinge-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the first variant Fc domain. The second monomer comprises VH-CH1-hinge-CH2-CH3 from the N-terminus to the C-terminus, wherein CH2-CH3 is the second variant Fc domain. The common light chain comprises VL-CL. In such antibodies, one variant Fc domain contains the amino acid substitution N208D / Q295E / N384D / Q418E / N421D, and the first and second variant Fc domains each contain the amino acid substitution E233P / L234V / L235A / G236del / S267K, one variant Fc domain contains the amino acid substitution S364K / E357Q, and the other variant Fc domain contains the amino acid substitution L368D / K370S, and the numbering follows the EU index as in the case of Kabat. scFv has a sequence selected from the group consisting of SEQ ID NOs: 87, 88, 98, 99, 109, 110, 120, 121, 131, 132, 142, and 143. Furthermore, the VH domain includes sequence number 81 or sequence number 82, and the VL domain includes sequence number 84.

[0020] In another embodiment, nucleic acid compositions comprising nucleic acids encoding either heterodimeric antibodies or antigen-binding domains described herein are provided herein.

[0021] In yet another embodiment, an expression vector comprising any of the nucleic acids described herein is provided herein.

[0022] In one embodiment, host cells transformed with any of the expression vectors or nucleic acids described herein are provided herein.

[0023] In another aspect, provided herein is a method of making a heterodimeric antibody or antigen-binding domain of the subject matter described herein. The method includes culturing a host cell transformed with any of the expression vectors or nucleic acids described herein under conditions in which the antibody or antigen-binding domain is expressed, and recovering the antibody or antigen-binding domain.

[0024] In some embodiments, provided herein is a method of treating cancer, the method comprising administering to a patient in need thereof any one of the antibodies of the subject matter described herein. In some embodiments, the cancer is gastric cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] [Figure 1A] (Including skews and pI variants) shows useful pairs of Fc heterodimerization variant sets. There are variants that do not have the corresponding "monomer 2" variant, and these are pI variants that can be used alone for any monomer. [Figure 1B] (Including skews and pI variants) shows useful pairs of Fc heterodimerization variant sets. There are variants that do not have the corresponding "monomer 2" variant, and these are pI variants that can be used alone for any monomer. [Figure 1C] (Including skews and pI variants) shows useful pairs of Fc heterodimerization variant sets. There are variants that do not have the corresponding "monomer 2" variant, and these are pI variants that can be used alone for any monomer. [Figure 1D] (Including skews and pI variants) shows useful pairs of Fc heterodimerization variant sets. There are variants that do not have the corresponding "monomer 2" variant, and these are pI variants that can be used alone for any monomer. [Figure 1E]This shows useful pairs of Fc heterodimerization variants (including skew and pI variants). Some variants do not have a corresponding "monomer 2" variant; these are pI variants that can be used independently for either monomer. [Figure 2] The constant regions and their respective substitutions for the isoelectron-distributed variant antibodies are shown. pI_(-) indicates a low pI variant, and pI_(+) indicates a high pI variant. These can be optionally and independently combined with other heterodimerized variants of the present invention (and other variant types outlined herein). [Figure 3] These variants exhibit useful removal variants that eliminate the FcγR bond (sometimes referred to as "knockout" or "KO" variants). Generally, removal variants are found on both monomers, but in some cases, they may be present on only one of the monomers. [Figure 4] Particularly useful embodiments of the "non-Fv" components of the present invention are shown. [Figure 5] As components, several charged scFv linkers are shown, which are used to increase or decrease the pI of heterodimer antibodies utilizing one or more scFvs. (+H) positive linkers are used in particular herein. A single prior art scFv linker with a single charge is referred to as "Whitlow" (Whitlow et al., Protein Engineering 6(8):989-995 (1993)). It should be noted that this linker was used to reduce aggregation in scFvs and to enhance proteolytic stability. It should also be noted that any or all of these linkers may be used as optional domain linkers considered herein, and in particular, those listed as uncharged "additional scFv linkers" may be used in particular. [Figure 6A]The following are sequences of several useful 1+1 Fab-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, lacking Fv sequences (e.g., scFv and VH on the Fab side). That is, the initial " / " slash of the "Fab-Fc side" indicates that the C-terminus of the VH outlined herein is bound at that point. Similarly, the initial " / " slash of the "scFv-Fc side" indicates that the C-terminus of the scFv (e.g., VH-scFv linker-VL or VL-scFv linker-VH) outlined herein is bound at that point. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, C220S on the strand having the S364K / E357Q scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 2 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368D / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.The skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, C220S on the strand having the D401K scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand having the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand with the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is based on human IgG4 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the S228P (EU numbering, which is S241P in Kabat) variant on both strands that removes the Fab arm exchange as known in the art. Skeleton 9 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant.Skeleton 10 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, and the S267K variant on both strands. Skeleton 11 is identical to skeleton 1 except that it includes the M428L / N434S Xtend mutation. The skeleton 12 is based on human IgG1 (356E / 358M allotype) and includes the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q:L368D / K370S scuba riant, C220S, and S364K / E357Q scuba riant, as well as the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as understood by those skilled in the art. In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 6B]The following are sequences of several useful 1+1 Fab-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, lacking Fv sequences (e.g., scFv and VH on the Fab side). That is, the initial " / " slash of the "Fab-Fc side" indicates that the C-terminus of the VH outlined herein is bound at that point. Similarly, the initial " / " slash of the "scFv-Fc side" indicates that the C-terminus of the scFv (e.g., VH-scFv linker-VL or VL-scFv linker-VH) outlined herein is bound at that point. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, C220S on the strand having the S364K / E357Q scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 2 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368D / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.The skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, C220S on the strand having the D401K scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand having the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand with the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is based on human IgG4 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the S228P (EU numbering, which is S241P in Kabat) variant on both strands that removes the Fab arm exchange as known in the art. Skeleton 9 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant.Skeleton 10 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, and the S267K variant on both strands. Skeleton 11 is identical to skeleton 1 except that it includes the M428L / N434S Xtend mutation. The skeleton 12 is based on human IgG1 (356E / 358M allotype) and includes the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q:L368D / K370S scuba riant, C220S, and S364K / E357Q scuba riant, as well as the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as understood by those skilled in the art. In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 6C]The following are sequences of several useful 1+1 Fab-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, lacking Fv sequences (e.g., scFv and VH on the Fab side). That is, the initial " / " slash of the "Fab-Fc side" indicates that the C-terminus of the VH outlined herein is bound at that point. Similarly, the initial " / " slash of the "scFv-Fc side" indicates that the C-terminus of the scFv (e.g., VH-scFv linker-VL or VL-scFv linker-VH) outlined herein is bound at that point. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, C220S on the strand having the S364K / E357Q scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 2 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368D / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.The skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, C220S on the strand having the D401K scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand having the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand with the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is based on human IgG4 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the S228P (EU numbering, which is S241P in Kabat) variant on both strands that removes the Fab arm exchange as known in the art. Skeleton 9 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant.Skeleton 10 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, and the S267K variant on both strands. Skeleton 11 is identical to skeleton 1 except that it includes the M428L / N434S Xtend mutation. The skeleton 12 is based on human IgG1 (356E / 358M allotype) and includes the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q:L368D / K370S scuba riant, C220S, and S364K / E357Q scuba riant, as well as the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as understood by those skilled in the art. In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 6D]The following are sequences of several useful 1+1 Fab-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, lacking Fv sequences (e.g., scFv and VH on the Fab side). That is, the initial " / " slash of the "Fab-Fc side" indicates that the C-terminus of the VH outlined herein is bound at that point. Similarly, the initial " / " slash of the "scFv-Fc side" indicates that the C-terminus of the scFv (e.g., VH-scFv linker-VL or VL-scFv linker-VH) outlined herein is bound at that point. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, C220S on the strand having the S364K / E357Q scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 2 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368D / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, C220S on the strand with the S364K scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.The skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, C220S on the strand having the D401K scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand having the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, C220S on the strand with the S364K / E357Q scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand with the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is based on human IgG4 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the S228P (EU numbering, which is S241P in Kabat) variant on both strands that removes the Fab arm exchange as known in the art. Skeleton 9 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant.Skeleton 10 is based on human IgG2 and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, and the S267K variant on both strands. Skeleton 11 is identical to skeleton 1 except that it includes the M428L / N434S Xtend mutation. The skeleton 12 is based on human IgG1 (356E / 358M allotype) and includes the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q:L368D / K370S scuba riant, C220S, and S364K / E357Q scuba riant, as well as the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as understood by those skilled in the art. In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 7A]The following shows the sequences of several useful 2+1 Fab2-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, without Fv and linker sequences. For example, on the "Fab-Fc side", "(VH domain / )" indicates that the C-terminus of the VH domain is bound at its position, which is the beginning of the CH1 domain of the heavy chain. Similarly, "VH-CH1-domain linker 1-scFv-domain linker 2 / " at the beginning of the "Fab-scFv-Fc side" indicates that these sequences are variant CH2-CH3 This indicates binding to the enumerated sequences, which are Fc domains. As discussed herein, the scFv domain can be in either the direction of the -VH-scFv linker-VL- or -VL-scFv linker-VH- as discussed herein. In addition, preferred domain linkers for “domain linker 2” are those enumerated in Figure 5 as “useful domain linkers”. Note that sequence identifiers are for the enumerated sequences only. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and has S364K / E357Q:L368D / K370S scuba riant, N208D / Q295E / N384D / Q418E / N421D on a strand having L368D / K370S scuba riant The pI variant and the E233P / L234V / L235A / G236del / S267K removal variant are included on both strands. The skeleton 2 is based on human IgG1 (356E / 358M allotype) and contains S364K:L368D / K370S scuba ant and N208D / Q295E / N384D / Q418E / N421D on the strand having L368D / K370S scuba ant. The pI variant includes the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.Skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is identical to skeleton 1 except that it contains the M428L / N434S Xtend mutation. Skeleton 9 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as will be understood by those skilled in the art.In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 7B]The following shows the sequences of several useful 2+1 Fab2-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, without Fv and linker sequences. For example, on the "Fab-Fc side", "(VH domain / )" indicates that the C-terminus of the VH domain is bound at its position, which is the beginning of the CH1 domain of the heavy chain. Similarly, "VH-CH1-domain linker 1-scFv-domain linker 2 / " at the beginning of the "Fab-scFv-Fc side" indicates that these sequences are variant CH2-CH3 This indicates binding to the enumerated sequences, which are Fc domains. As discussed herein, the scFv domain can be in either the direction of the -VH-scFv linker-VL- or -VL-scFv linker-VH- as discussed herein. In addition, preferred domain linkers for “domain linker 2” are those enumerated in Figure 5 as “useful domain linkers”. Note that sequence identifiers are for the enumerated sequences only. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and has S364K / E357Q:L368D / K370S scuba riant, N208D / Q295E / N384D / Q418E / N421D on a strand having L368D / K370S scuba riant The pI variant and the E233P / L234V / L235A / G236del / S267K removal variant are included on both strands. The skeleton 2 is based on human IgG1 (356E / 358M allotype) and contains S364K:L368D / K370S scuba ant and N208D / Q295E / N384D / Q418E / N421D on the strand having L368D / K370S scuba ant. The pI variant includes the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.Skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is identical to skeleton 1 except that it contains the M428L / N434S Xtend mutation. Skeleton 9 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as will be understood by those skilled in the art.In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 7C]The following shows the sequences of several useful 2+1 Fab2-scFv-Fc bispecific antibody format heavy chain skeletons based on human IgG1, without Fv and linker sequences. For example, on the "Fab-Fc side", "(VH domain / )" indicates that the C-terminus of the VH domain is bound at its position, which is the beginning of the CH1 domain of the heavy chain. Similarly, "VH-CH1-domain linker 1-scFv-domain linker 2 / " at the beginning of the "Fab-scFv-Fc side" indicates that these sequences are variant CH2-CH3 This indicates binding to the enumerated sequences, which are Fc domains. As discussed herein, the scFv domain can be in either the direction of the -VH-scFv linker-VL- or -VL-scFv linker-VH- as discussed herein. In addition, preferred domain linkers for “domain linker 2” are those enumerated in Figure 5 as “useful domain linkers”. Note that sequence identifiers are for the enumerated sequences only. Skeleton 1 is based on human IgG1 (356E / 358M allotype) and has S364K / E357Q:L368D / K370S scuba riant, N208D / Q295E / N384D / Q418E / N421D on a strand having L368D / K370S scuba riant The pI variant and the E233P / L234V / L235A / G236del / S267K removal variant are included on both strands. The skeleton 2 is based on human IgG1 (356E / 358M allotype) and contains S364K:L368D / K370S scuba ant and N208D / Q295E / N384D / Q418E / N421D on the strand having L368D / K370S scuba ant. The pI variant includes the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 3 is based on human IgG1 (356E / 358M allotype) and includes the S364K:L368E / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368E / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands.Skeleton 4 is based on human IgG1 (356E / 358M allotype) and includes the D401K:K360E / Q362E / T411E scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the K360E / Q362E / T411E scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 5 is based on human IgG1 (356D / 358L allotype) and includes the S364K / E357Q:L368D / K370S scuba ant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand having the L368D / K370S scuba ant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Skeleton 6 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, the N208D / Q295E / N384D / Q418E / N421D pI variant on the strand containing the L368D / K370S scuba riant, the E233P / L234V / L235A / G236del / S267K removal variant on both strands, and the N297A variant on both strands. Skeleton 7 is identical to 6 except that the mutation is N297S. Skeleton 8 is identical to skeleton 1 except that it contains the M428L / N434S Xtend mutation. Skeleton 9 is based on human IgG1 (356E / 358M allotype) and includes the S364K / E357Q:L368D / K370S scuba riant, the P217R / P229R / N276K pI variant on the strand having the S364K / E357Q scuba riant, and the E233P / L234V / L235A / G236del / S267K removal variant on both strands. Each of these skeletons contains a sequence that is 90, 95, 98, and 99% identical (as defined herein) to the enumerated sequences and / or includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 additional amino acid substitutions (compared to the “parent” in the figure, which already includes some amino acid modifications compared to the parent human IgG1 (or IgG2 or IgG4 depending on the skeleton), as will be understood by those skilled in the art.In other words, the listed skeletons may include additional amino acid modifications (generally amino acid substitutions) in addition to the scuba riants, pI variants, and removal variants contained within the skeletons in this diagram. [Figure 8] This shows the "non-Fv" skeleton of the homologous light chain (i.e., constant light chain) used in the 1+1 Fab-scFv-Fc and 2+1 Fab2-scFv-Fc bispecific antibodies of the present invention. [Figure 9] The sequences of A) human claudin 18 isoform A2 (CLDN18.2), B) mouse claudin 18 isoform A2.1, C) mouse claudin 18 isoform A2.2, and D) cynomolgus monkey claudin 18 are shown to facilitate the development of cross-reactive antigen-binding domains for easier clinical development. [Figure 10] The variable heavy and light chains of exemplary anti-CLDN18.2 ABD used in the anti-CLDN 18.2× anti-CD3 bispecific antibody of the present invention are shown. CDRs are underlined. As is the case with any sequence described herein that includes a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. The H0 variable heavy chain and L0 variable light chain are mouse, while the H1 variable heavy chain, H2 variable heavy chain, and L1 variable light chain are humanized. [Figure 11] The sequence of the mouse anti-CLDN18.2 antibody with the removal variant (E233P / L234V / L235A / G236del / S267K, "IgG1_PVA_ / S267K") is shown. The CDR is underlined. As is the case with all sequences described herein that include CDRs, the precise identification of the CDR location may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. [Figure 12A]The following are exemplary anti-CD3 scFv sequences suitable for use with the bispecific antibodies of the present invention. The CDR is underlined, the scFv linker is double-underlined (in the sequence, the scFv linker is the positively charged scFv(GKPGS)4 linker (SEQ ID NO: 10), but as will be understood by those skilled in the art, this linker can be replaced by other linkers, some of which are uncharged or charged linkers, as shown in Figure 5), and a slash indicates the boundary of a variable domain. In addition, the nomenclature convention indicates the direction of the scFv from the N-terminus to the C-terminus. The scFv sequence is shown in both directions as shown. As is true for all sequences described herein that include a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. Furthermore, for all sequences shown in the figure, these VH and VL sequences can be used in either scFv format or Fab format. [Figure 12B] The following are exemplary anti-CD3 scFv sequences suitable for use with the bispecific antibodies of the present invention. The CDR is underlined, the scFv linker is double-underlined (in the sequence, the scFv linker is the positively charged scFv(GKPGS)4 linker (SEQ ID NO: 10), but as will be understood by those skilled in the art, this linker can be replaced by other linkers, some of which are uncharged or charged linkers, as shown in Figure 5), and a slash indicates the boundary of a variable domain. In addition, the nomenclature convention indicates the direction of the scFv from the N-terminus to the C-terminus. The scFv sequence is shown in both directions as shown. As is true for all sequences described herein that include a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. Furthermore, for all sequences shown in the figure, these VH and VL sequences can be used in either scFv format or Fab format. [Figure 12C] The following are exemplary anti-CD3 scFv sequences suitable for use with the bispecific antibodies of the present invention. The CDR is underlined, the scFv linker is double-underlined (in the sequence, the scFv linker is the positively charged scFv(GKPGS)4 linker (SEQ ID NO: 10), but as will be understood by those skilled in the art, this linker can be replaced by other linkers, some of which are uncharged or charged linkers, as shown in Figure 5), and a slash indicates the boundary of a variable domain. In addition, the nomenclature convention indicates the direction of the scFv from the N-terminus to the C-terminus. The scFv sequence is shown in both directions as shown. As is true for all sequences described herein that include a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. Furthermore, for all sequences shown in the figure, these VH and VL sequences can be used in either scFv format or Fab format. [Figure 12D]The following are exemplary anti-CD3 scFv sequences suitable for use with the bispecific antibodies of the present invention. The CDR is underlined, the scFv linker is double-underlined (in the sequence, the scFv linker is the positively charged scFv(GKPGS)4 linker (SEQ ID NO: 10), but as will be understood by those skilled in the art, this linker can be replaced by other linkers, some of which are uncharged or charged linkers, as shown in Figure 5), and a slash indicates the boundary of a variable domain. In addition, the nomenclature convention indicates the direction of the scFv from the N-terminus to the C-terminus. The scFv sequence is shown in both directions as shown. As is true for all sequences described herein that include a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. Furthermore, for all sequences shown in the figure, these VH and VL sequences can be used in either scFv format or Fab format. [Figure 12E] The following are exemplary anti-CD3 scFv sequences suitable for use with the bispecific antibodies of the present invention. The CDR is underlined, the scFv linker is double-underlined (in the sequence, the scFv linker is the positively charged scFv(GKPGS)4 linker (SEQ ID NO: 10), but as will be understood by those skilled in the art, this linker can be replaced by other linkers, some of which are uncharged or charged linkers, as shown in Figure 5), and a slash indicates the boundary of a variable domain. In addition, the nomenclature convention indicates the direction of the scFv from the N-terminus to the C-terminus. The scFv sequence is shown in both directions as shown. As is true for all sequences described herein that include a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. Furthermore, for all sequences shown in the figure, these VH and VL sequences can be used in either scFv format or Fab format. [Figure 12F] The following are exemplary anti-CD3 scFv sequences suitable for use with the bispecific antibodies of the present invention. The CDR is underlined, the scFv linker is double-underlined (in the sequence, the scFv linker is the positively charged scFv(GKPGS)4 linker (SEQ ID NO: 10), but as will be understood by those skilled in the art, this linker can be replaced by other linkers, some of which are uncharged or charged linkers, as shown in Figure 5), and a slash indicates the boundary of a variable domain. In addition, the nomenclature convention indicates the direction of the scFv from the N-terminus to the C-terminus. The scFv sequence is shown in both directions as shown. As is true for all sequences described herein that include a CDR, the precise identification of the CDR position may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. Furthermore, for all sequences shown in the figure, these VH and VL sequences can be used in either scFv format or Fab format. [Figure 13A] Two useful formats of the present invention are shown. Figure 13A shows the “1+1 Fab-scFv-Fc” format, where the first Fab arm is coupled to CLDN18.2 and the second scFv arm is coupled to CD3. Figure 13B shows the “2+1 Fab2-scFv-Fc” format, having a first Fab arm and a second Fab-scFv arm coupled to CLDN18.2, where the Fab is coupled to CLDN18.2 and the scFv is coupled to CD3. It should be noted that additional formats may also be used, as commonly shown in Figure 42 of this specification. [Figure 13B]Two useful formats of the present invention are shown. Figure 13A shows the “1+1 Fab-scFv-Fc” format, where the first Fab arm is coupled to CLDN18.2 and the second scFv arm is coupled to CD3. Figure 13B shows the “2+1 Fab2-scFv-Fc” format, having a first Fab arm and a second Fab-scFv arm coupled to CLDN18.2, where the Fab is coupled to CLDN18.2 and the scFv is coupled to CD3. It should be noted that additional formats may also be used, as commonly shown in Figure 42 of this specification. [Figure 14] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 1+1 Fab-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using the Fab variable region 1 and scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15A] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15B]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15C] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15D]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15E] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15F]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15G] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15H]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15I] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15J]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15K] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15L]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15M] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15N]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15O] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15P]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15Q] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15R]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15S] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15T]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15U] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15V]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15W] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15X]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15Y] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15Z]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AA] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AB]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AC] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AD]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AE] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AF]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AG] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AH]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AI] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AJ]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AK] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AL]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AM] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AN]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AO] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AP]The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 15AQ] The amino acid sequences of prototype anti-CLDN18.2× anti-CD3 bispecific antibodies in 2+1 Fab2-scFv-Fc format (based on the mouse CLDN18.2 ABD shown in Figure 10) are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 16] The amino acid sequences of control anti-RSV × high CD3 bispecific antibodies in 1+1 Fab-scFv-Fc format are shown. The antibodies are named using the Fab variable region 1 and scFv variable region 2, separated by a dash. CDR is underlined, and a slash indicates the boundary of the variable region. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 17A]A) KP-4 cells and B) NUGC-4 cells are shown to be bound to prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies containing mouse CLDN18.2 ABD XENP24645 (1+1 Fab-scFv-Fc containing high CD3), XENP24646 (1+1 Fab-scFv-Fc containing high-medium CD3 #1), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3), XENP24648 (2+1 Fab2-scFv-Fc containing high-medium CD3 #1), and XENP24649 (2+1 Fab2-scFv-Fc containing medium CD3). Controls included anti-RSV × anti-CD3 bispecific antibody, cells only, and secondary antibody only. The data show that the prototype anti-CLDN18.2×anti-CD3 bsAb binds to NUGC-4 cells in a dose-dependent manner and is close to baseline binding to KP-4 cells at all concentrations tested. Notably, bsAbs in the "2+1 Fab2-scFv-Fc" format (i.e., XENP24647, XENP24648, and XENP24949) bound to NUGC-4 cells much more strongly than bsAbs in the "1+1 Fab2-scFv-Fc" format (i.e., XENP24645 and XENP24646), due to the extra avidity carried by the "2+1 Fab2-scFv-Fc" format. [Figure 17B]A) KP-4 cells and B) NUGC-4 cells are shown to be bound to prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies containing mouse CLDN18.2 ABD XENP24645 (1+1 Fab-scFv-Fc containing high CD3), XENP24646 (1+1 Fab-scFv-Fc containing high-medium CD3 #1), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3), XENP24648 (2+1 Fab2-scFv-Fc containing high-medium CD3 #1), and XENP24649 (2+1 Fab2-scFv-Fc containing medium CD3). Controls included anti-RSV × anti-CD3 bispecific antibody, cells only, and secondary antibody only. The data show that the prototype anti-CLDN18.2×anti-CD3 bsAb binds to NUGC-4 cells in a dose-dependent manner and is close to baseline binding to KP-4 cells at all concentrations tested. Notably, bsAbs in the "2+1 Fab2-scFv-Fc" format (i.e., XENP24647, XENP24648, and XENP24949) bound to NUGC-4 cells much more strongly than bsAbs in the "1+1 Fab2-scFv-Fc" format (i.e., XENP24645 and XENP24646), due to the extra avidity carried by the "2+1 Fab2-scFv-Fc" format. [Figure 18A] This study demonstrates the induction of RTCCs on A) KP-4 cells and B) NUGC-4 cells by prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies containing mouse CLDN18.2 ABD XENP24645 (1+1 Fab-scFv-Fc with high CD3-), XENP24646 (1+1 Fab-scFv-Fc with high-medium CD3-#1), XENP24647 (2+1 Fab2-scFv-Fc with high CD3-), XENP24648 (2+1 Fab2-scFv-Fc with high-medium CD3-#1), and XENP24649 (2+1 Fab2-scFv-Fc with medium CD3-). The data show that the prototype anti-CLDN18.2×anti-CD3 bsAb dose-dependently induced RTCC on NUGC-4 cells, but did not induce RTCC on KP-4 cells. [Figure 18B]This study demonstrates the induction of RTCCs on A) KP-4 cells and B) NUGC-4 cells by prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies containing mouse CLDN18.2 ABD XENP24645 (1+1 Fab-scFv-Fc with high CD3-), XENP24646 (1+1 Fab-scFv-Fc with high-medium CD3-#1), XENP24647 (2+1 Fab2-scFv-Fc with high CD3-), XENP24648 (2+1 Fab2-scFv-Fc with high-medium CD3-#1), and XENP24649 (2+1 Fab2-scFv-Fc with medium CD3-). The data show that the prototype anti-CLDN18.2×anti-CD3 bsAb dose-dependently induced RTCC on NUGC-4 cells, but did not induce RTCC on KP-4 cells. [Figure 19A] The expression levels on A) NUGC-4 cells and B) SNU-601 cells, as determined by flow cytometry, are shown. The data indicate that SNU-601 expresses more CLDN18.2 than NUGC-4. [Figure 19B] The expression levels on A) NUGC-4 cells and B) SNU-601 cells, as determined by flow cytometry, are shown. The data indicate that SNU-601 expresses more CLDN18.2 than NUGC-4. [Figure 20A]This shows the binding of prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies containing mouse CLDN18.2 ABD XENP24645 (1+1 Fab-scFv-Fc containing high CD3), XENP24646 (1+1 Fab-scFv-Fc containing high-medium CD3 #1), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3), and XENP24649 (2+1 Fab2-scFv-Fc containing medium CD3) to NUGC-4 cells and SNU-601 cells. Controls included anti-RSV × anti-CD3 bispecific antibody, cells only, and secondary antibody only. The data show that each bsAb binds to both NUGC-4 and SNU-601 in a dose-dependent manner, with higher maximum binding to SNU-601 cells than to NUGC-4, which is consistent with the respective CLDN18.2 expression levels on each cell line. [Figure 20B] This shows the binding of prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies containing mouse CLDN18.2 ABD XENP24645 (1+1 Fab-scFv-Fc containing high CD3), XENP24646 (1+1 Fab-scFv-Fc containing high-medium CD3 #1), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3), and XENP24649 (2+1 Fab2-scFv-Fc containing medium CD3) to NUGC-4 cells and SNU-601 cells. Controls included anti-RSV × anti-CD3 bispecific antibody, cells only, and secondary antibody only. The data show that each bsAb binds to both NUGC-4 and SNU-601 in a dose-dependent manner, with higher maximum binding to SNU-601 cells than to NUGC-4, which is consistent with the respective CLDN18.2 expression levels on each cell line. [Figure 21A]The following shows the induction of RTCC (indicated by a decrease in viable target cells) on A) NUGC-4 cells and B) SNU-601 cells after 24-hour incubation with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies having mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by a decrease in viable cells. [Figure 21B] The following shows the induction of RTCC (indicated by a decrease in viable target cells) on A) NUGC-4 cells and B) SNU-601 cells after 24-hour incubation with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies having mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by a decrease in viable cells. [Figure 22A]The following shows the induction of RTCC (indicated by an increase in dead target cells) on A) NUGC-4 cells and B) SNU-601 cells after 24-hour incubation with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies having mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by an increase in dead / dying cells. [Figure 22B] The following shows the induction of RTCC (indicated by an increase in dead target cells) on A) NUGC-4 cells and B) SNU-601 cells after 24-hour incubation with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies having mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by an increase in dead / dying cells. [Figure 23A]The following shows the induction of RTCC (indicated by a decrease in viable target cells) on A) NUGC-4 cells and B) SNU-601 cells after 48-hour incubation with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by a decrease in viable cells. [Figure 23B] The following shows the induction of RTCC (indicated by a decrease in viable target cells) on A) NUGC-4 cells and B) SNU-601 cells after 48-hour incubation with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by a decrease in viable cells. [Figure 24A]This document shows the induction of RTCCs (indicated by an increase in dead target cells) on A) NUGC-4 cells and B) SNU-601 cells after 48 hours of incubation with the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies having mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc containing CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc containing CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc containing CD3-high #1), and XENP24649 (2+1 Fab2-scFv-Fc containing CD3-medium #1) in human PBMCs (effector-to-target cell ratio of 20:1) and mouse CLDN18.2 ABD. Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by an increase in dead / dying cells. [Figure 24B] This document shows the induction of RTCCs (indicated by an increase in dead target cells) on A) NUGC-4 cells and B) SNU-601 cells after 48 hours of incubation with the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies having mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc containing CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc containing CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc containing CD3-high #1), and XENP24649 (2+1 Fab2-scFv-Fc containing CD3-medium #1) in human PBMCs (effector-to-target cell ratio of 20:1) and mouse CLDN18.2 ABD. Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data show that bsAb enhances the death of target cells (i.e., NUGC-4 and SNU-601 cells), as indicated by an increase in dead / dying cells. [Figure 25A]The percentages of CD4+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of NUGC-4 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 25B] The percentages of CD4+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of NUGC-4 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 25C]The percentages of CD4+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of NUGC-4 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 26A] The percentages of CD8+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of NUGC-4 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 26B]The percentages of CD8+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of NUGC-4 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 26C] The percentages of CD8+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of NUGC-4 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 27A]The percentages of CD4+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of SNU-601 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 27B] The percentages of CD4+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of SNU-601 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 27C]The percentages of CD4+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of SNU-601 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high #1), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 28A] The percentages of CD8+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of SNU-601 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 28B]The percentages of CD8+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of SNU-601 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 28C] The percentages of CD8+ T cells expressing A) CD69, B) CD25, and C) CD107a are shown after 48-hour incubation of SNU-601 cells with human PBMCs (effector-to-target cell ratio of 20:1) and the following prototype anti-CLDN18.2 × anti-CD3 bispecific antibodies with mouse CLDN18.2 ABD: XENP24645 (1+1 Fab-scFv-Fc including CD3-high), XENP24646 (1+1 Fab-scFv-Fc including CD3-high-medium #1), XENP24647 (2+1 Fab2-scFv-Fc including CD3-high), and XENP24649 (2+1 Fab2-scFv-Fc including CD3-medium). Controls included anti-RSV × anti-CD3 bispecific antibody, target cells only, and target cells and effector cells only. The data showed a trend consistent with RTCC, namely, higher affinity CD3 binding and / or divalent CLDN18.2 binding enhances T cell activation and degranulation. [Figure 29]The sequence of the anti-CLDN18.2 antibody having a humanized variable region with a removed variant (E233P / L234V / L235A / G236del / S267K, "IgG1_PVA_ / S267K") is shown. The CDR is underlined. As is the case with all sequences described herein that include CDRs, the precise identification of the CDR location may vary slightly depending on the numbering used, as shown in Table 1, and therefore, not only the underlined CDRs but also CDRs contained within the VH and VL domains using other numbering systems are included herein. [Figure 30] This demonstrates germline identity between humanized CLDN18.2 ABD and mouse CLDN18.2 ABD. [Figure 31A] The amino acid sequences of anti-CLDN 18.2× anti-CD3 bispecific antibodies having a humanized variable region in a 1+1 Fab-scFv-Fc format are shown. The antibodies are named using the Fab variable region 1 and scFv variable region 2, separated by a dash. CDR is underlined, and a slash indicates the boundary of the variable region. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 31B] The amino acid sequences of anti-CLDN 18.2× anti-CD3 bispecific antibodies having a humanized variable region in a 1+1 Fab-scFv-Fc format are shown. The antibodies are named using the Fab variable region 1 and scFv variable region 2, separated by a dash. CDR is underlined, and a slash indicates the boundary of the variable region. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 31C]The amino acid sequences of anti-CLDN 18.2× anti-CD3 bispecific antibodies having a humanized variable region in a 1+1 Fab-scFv-Fc format are shown. The antibodies are named using the Fab variable region 1 and scFv variable region 2, separated by a dash. CDR is underlined, and a slash indicates the boundary of the variable region. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 32A] The amino acid sequences of anti-CLDN18.2× anti-CD3 bispecific antibodies having humanized variable regions in a 2+1 Fab2-scFv-Fc format are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 32B] The amino acid sequences of anti-CLDN18.2× anti-CD3 bispecific antibodies having humanized variable regions in a 2+1 Fab2-scFv-Fc format are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 32C]The amino acid sequences of anti-CLDN18.2× anti-CD3 bispecific antibodies having humanized variable regions in a 2+1 Fab2-scFv-Fc format are shown. The antibodies are named using Fab variable region 1 and Fab-scFv variable region 2, separated by dashes. CDR is underlined, and slashes indicate the boundaries of the variable regions. The scFv domain has a VH-scFv linker-VL orientation (from N-terminus to C-terminus), but this can be reversed. In addition, each sequence outlined herein may contain or exclude the M428L / N434S variant in one or preferably both Fc domains, which results in a longer half-life in serum. [Figure 33A] The expression levels on A) SNU-601 cells and B) SNU-601(2E4) cells (enriched for the CLDN18.2-expressing population), as determined by flow cytometry, are shown. The data indicate that SNU-601(2E4) contains a substantially higher population of CLDN18.2+ cells. The experiments in this section are performed using SNU-601(2E4) cells. [Figure 33B] The expression levels on A) SNU-601 cells and B) SNU-601(2E4) cells (enriched for the CLDN18.2-expressing population), as determined by flow cytometry, are shown. The data indicate that SNU-601(2E4) contains a substantially higher population of CLDN18.2+ cells. The experiments in this section are performed using SNU-601(2E4) cells. [Figure 34]The following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-high medium #1 1+1 Fab-scFv-Fc), XENP29475 (H2L1 CLDN18.2 ABD, CD3-high medium #1 1+1 Fab-scFv-Fc), XENP29476 (H1L1 CLDN18.2 ABD, CD3-high 2+1 This shows the binding of Fab2-scFv-Fc), XENP29477 (2+1 Fab2-scFv-Fc containing H2L1 CLDN18.2 ABD and CD3-high), XENP29478 (2+1 Fab2-scFv-Fc containing H1L1 CLDN18.2 ABD and CD3-high #1), and XENP29479 (2+1 Fab2-scFv-Fc containing H2L1 CLDN18.2 ABD and CD3-high #1) to SNU-601(2E4) cells. The controls used were XENP24644 (H0L0 CLDN18.2, bivalent mAb), XENP29470 (H1L1 CLDN18.2, bivalent mAb), XENP29471 (H2L1 CLDN18.2, bivalent mAb), XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells only, and secondary antibodies only. The data showed that bsAbs with humanized CLDN18.2 ABD had similar binding to SNU-601(2E4) cells as bsAbs with mouse CLDN18.2 ABD, indicating that humanization preserved the binding efficacy of bsAbs. Notably, bsAb in the "2+1 Fab2-scFv-Fc" format showed similar binding to divalent anti-CLDN18.2 mAb.In addition, the data show that bsAbs based on H1L1 humanized variants (e.g., XENP29472, XENP29474, XENP28476, and XENP29478) maintained better binding than bsAbs based on H2L1 humanized variants (e.g., XENP29473, XENP29475, XENP29477, and XENP29479). [Figure 35A]This demonstrates the induction of RTCC on SNU-601(2E4) cells, which is shown by A) a decrease in the number of CFSE+ SNU-601(2E4) cells, B) the percentage of CFSE+ SNU-601(2E4) cells stained with Zombie Aqua, and C) human PBMCs (effector vs. target cell ratio of 10:1), as well as Zombie Aqua MFI on CFSE+ SNU-601(2E4) cells after 24-hour incubation with the following anti-CLDN18.2 × anti-CD3 bispecific antibodies against humanized CLDN18.2 ABD: XENP29472(H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473(H2L1 CLDN18.2 ABD, CD3-High including 1+1 Fab-scFv-Fc), XENP29474(H1L1 CLDN18.2 ABD, CD3-High Medium #1 including 1+1 Fab-scFv-Fc), XENP29475(H2L1 CLDN18.2 ABD, CD3-High Medium #1 including 1+1 Fab-scFv-Fc), XENP29476(H1L1 CLDN18.2 ABD, CD3-High including 2+1 Fab2-scFv-Fc), XENP29477(H2L1 CLDN18.2 ABD, CD3-High including 2+1 Fab2-scFv-Fc), XENP29478(H1L1 CLDN18.2 ABD, CD3-High Medium #1 including 2+1 The subjects were Fab2-scFv-Fc and XENP29479 (2+1 Fab2-scFv-Fc containing H2L1 CLDN18.2 ABD and CD3-high #1). The controls used were XENP24645 (1+1 Fab2-scFv-Fc containing CD3-high), XENP24647 (2+1 Fab2-scFv-Fc containing CD3-high), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved the induction of RTCC by bsAb with humanized CLDN18.2 ABD. [Figure 35B]This demonstrates the induction of RTCC on SNU-601(2E4) cells, which is shown by A) a decrease in the number of CFSE+ SNU-601(2E4) cells, B) the percentage of CFSE+ SNU-601(2E4) cells stained with Zombie Aqua, and C) human PBMCs (effector vs. target cell ratio of 10:1), as well as Zombie Aqua MFI on CFSE+ SNU-601(2E4) cells after 24-hour incubation with the following anti-CLDN18.2 × anti-CD3 bispecific antibodies against humanized CLDN18.2 ABD: XENP29472(H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473(H2L1 CLDN18.2 ABD, CD3-High including 1+1 Fab-scFv-Fc), XENP29474(H1L1 CLDN18.2 ABD, CD3-High Medium #1 including 1+1 Fab-scFv-Fc), XENP29475(H2L1 CLDN18.2 ABD, CD3-High Medium #1 including 1+1 Fab-scFv-Fc), XENP29476(H1L1 CLDN18.2 ABD, CD3-High including 2+1 Fab2-scFv-Fc), XENP29477(H2L1 CLDN18.2 ABD, CD3-High including 2+1 Fab2-scFv-Fc), XENP29478(H1L1 CLDN18.2 ABD, CD3-High Medium #1 including 2+1 The subjects were Fab2-scFv-Fc and XENP29479 (2+1 Fab2-scFv-Fc containing H2L1 CLDN18.2 ABD and CD3-high #1). The controls used were XENP24645 (1+1 Fab2-scFv-Fc containing CD3-high), XENP24647 (2+1 Fab2-scFv-Fc containing CD3-high), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved the induction of RTCC by bsAb with humanized CLDN18.2 ABD. [Figure 35C]This demonstrates the induction of RTCC on SNU-601(2E4) cells, which is shown by A) a decrease in the number of CFSE+ SNU-601(2E4) cells, B) the percentage of CFSE+ SNU-601(2E4) cells stained with Zombie Aqua, and C) human PBMCs (effector vs. target cell ratio of 10:1), as well as Zombie Aqua MFI on CFSE+ SNU-601(2E4) cells after 24-hour incubation with the following anti-CLDN18.2 × anti-CD3 bispecific antibodies against humanized CLDN18.2 ABD: XENP29472(H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473(H2L1 CLDN18.2 ABD, CD3-High including 1+1 Fab-scFv-Fc), XENP29474(H1L1 CLDN18.2 ABD, CD3-High Medium #1 including 1+1 Fab-scFv-Fc), XENP29475(H2L1 CLDN18.2 ABD, CD3-High Medium #1 including 1+1 Fab-scFv-Fc), XENP29476(H1L1 CLDN18.2 ABD, CD3-High including 2+1 Fab2-scFv-Fc), XENP29477(H2L1 CLDN18.2 ABD, CD3-High including 2+1 Fab2-scFv-Fc), XENP29478(H1L1 CLDN18.2 ABD, CD3-High Medium #1 including 2+1 The subjects were Fab2-scFv-Fc and XENP29479 (2+1 Fab2-scFv-Fc containing H2L1 CLDN18.2 ABD and CD3-high #1). The controls used were XENP24645 (1+1 Fab2-scFv-Fc containing CD3-high), XENP24647 (2+1 Fab2-scFv-Fc containing CD3-high), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved the induction of RTCC by bsAb with humanized CLDN18.2 ABD. [Figure 36A]The following describes the activation of CD4+ T cells after 24-hour incubation of SNU-601(2E4) with human PBMCs (10:1 effector-to-target cell ratio) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD69 MFI on CD4+ T cells, and B) the percentage of CD4+ T cells expressing CD69]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved T cell activation by bsAb with humanized CLDN18.2 ABD. [Figure 36B]The following describes the activation of CD4+ T cells after 24-hour incubation of SNU-601(2E4) with human PBMCs (10:1 effector-to-target cell ratio) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD69 MFI on CD4+ T cells, and B) the percentage of CD4+ T cells expressing CD69]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved T cell activation by bsAb with humanized CLDN18.2 ABD. [Figure 37A]The following describes the degranulation of CD4+ T cells after 24 hours of incubation of SNU-601(2E4) with human PBMCs (10:1 effector-to-target cell ratio) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD107a MFI on CD4+ T cells, and B) the percentage of CD4+ T cells expressing CD107a]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells alone, and secondary antibodies alone. [Figure 37B]The following describes the degranulation of CD4+ T cells after 24 hours of incubation of SNU-601(2E4) with human PBMCs (10:1 effector-to-target cell ratio) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD107a MFI on CD4+ T cells, and B) the percentage of CD4+ T cells expressing CD107a]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells alone, and secondary antibodies alone. [Figure 38A]The following describes the activation of CD8+ T cells after 24-hour incubation of SNU-601(2E4) with human PBMCs (effector-to-target cell ratio of 10:1) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD69 MFI on CD8+ T cells, and B) the percentage of CD8+ T cells expressing CD69]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved T cell activation by bsAb with humanized CLDN18.2 ABD. [Figure 38B]The following describes the activation of CD8+ T cells after 24-hour incubation of SNU-601(2E4) with human PBMCs (effector-to-target cell ratio of 10:1) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD69 MFI on CD8+ T cells, and B) the percentage of CD8+ T cells expressing CD69]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells only, and secondary antibodies only. Consistent with the binding data, humanization preserved T cell activation by bsAb with humanized CLDN18.2 ABD. [Figure 39A]The following describes the degranulation of CD8+ T cells after 24 hours of incubation of SNU-601(2E4) with human PBMCs (effector-to-target cell ratio of 10:1) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD107a MFI on CD8+ T cells, and B) the percentage of CD8+ T cells expressing CD107a]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells alone, and secondary antibodies alone. [Figure 39B]The following describes the degranulation of CD8+ T cells after 24 hours of incubation of SNU-601(2E4) with human PBMCs (effector-to-target cell ratio of 10:1) and the following anti-CLDN18.2 × anti-CD3 bispecific antibodies containing humanized CLDN18.2 ABD [A) CD107a MFI on CD8+ T cells, and B) the percentage of CD8+ T cells expressing CD107a]: XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium 1+1 #1) Fab-scFv-Fc), XENP29475 (1+1 Fab-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1), XENP29476 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High), XENP29477 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High), XENP29478 (2+1 Fab2-scFv-Fc including H1L1 CLDN18.2 ABD, CD3-High Middle #1), and XENP29479 (2+1 Fab2-scFv-Fc including H2L1 CLDN18.2 ABD, CD3-High Middle #1). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-), XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-), cells alone, and secondary antibodies alone. [Figure 40A]The following antibodies, possessing human PBMCs (10:1 effector-to-target cell ratio) and humanized CLDN18.2 ABD, were used to induce 24-hour incubation of SNU-601(2E4) cells, resulting in A) IFNγ and B) TNFα secretion. These antibodies were used against SNU-601(2E4) cells. The antibodies used were XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium high 1+1 Fab-scFv-Fc), and XENP29475 (H2L1 CLDN18.2 ABD, CD3-High Middle #1 included 1+1 Fab-scFv-Fc), XENP29476(H1L1 CLDN18.2 ABD, CD3-High included 2+1 Fab2-scFv-Fc), XENP29477(H2L1 CLDN18.2 ABD, CD3-High included 2+1 Fab2-scFv-Fc), XENP29478(H1L1 CLDN18.2 ABD, CD3-High Middle #1 included 2+1 Fab2-scFv-Fc), and XENP29479(H2L1 CLDN18.2 ABD, CD3-High Middle #1 included 2+1 Fab2-scFv-Fc). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-) and XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-). Consistent with the binding data, humanization preserved the induction of cytokine secretion by bsAb with humanized CLDN18.2 ABD. [Figure 40B]The following antibodies, possessing human PBMCs (10:1 effector-to-target cell ratio) and humanized CLDN18.2 ABD, were used to induce 24-hour incubation of SNU-601(2E4) cells, resulting in A) IFNγ and B) TNFα secretion. These antibodies were used against SNU-601(2E4) cells. The antibodies used were XENP29472 (H1L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29473 (H2L1 CLDN18.2 ABD, CD3-high 1+1 Fab-scFv-Fc), XENP29474 (H1L1 CLDN18.2 ABD, CD3-medium high 1+1 Fab-scFv-Fc), and XENP29475 (H2L1 CLDN18.2 ABD, CD3-High Middle #1 included 1+1 Fab-scFv-Fc), XENP29476(H1L1 CLDN18.2 ABD, CD3-High included 2+1 Fab2-scFv-Fc), XENP29477(H2L1 CLDN18.2 ABD, CD3-High included 2+1 Fab2-scFv-Fc), XENP29478(H1L1 CLDN18.2 ABD, CD3-High Middle #1 included 2+1 Fab2-scFv-Fc), and XENP29479(H2L1 CLDN18.2 ABD, CD3-High Middle #1 included 2+1 Fab2-scFv-Fc). The controls used were XENP24645 (1+1 Fab-scFv-Fc containing high CD3-) and XENP24647 (2+1 Fab2-scFv-Fc containing high CD3-). Consistent with the binding data, humanization preserved the induction of cytokine secretion by bsAb with humanized CLDN18.2 ABD. [Figure 41A]The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41B] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41C] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41D]The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41E] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41F] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41G]The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41H] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41I] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41J]The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41K] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41L] The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 41M]The following sequences represent several useful 2+1 Fab2-scFv-Fc bispecific antibody formats based on human IgG1, containing six different anti-CD3 ABD scFvs in both directions, as well as sequences containing and excluding the "XTEND®" FcRn variant 428L / 434S (sometimes referred to as the "LS" variant), but also sequences that exclude ABDs of other antigens. Specifically, chain 1 of each set is a -CH1-hinge-CH2-CH3 sequence ("Fab-Fc side"), to which a VH1 sequence can be appended (e.g., the C-terminus of VH1 is ligated with a slash " / ") to form the complete heavy chain VH1-CH1-hinge-CH2-CH3. Chain 2 of each set is anti-CD3 scFv-domain linker-CH2-CH3, to which VH1-CH1-optional domain linker- can be added (for example, by ligating the C-terminus of the domain linker with a slash " / ") to form the complete chain VH1-CH1-domain linker-scFv-domain linker-CH2-CH3. In this embodiment, scFv can be in either direction so that the complete chain is either VH1-CH1-domain linker-VH2-scFv linker-VL2-domain linker-CH2-CH3 or VH1-CH1-domain linker-VL2-scFv linker-VH2-domain linker-CH2-CH3 (note that the sequence in Figure 41 includes both options). Chain 3 is an LC domain to which VL1 can be added to form VL1-CL. CDRs are underlined, linkers are double underlined, and domain boundaries are indicated with slashes. It should be noted that all of the strand 2 sequences include the sequence GGGGSGGGGSKTHTCPPCP (SEQ ID NO: 29), which is a “mobile half-hinge” domain linker between the C-terminus of the scFv and the N-terminus of the CH2 domain. However, this linker can be replaced in any of the strand 2 sequences in Figure 41 with any of the “useful domain linkers” in Figure 5, and in some embodiments, SEQ ID NO: 29 is replaced with SEQ ID NO: 28, the “full-hinge C220S variant.” Each of these sequences includes preferred skew, pI, and elimination variants. [Figure 42A] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42B] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42C] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42D] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42E] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42F]Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42G] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42H] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42I] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Figure 42J] Several formats of the present invention are shown. First, there is the 1+1 Fab-scFv-Fc format having first and second anti-antigen binding domains. In addition, mAb-Fv, mAb-scFv, central scFv, central Fv, one-armed central scFv, one-scFv-mAb, scFv-mAb, and dual scFv are all shown. For all scFv domains shown, they can be either N-terminal to C-terminal variable heavy chain-(optional linker)-variable light chain, or vice versa. In addition, in the case of one-armed scFv-mAb, the scFv can bind to either the N-terminus of the heavy chain monomer or the N-terminus of the light chain. [Modes for carrying out the invention]

[0026] A. Overview Anti-bispecific antibodies that co-link CD3 to tumor antigen targets are used to redirect T cells to attack and lyse targeted tumor cells. Examples include the BiTE® and DART formats, which monovalently link CD3 to tumor antigens. While CD3-targeted approaches show considerable promise, a common side effect of such therapies is the production of associated cytokines, often leading to toxic cytokine release syndrome. Because the anti-CD3 binding domain of bispecific antibodies links to all T cells, a highly cytokine-producing CD4 T cell subset is recruited. Furthermore, this CD4 T cell subset includes regulatory T cells, and their recruitment and proliferation can lead to immunosuppression and negatively impact long-term tumor suppression. In addition, these formats do not contain an Fc domain, indicating a very short serum half-life in patients.

[0027] In particular, various formats of anti-CD3, anti-CLDN18.2 bispecific antibodies, such as those shown in Figures 13 and 42, are provided herein. These bispecific antibodies are useful in the treatment of cancer, especially cancers such as gastric cancer, esophageal cancer, and pancreatic cancer. Such antibodies are used to direct CD3+ effector T cells to CLDN18.2+ tumors, thereby enabling CD3+ effector T cells to attack and lyse the CLDN18.2+ tumors.

[0028] In addition, the present invention provides bispecific antibodies having different binding affinities to human CD3 that can modify or reduce the potential side effects of anti-CD3 therapy. Specifically, in some embodiments, the present invention provides antibody constructs comprising an anti-CD3 antigen-binding domain that is a “strong” or “high affinity” conjugate to CD3 (for example, one example being a heavy-chain and light-chain variable domain indicated as H1.30_L1.47 (optionally, including a charged linker as appropriate)) and also binds to CLDN18.2. In other embodiments, the present invention provides antibody constructs comprising an anti-CD3 antigen-binding domain that is a “light” or “low affinity” conjugate to CD3. Additional embodiments provide antibody constructs comprising an anti-CD3 antigen-binding domain having an intermediate or “medium” affinity to CD3 that also binds to CD38. While a very large number of anti-CD3 antigen-binding domains (ABDs) can be used, particularly useful embodiments utilize six different anti-CD3 ABDs, which can be used in two scFv directions as described herein. Affinity is generally measured using a Biacore assay.

[0029] It should be understood that the “high, medium, and low” anti-CD3 sequences of the present invention can be used in various heterodimerization formats shown in the figures. Generally, due to the potential side effects of T cell recruitment, preferred embodiments utilize a format that binds to CD3 only in a monovalent state, as shown in Figures 13A and 13B, where the format shown herein is CD3 ABD, as fully described herein, where scFv is. In contrast, the bispecific antibodies of the present invention can bind to CLDN18.2 in either a monovalent (e.g., Figure 13A) or a bivalent (e.g., Figure 13B) state.

[0030] Accordingly, in one embodiment, heterodimer antibodies that bind to two different antigens are provided herein, for example, the antibodies are “bispecific” in that they bind to two different target antigens described herein, e.g., CD3 and CLDN18.2. These heterodimer antibodies can bind to these target antigens in either a monovalent (e.g., having a single antigen-binding domain such as a pair of variable heavy-chain and variable light-chain domains) or a bivalent (having two antigen-binding domains, each independently binding to the antigen). The heterodimer antibodies provided herein are based on the use of different monomers containing amino acid substitutions that “distort” the formation of heterodimers on homodimers, as will be outlined more fully below, and are linked to “pI variants” that enable the simple purification of heterodimers detached from homodimers, as will be outlined similarly below. The heterodimer bispecific antibodies provided generally rely on the use of manipulated or variant Fc domains that can self-assemble in producing cells to produce heterodimer proteins, and on methods for generating and purifying such heterodimer proteins.

[0031] C. Nomenclature The bispecific antibodies of the present invention are listed in several different formats. Each polypeptide is assigned a unique "XENP" number, although, as understood in the art, longer sequences may contain shorter sequences. For example, in the 1+1 Fab-scFv-Fc format of a given sequence, the heavy chain of the scFv side monomer will have a first XENP number, while the scFv domain will have a different XENP number. Since some molecules have three polypeptides, the XENP numbers are used as names along with the components. Thus, the molecule XENP29472 in bottle-opener format contains three sequences: XENP29472 chain 1, XENP29472 chain 2, and XENP29472 chain 3 (Figure 31A). These XENP numbers are found in the sequence listing and identifiers and are used in the figures. In addition, one molecule containing three components generates multiple sequence identifiers. For example, the Fab monomer list has three CDRs: the full-length sequence, the variable heavy chain sequence, and the variable heavy chain sequence. The light chain has three CDRs: a full-length sequence, a variable light chain sequence, and a variable light chain sequence. The scFv-Fc domain has a full-length sequence, an scFv sequence, a variable light chain sequence, three light chain CDRs, an scFv linker, a variable heavy chain sequence, and three heavy chain CDRs. Note that all molecules in this specification that have an scFv domain use a single charged scFv linker (+H), although others may be used. In addition, the nomenclature for specific variable domains uses the format of type "Hx.xx_Ly.yy", where the number is a unique identifier for a particular variable chain sequence. Thus, the variable domain on the Fab side of XENP29472 is "H1L1", which indicates that the variable heavy chain domain H1 is combined with the light chain domain L1. When these sequences are used as scFv, the name "H1L1" indicates that the variable heavy chain domain H1 is combined with the light chain domain L1 and is oriented in the VH-linker-VL direction from the N-terminus to the C-terminus. This molecule, which has the same sequence of heavy and light chain variable domains but in reverse order, is named "L1H1". Similarly, as is evident from the sequence listing and figures, different constructs can "mix and adapt" the heavy and light chains.

[0032] D. Definition To facilitate a more complete understanding of this application, several definitions are provided below. Such definitions are intended to encompass grammatical equivalents.

[0033] In this specification, “removal” means a reduction or elimination of activity. Therefore, for example, “cleaving the FcγR bond” means that the Fc region amino acid variant has less than 50% of the initiation bond compared to the Fc region without the specific variant, and preferably the activity is lost by more than 70-80-90-95-98%, and generally the activity is below a detectable binding level in Biacore, SPR, or BLI assays. The compounds particularly used for FcγR bond removal are shown in Figure 3, and these are generally attached to both monomers.

[0034] As used herein, “ADCC” or “antibody-dependent cell-mediated cytotoxicity” refers to a cell-mediated response in which nonspecific cytotoxic cells expressing FcγR recognize a bound antibody on target cells, subsequently causing lysis of the target cells. ADCC correlates with binding to FcγRIIIa, and increased binding to FcγRIIIa results in increased ADCC activity.

[0035] As used herein, "ADCP" or antibody-dependent cell-mediated phagocytosis refers to a cell-mediated response in which nonspecific phagocytic cells expressing FcγR recognize a bound antibody on a target cell, subsequently triggering phagocytosis of the target cell.

[0036] In this specification, “antigen-binding domain” or “ABD” means a set of six complementary determining regions (CDRs) that, when present as part of a polypeptide sequence, specifically bind to the target antigens considered herein. Thus, the “checkpoint antigen-binding domain” binds to the target checkpoint antigen as outlined herein. As is known in the art, these CDRs generally exist as a first set of variable heavy chain CDRs (VHCDR or VHCDR) and a second set of variable light chain CDRs (VLCDR or VLCDR), each containing three CDRs: VHCDR1, VHCDR2, and VHCDR3 for the heavy chain and VLCDR1, VLCDR2, and VLCDR3 for the light chain. The CDRs are present in the variable heavy chain domain and the variable light chain domain, respectively, and together form the Fv region. (See Table 1 and the relevant discussion above for the CDR numbering scheme). Thus, in some cases, the six CDRs of the antigen-binding domain are contributed to by the variable heavy chain and the variable light chain domains. In the “Fab” format, the set of six CDRs is contributed by two different polypeptide sequences, a variable heavy chain domain (including VH or VH, VHCDR1, VHCDR2, and VHCDR3), and a variable light chain domain (including VL or VL, VLCDR1, VLCDR2, and VLCDR3), where the C-terminus of the VH domain is bound to the N-terminus of the CH1 domain of the heavy chain, and the C-terminus of the VL domain is bound to the N-terminus of the constant light chain domain (thus forming the light chain). In the scFv format, the VH and VL domains are covalently bound to a single polypeptide sequence through the use of a linker ("scFv linker"), as generally outlined herein, which can be either VH-linker-VL or VL-linker-VH (starting from the N-terminus), with the former being generally preferred (including optional domain linkers on both sides, depending on the format used (e.g., from Figure 13)). Generally, the C-terminus of the scFv domain is bound to the N-terminus of the hinge of the second monomer.

[0037] In this specification, “modification” means the substitution, insertion, and / or deletion of amino acids in a polypeptide sequence, or an alteration of a portion chemically linked to a protein. For example, the modification may be an altered carbohydrate or PEG structure linked to a protein. In this specification, “amino acid modification” means the substitution, insertion, and / or deletion of amino acids in a polypeptide sequence. For clarity, unless otherwise stated, amino acid modification always refers to amino acids encoded by DNA, e.g., the 20 amino acids that have codons in DNA and RNA.

[0038] In this specification, “amino acid substitution” or “substitution” means replacing an amino acid at a specific position in a parent polypeptide sequence with a different amino acid. In particular, in some embodiments, the substitution is for an amino acid that does not naturally exist at a particular position (either not naturally present in an organism or not present in any organism). For example, substitution E272Y refers to a variant polypeptide in which the glutamic acid at position 272 is substituted with tyrosine, in this case the Fc variant. For clarity, a protein that is manipulated to alter the nucleic acid coding sequence but not alter the starting amino acid (e.g., replacing CGG (coding arginine) with CGA (still encoding arginine) to increase the expression level in a host organism) is not an “amino acid substitution.” That is, if a protein has the same amino acid at its particular starting position despite the creation of a new gene encoding the same protein, it is not an amino acid substitution.

[0039] As used herein, “amino acid insertion” or “insertion” means the addition of an amino acid sequence to a specific position in the parent polypeptide sequence. For example, -233E or 233E indicates an insertion of glutamic acid after position 233 and before position 234. In addition, -233ADE or A233ADE indicates an insertion of AlaAspGlu after position 233 and before position 234.

[0040] As used herein, "amino acid deletion" or "deletion" means the removal of an amino acid sequence at a specific position in the parent polypeptide sequence. For example, E233- or E233#, E233() or E233del indicate the deletion of glutamic acid at position 233. In addition, EDA233- or EDA233# indicates the deletion of the sequence GluAspAla, which begins at position 233.

[0041] As used herein, “variant protein,” “protein variant,” or “variant” means a protein that differs from that of the parent protein based on at least one amino acid modification. A protein variant has at least one amino acid modification compared to the parent protein, but not so many that the variant protein cannot align with the parent protein using an alignment program such as those described below. Generally, the variant proteins outlined herein (such as the variant Fc domain) are generally at least 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, or 99% identical to the parent protein using an alignment program such as BLAST, as described below.

[0042] As described below, in some embodiments, the parent polypeptide, for example, the Fc parent polypeptide, is a human wild-type sequence such as a heavy chain constant domain or Fc region derived from IgG1, IgG2, IgG3, or IgG4. However, human sequences having variants can also function as "parent polypeptides," including, for example, the IgG1 / 2 hybrid described in U.S. Patent Application Publication No. 2006 / 0134105. The protein variant sequences described herein preferably have at least about 80% identity with the parent protein sequence, most preferably at least about 90% identity, and more preferably at least about 95-98-99% identity. Therefore, as used herein, “antibody variant” or “variant antibody” means an antibody that is different from the parent antibody based on at least one amino acid modification; “IgG variant” or “variant IgG” means an antibody that is different from the parent IgG (which, again, is often derived from a human IgG sequence) based on at least one amino acid modification; and “immunoglobulin variant” or “variant immunoglobulin” means an immunoglobulin sequence that is different from that of the parent immunoglobulin sequence based on at least one amino acid modification. As used herein, “Fc variant” or “variant Fc” means a protein that has an amino acid modification in its Fc domain compared to the Fc domain of human IgG1, IgG2, or IgG4.

[0043] The Fc variants of the present invention are defined according to the amino acid modifications that constitute them. For example, N434S or 434S is an Fc variant having a substituted serine at position 434 relative to the parent Fc polypeptide, and the numbering follows the EU index. Similarly, M428L / N434S defines an Fc variant having substitutions M428L and N434S relative to the parent Fc polypeptide. The identity of the WT amino acids does not need to be specified, in which case the aforementioned variant is referred to as 428L / 434S. It should be noted that the order in which substitutions are provided is arbitrary, i.e., for example, N434S / M428L is the same Fc variant as M428L / N434S. For all positions considered in this invention relating to antibodies, unless otherwise specified, the numbering of amino acid positions follows the EU index. The EU index, or EU numbering scheme as in the case of Kabat, refers to the numbering of EU antibodies. Kabat et al. collected numerous primary sequences of the variable regions of the heavy and light chains. Based on the degree of sequence conservation, Kabat et al. classified the individual primary sequences into CDRs and frameworks and compiled a list (see Sequences of Immunological Interest, 5th edition, NIH publication, No. 91-3242, E. Kabat et al., which is incorporated herein by reference in its entirety). See also Edelman et al., 1969, Proc Natl Acad Sci USA 63:78-85 (which is incorporated herein by reference in its entirety). Modifications may be additions, deletions, or substitutions.

[0044] In this specification, "protein" means at least two covalently bonded amino acids, including proteins, polypeptides, oligopeptides, and peptides. In addition, polypeptides constituting the antibodies of the present invention may include synthetic derivatization of one or more side chains or terminals, glycosylation, PEGylation, circular permutation, cyclization, linking to other molecules, fusion to proteins or protein domains, and addition of peptide tags or labels.

[0045] As used herein, "residue" refers to the position in a protein and its associated amino acid identity. For example, asparagine 297 (also known as Asn297 or N297) is the residue at position 297 in the human antibody IgG1.

[0046] As used herein, “Fab” or “Fab region” generally refers to a polypeptide comprising the VH, CH1, VL, and CL immunoglobulin domains on two different polypeptide chains (e.g., VH-CH1 on one chain and VL-CL on the other). Fab may refer to this region alone or to this region in relation to the bispecific antibody of the present invention. In relation to Fab, Fab includes the Fv region in addition to the CH1 and CL domains.

[0047] As used herein, “Fv,” “Fv fragment,” or “Fv region” means a polypeptide containing the VL and VH domains of an ABD. An Fv region can be formatted as both a Fab (two different polypeptides, including the constant region outlined above) and an scFv, with the VL and VH domains being combined (generally with the linker discussed herein) to form an scFv.

[0048] In this specification, "single-stranded Fv" or "scFv" generally refers to a variable heavy-chain domain covalently bonded to a variable light-chain domain, which forms an scFv or scFv domain using the scFv linker discussed herein. The scFv domain can be in either direction (VH-linker-VL or VL-linker-VH) from the N-terminus to the C-terminus. In the sequences shown in the sequence listings and figures, the order of the VH and VL domains is indicated in the name. For example, H.X_L.Y means VH-linker-VL from the N-terminus to the C-terminus, and L.Y_H.X means VL-linker-VH.

[0049] As used herein, "IgG subclass modification" or "isotype modification" refers to an amino acid modification that converts one amino acid of one IgG isotype to a corresponding amino acid of a different, matching IgG isotype. For example, since IgG1 contains tyrosine at position 296 of EU and IgG2 contains phenylalanine, the F296Y substitution in IgG2 is considered an IgG subclass modification.

[0050] As used herein, “modifications not occurring naturally” means non-isotypic amino acid modifications. For example, since none of the human IgGs contain serine at position 434, the substitution 434S in IgG1, IgG2, IgG3, or IgG4 (or their hybrids) is considered a modification not occurring naturally.

[0051] As used herein, "amino acid" and "amino acid identity" mean one of the 20 naturally occurring amino acids encoded by DNA and RNA.

[0052] As used herein, “effector function” refers to a biochemical event resulting from the interaction between an antibody Fc region and an Fc receptor or ligand. Effector functions include, but are not limited to, ADCC, ADCP, and CDC.

[0053] As used herein, “IgG Fc ligand” means any biologically derived molecule, preferably a polypeptide, that binds to the Fc region of an IgG antibody to form an Fc / Fc ligand complex. Fc ligands include, but are not limited to, FcγRI, FcγRII, FcγRIII, FcRn, C1q, C3, mannan-binding lectins, mannose receptors, staphylococcal protein A, streptococcal protein G, and viral FcγR. Fc ligands also include Fc receptor homologs (FcRH), a family of Fc receptors homologous to FcγR (the whole is incorporated by reference, Davis et al., 2002, Immunological Reviews 190:123-136). Fc ligands may include undiscovered molecules that bind to Fc. Specific IgG Fc ligands are FcRn and Fc gamma receptors. As used herein, "Fc ligand" means any biologically derived molecule, preferably a polypeptide, that binds to the Fc region of an antibody to form an Fc / Fc ligand complex.

[0054] As used herein, “Fc gamma receptor,” “FcγR,” or “Fc gamma R” means any member of the protein family that binds to the Fc region of an IgG antibody and is encoded by the FcγR gene. In humans, this family includes, but is not limited to, FcγRI(CD64), which contains the isoforms FcγRIa, FcγRIb, and FcγRIC; FcγRII(CD32), which contains the isoforms FcγRIIa (including the allotypes H131 and R131), FcγRIIb (including FcγRIIb-1 and FcγRIIb-2), and FcγRIIc; and FcγRIII(CD16), which contains the isoforms FcγRIIIa (including the allotypes V158 and F158), and FcγRIIIb (including the allotypes FcγRIIb-NA1 and FcγRIIb-NA2) (the whole is incorporated by reference Jefferis et al., 2002, Immunol Lett 82:57-65), as well as any undiscovered human FcγR or FcγR isoform or allotype. FcγR may originate from any organism, including but not limited to humans, mice, rats, rabbits, and monkeys. Mouse FcγR includes, but is not limited to, FcγRI(CD64), FcγRII(CD32), FcγRIII(CD16), and FcγRIII-2(CD16-2), as well as any undiscovered mouse FcγR or FcγR isoform or allotype.

[0055] As used herein, “FcRn” or “neonatal Fc receptor” means a protein that binds to the Fc region of an IgG antibody and is at least partially encoded by the FcRn gene. FcRn may be derived from any organism, including but not limited to humans, mice, rats, rabbits, and monkeys. As is known in the art, a functional FcRn protein often comprises two polypeptides, referred to as a heavy chain and a light chain. The light chain is β-2-microglobulin, and the heavy chain is encoded by the FcRn gene. Unless otherwise stated herein, FcRn or FcRn protein refers to the complex of the FcRn heavy chain and β-2-microglobulin. Various FcRn variants are used to increase binding to the FcRn receptor, and in some cases to increase the serum half-life. “FcRn variant” means one that increases binding to the FcRn receptor, and preferred FcRn variants are listed below.

[0056] As used herein, “parent polypeptide” means a starting polypeptide that is subsequently modified to produce a variant. The parent polypeptide may be a naturally occurring polypeptide, or a variant or engineered version of a naturally occurring polypeptide. Accordingly, as used herein, “parent immunoglobulin” means an unmodified immunoglobulin polypeptide that is modified to produce a variant, and as used herein, “parent antibody” means an unmodified antibody that is modified to produce a variant antibody. It should be noted that “parent antibody” includes known commercially available recombinant antibodies, as outlined below. In this regard, “parent Fc domain” is related to the listed variants, and therefore “variant human IgG1 Fc domain” is compared to the parent Fc domain of human IgG1, “variant human IgG4 Fc domain” is compared to the parent Fc domain of human IgG4, and so on.

[0057] As used herein, “Fc,” “Fc region,” or “Fc domain” means a polypeptide containing the CH2-CH3 domain of an IgG molecule, and possibly including the hinge. In the EU numbering of human IgG1, the CH2-CH3 domain contains amino acids 231-447, and the hinge contains amino acids 216-230. Therefore, the definition of “Fc domain” includes both amino acids 231-447 (CH2-CH3) or 216-447 (hinge-CH2-CH3), or fragments thereof. In this context, “Fc fragment” may contain fewer amino acids from either or both of the N-terminus and C-terminus, but still retain the ability to form dimers with other Fc domains or Fc fragments so that they can be detected using standard size-based methods (e.g., non-denaturing chromatography, size exclusion chromatography, etc.). Human IgG Fc domains are particularly useful in the present invention and may be Fc domains derived from human IgG1, IgG2, or IgG4.

[0058] A "variant Fc region" contains amino acid modifications compared to the parent Fc domain. Therefore, a "variant human IgG1 Fc domain" contains amino acid modifications (including amino acid deletions in the case of excised variants, but generally amino acid substitutions) compared to the human IgG1 Fc domain. Generally, a variant Fc domain has at least about 80, 85, 90, 95, 97, 98, or 99 percent identity with the corresponding parent human IgG Fc domain (using the identity algorithms discussed below with default parameters (in one embodiment, the BLAST algorithm known in the art is used)). Alternatively, a variant Fc domain may have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 amino acid modifications compared to the parent Fc domain. Alternatively, the variant Fc domain may have up to 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 amino acid modifications compared to the parent Fc domain. In addition, as discussed herein, the variant Fc domains herein still retain the ability to form dimers with other Fc domains, which can be measured using known techniques described herein, such as non-denaturing gel electrophoresis.

[0059] In this specification, “heavy chain constant region” means the CH1-hinge-CH2-CH3 portion of an antibody (or fragment thereof), excluding the variable heavy chain domain, which in the EU numbering of human IgG1 corresponds to amino acids 118-447. In this specification, “heavy chain constant region fragment” means a heavy chain constant region having fewer amino acids from either or both of the N-terminus and / or C-terminus, but still retaining the ability to form dimers with another heavy chain constant region.

[0060] As used herein, “position” means a location within a protein sequence. Positions may be numbered sequentially or by an established format, such as the EU Index for antibody numbering.

[0061] As used herein, "target antigen" refers to a molecule that is specifically bound by the antibody-binding domain, which includes the variable region of a given antibody. In this example, as will be discussed below, the target antigen is a checkpoint inhibitor protein.

[0062] In the context of the monomers of the heterodimer antibodies of the present invention as used herein, “strandedness” means incorporating the heterodimerizing variant into each monomer in such a way that it retains the ability to “match” and form a heterodimer, similar to the double strands of “matching” DNA. For example, if several pI variants are manipulated into monomer A (e.g., to increase the pI), then stereovariants that are “charge pairs,” which can be similarly utilized, do not interfere with the pI variants, and for example, the charge variants with increased pI are placed on the same “strand” or “monomer” and retain both functionality. Similarly, for “skew” variants that form sets of pairs, as outlined in more detail below, those skilled in the art will consider the pI when determining which strand or monomer one set of pairs goes into, and therefore, pI separation is also maximized using the pI of the skew.

[0063] As used herein, "target cells" refers to cells that express a target antigen.

[0064] In the context of producing the bispecific antibody according to the present invention, "host cell" means a cell that contains exogenous nucleic acids encoding the components of the bispecific antibody and is capable of expressing the bispecific antibody under suitable conditions. Suitable host cells are described below.

[0065] As used herein, “variable region” or “variable domain” means a region of immunoglobulin that includes one or more Ig domains substantially encoded by any of the Vκ, Vλ, and / or VH genes constituting the kappa, lambda, and heavy chain immunoglobulin loci, respectively, and includes a CDR that confers antigen specificity. Thus, a “variable heavy chain domain” pairs with a “variable light chain domain” to form an antigen-binding domain ("ABD"). In addition, each variable domain includes three hypervariable regions ("complementarity-determining regions," "CDRs") (VHCDR1, VHCDR2, and VHCDR3 in the variable heavy chain domain; VLCDR1, VLCDR2, and VLCDR3 in the variable light chain domain) and four framework (FR) regions, arranged in the order FR1-CDR1-FR2-CDR2-FR3-CDR3-FR4 from the amino terminus to the carboxy terminus.

[0066] In this specification, “wild-type” or “WT” means a naturally occurring amino acid or nucleotide sequence, including allelic mutations. WT proteins have an amino acid or nucleotide sequence that has not been intentionally modified.

[0067] The present invention provides several antigen-binding domains having sequence identity with human antibody domains. Sequence identity between two similar sequences (e.g., antibody variable domains) can be determined using the following methods: Smith, TF & Waterman, MS (1981) "Comparison Of Biosequences," Adv.Appl.Math.2:482 [Local Homology Algorithm], Needleman, SB & Wunsch, CD (1970) "A General Method Applicable To The Search For Similarities In The Amino Acid Sequence Of Two Proteins," J.Mol.Biol.48:443 [Homology-Alignment Algorithm], Pearson, WR & Lipman, DJ (1988) "Improved Tools For Biological Sequence Comparison," Proc.Natl.Acad.Sci.(USA)85:2444 [Similarity Search Method], or Altschul, SF et al, (1990) "Basic Local Alignment Search Sequence identity can be measured by algorithms such as the "BLAST" algorithm (see https: / / blast.ncbi.nlm.nih.gov / Blast.cgi), which is described in Tool, J.Mol.Biol.215:403-10. When using any of the aforementioned algorithms, default parameters (window length, gap penalty, etc.) are used. In one embodiment, sequence identity is measured using the BLAST algorithm with default parameters.

[0068] The antibodies of the present invention are generally isolated or recombinant. As used herein to describe the various polypeptides disclosed, "isolated" means a polypeptide that has been identified, separated, and / or recovered from the cells or cell culture in which it was expressed. Typically, an isolated polypeptide will be prepared by at least one purification step. An "isolated antibody" refers to an antibody that substantially does not contain other antibodies having different antigen specificities. "Recombinant" means that an antibody is produced using recombinant nucleic acid techniques in a foreign host cell, and they can also be isolated.

[0069] "Specific binding" to a particular antigen or epitope, or "specifically binds to" or "is specific for" a particular antigen or epitope, means a binding that is measurably different from non-specific interactions. Specific binding can be measured, for example, generally by determining the binding of a molecule compared to the binding of a control molecule that is a molecule of a similar structure having no binding activity. For example, specific binding can be determined by competition with a control molecule similar to the target.

[0070] Specific binding to a particular antigen or epitope can be, for example, at least about 10 -4 M, at least about 10 -5 M, at least about 10 -6 M, at least about 10 -7 M, at least about 10 -8 M, at least about 10 -9 M, alternatively at least about 10 -10 M, at least about 10 -11 M, at least about 10 -12 M, or more, and can be shown by an antibody having a KD, where KD refers to the dissociation rate of a particular antibody-antigen interaction. Typically, an antibody that specifically binds to an antigen will have a KD that is 20-fold, 50-fold, 100-fold, 500-fold, 1000-fold, 5,000-fold, 10,000-fold, or more times that of the control molecule compared to the antigen or epitope.

[0071] Furthermore, specific binding to a particular antigen or epitope may be demonstrated, for example, by an antibody whose KA or Ka for the antigen or epitope is at least 20-fold, 50-fold, 100-fold, 500-fold, 1000-fold, 5,000-fold, 10,000-fold, or more than 20-fold, 50-fold, 10,000-fold, or higher compared to a control, where KA or Ka refers to the association rate of a particular antibody-antigen interaction. Binding affinity is generally measured using Biacore, SPR, or BLI assays.

[0072] E. Antibodies In one embodiment, bispecific antibodies that bind to CLDN18.2 and CD3 in various formats, as outlined below and generally shown in Figures 13 and 42, are provided herein. These bispecific heterodimer antibodies include a CLDN18.2-binding domain. In certain embodiments, the CLDN18.2-binding domain includes VHCDR1, VHCDR2, VHCDR3, VLCDR1, VLCDR2, and VLCDR3 sequences of CLDN18.2-binding domains selected from the group shown in Figure 10. In some embodiments, the CLDN18.2-binding domain includes underlined VHCDR1, VHCDR2, VHCDR3, VLCDR1, VLCDR2, and VLCDR3 sequences of CLDN18.2-binding domains selected from those shown in Figure 10.

[0073] These bispecific heterodimer antibodies bind to CLDN18.2 and CD3. Such antibodies include a CD3-binding domain and at least one CLDN18.2-binding domain. Any suitable CLDN18.2-binding domain can be included in the anti-CLDN18.2 × anti-CD3 bispecific antibody. In some embodiments, the anti-CLDN18.2 × anti-CD3 bispecific antibody includes, but is not limited to, the CLDN18.2-binding domains shown in Figure 10, and includes one, two, three, four or more CLDN18.2-binding domains. In a particular embodiment, the anti-CLDN18.2 × anti-CD3 antibody includes a CLDN18.2-binding domain comprising the VHCDR1, VHCDR2, VHCDR3, VLCDR1, VLCDR2, and VLCDR3 sequences of CLDN18.2-binding domains selected from the group consisting of those shown in Figure 10. In some embodiments, the anti-CLDN18.2×anti-CD3 antibody includes a CLDN18.2 binding domain comprising underlined VHCDR1, VHCDR2, VHCDR3, VLCDR1, VLCDR2, and VLCDR3 sequences of the CLDN18.2 domain selected from the group shown in Figure 10. In some embodiments, the anti-CLDN18.2×anti-CD3 antibody includes a CLDN18.2 binding domain comprising a variable heavy chain domain and a variable light chain domain of the CLDN18.2 binding domain selected from the group shown in Figure 10. In exemplary embodiments, the anti-CLDN18.2×anti-CD3 antibody includes an anti-CLDN18.2 H1L1 or H2L1 binding domain.

[0074] The anti-CLDN18.2× anti-CD3 antibodies provided herein may contain any preferred CD3-binding domain. In certain embodiments, the anti-CLDN18.2× anti-CD3 antibody includes a CD3-binding domain comprising the VHCDR1, VHCDR2, VHCDR3, VLCDR1, VLCDR2, and VLCDR3 sequences of a CD3-binding domain selected from the group shown in Figure 12. In some embodiments, the anti-CLDN18.2× anti-CD3 antibody includes a CD3-binding domain comprising the underlined VHCDR1, VHCDR2, VHCDR3, VLCDR1, VLCDR2, and VLCDR3 sequences of a CD3-binding domain selected from the group shown in Figure 12. In some embodiments, the anti-CLDN18.2× anti-CD3 antibody includes a CD3-binding domain comprising a variable heavy chain domain and a variable light chain domain of a CD3-binding domain selected from the group shown in Figure 12. In some embodiments, the CD3 binding domain is selected from anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.48, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47. As outlined herein, these anti-CD3 antigen-binding domains (CD3-ABD) can be used in scFv format in either direction (e.g., N-terminus to C-terminus, VH-scFv linker-VL, or VL-scFv linker-VH).

[0075] As used herein, the term "antibody" is used in general terms. The antibodies used in the present invention may take several forms as described herein, including conventional antibodies, as well as antibody derivatives, fragments, and mimetic compounds as described herein.

[0076] Conventional antibody structural units typically contain tetramers. Each tetramer typically consists of two identical polypeptide chain pairs, each pair having one "light" chain (typically with a molecular weight of about 25 kDa) and one "heavy" chain (typically with a molecular weight of about 50–70 kDa). Human light chains are classified as kappa light chains and lambda light chains. This invention covers the IgG class, which includes, but is not limited to, several subclasses: IgG1, IgG2, IgG3, and IgG4. It should be noted that IgG1 has different allotypes with polymorphisms at 356 (D or E) and 358 (L or M). The sequences shown herein use the 356D / 358M allotype, but other allotypes are included herein. That is, any sequence containing an IgG1 Fc domain included herein may have 356E / 358L instead of 356D / 358M allotype.

[0077] In addition, many of the antibodies described herein have at least one cysteine ​​at position 220 replaced by serine. Generally, this is on the "scFv monomer" side for most of the sequences shown herein, but it may also be on the "Fab monomer" side, or both, to reduce disulfide formation. Specifically included in the sequences described herein are those in which one or both of these cysteines are replaced (C220S).

[0078] Therefore, as used herein, “isotype” means any subclass of immunoglobulin defined by the chemical and antigenic characteristics of their constant regions. It should be understood that therapeutic antibodies may also include hybrids of isotypes and / or subclasses. For example, as shown in U.S. Publication No. 2009 / 0163699 incorporated by reference, the present invention includes the use of human IgG1 / G2 hybrids.

[0079] The hypervariable region generally consists of approximately amino acid residues 24-34 (LCDR1; "L" indicates the light chain), 50-56 (LCDR2), and 89-97 (LCDR3) in the light chain variable region, and approximately 31-35B (HCDR1; "H" indicates the heavy chain), 50-65 (HCDR2), and 95-102 (HCDR3) in the heavy chain variable region; Kabat et al., SEQUENCES OF PROTEINS OF IMMUNOLOGICAL INTEREST, 5th Ed. Public Health Service, National Institutes of This includes residues that form hypervariable loops (e.g., residues 26-32 (LCDR1), 50-52 (LCDR2), and 91-96 (LCDR3) in the light chain variable region, and residues 26-32 (HCDR1), 53-55 (HCDR2), and 96-101 (HCDR3) in the heavy chain variable region; Chothia and Lesk (1987) J.Mol.Biol.196:901-917). Specific CDRs of the present invention are described below.

[0080] As will be understood by those skilled in the art, the exact numbering and arrangement of CDRs may differ between different numbering systems. However, it should be understood that the disclosure of variable heavy chain sequences and / or variable light chain sequences includes the disclosure of the relevant (unique) CDRs. Thus, the disclosure of each variable heavy chain region is a disclosure of VHCDRs (e.g., VHCDR1, VHCDR2, and VHCDR3), and the disclosure of each variable light chain region is a disclosure of VLCDRs (e.g., VLCDR1, VLCDR2, and VLCDR3). A useful comparison of CDR numbering is given below; see Lafranc et al., Dev. Comp. Immunol. 27(1):55-77 (2003).

[0081] [Table 1]

[0082] Throughout this specification, the Kabat numbering system is generally used when referring to residues within the variable domain (approximately residues 1-107 in the light chain variable region and residues 1-113 in the heavy chain variable region), while the EU numbering system is for the Fc region (see, for example, Kabat et al. (1991) above).

[0083] Another type of heavy chain Ig domain is the hinge region. In this specification, “hinge,” “hinge region,” “antibody hinge region,” or “hinge domain” refers to a mobile polypeptide containing amino acids between the first and second constant domains of an antibody. Structurally, the IgG CH1 domain terminates at position 215 of EU, and the IgG CH2 domain begins at residue position 231 of EU. Therefore, for IgG, the antibody hinge is defined herein as encompassing positions 216 (E216 in IgG1) to 230 (p230 in IgG1), with numbering following the EU index, as in the case of Kabat. In some cases, a “hinge fragment” is used, containing fewer amino acids than either or both of the N-terminus and C-terminus of the hinge domain. The hinge may also be a domain linker, as will be discussed below. In these embodiments, useful hinge-based domain linkers are shown in Figure 5, and in particular, Sequence ID No. 28 can be used specifically in the “1+1 Fab-scFv-Fc” and “2+1 Fab2-scFv-Fc” formats. As described herein, pI variants can also be fabricated in the hinge region.

[0084] The light chain generally consists of two domains: a variable light chain domain (containing the light chain CDR, which together with the variable heavy chain domain form the Fv region) and a constant light region (often referred to as CL or Cκ).

[0085] Another region of interest for the additional substitution outlined below is the Fc region.

[0086] The present invention provides a number of different CDR sets. In this case, a “complete CDR set” includes three variable light chain and three variable heavy chain CDRs, e.g., VLCDR1, VLCDR2, VLCDR3, VHCDR1, VHCDR2, and VHCDR3. These may each be part of a larger variable light chain or variable heavy chain domain. In addition, as fully outlined herein, the variable heavy chain domain and variable light chain domain may be on separate polypeptide chains when the heavy chain and light chain are used (e.g., when Fab is used), or on a single polypeptide chain in the case of scFv sequences.

[0087] CDRs contribute to the formation of antigen binding, or more specifically, the formation of the antibody's epitope binding site. An "epitope" refers to a determinant that interacts with a specific antigen-binding site within the variable region of an antibody molecule, also known as a paratope. Epitopes are groups of molecules, such as amino acids or sugar side chains, and typically possess specific structural and charge properties. A single antigen may have more than one epitope.

[0088] An epitope may include amino acid residues directly involved in binding (also referred to as the immunodominant components of the epitope) and other amino acid residues not directly involved in binding, such as amino acid residues that are effectively blocked by specific antigen-binding peptides. In other words, the amino acid residues are within the footprint of the specific antigen-binding peptide.

[0089] Epitopes can be either conformational or linear. Conformational epitopes are produced by spatially juxtaposed amino acids from different segments of a linear polypeptide chain. Linear epitopes are produced by adjacent amino acid residues within a polypeptide chain. Conformational and non-conformational epitopes can be distinguished in that binding to the former is lost in the presence of a denaturing solvent, while binding to the latter is not.

[0090] An epitope typically contains at least three, more commonly, at least five, or eight to ten amino acids within its unique spatial structure. Antibodies that recognize the same epitope can be validated in a simple immunoassay demonstrating the ability of one antibody to block the binding of another antibody to its target antigen, for example, "binning." As outlined below, the present invention includes not only the enumerated antigen-binding domains and the antibodies herein, but also those that compete for binding to epitopes bound by the enumerated antigen-binding domains.

[0091] Accordingly, the present invention provides different antibody domains. As described herein and known in the art, the heterodimeric antibodies of the present invention contain different domains in the heavy chain and light chain, which may also overlap. These domains include, but are not limited to, Fc domains, CH1 domains, CH2 domains, CH3 domains, hinge domains, heavy chain constant domains (CH1-hinge-Fc domain or CH1-hinge-CH2-CH3), variable heavy chain domains, variable light chain domains, light chain constant domains, Fab domains, and scFv domains.

[0092] Therefore, the “Fc domain” includes the -CH2-CH3 domain and optionally a hinge domain (-hinge domain-CH2-CH3) (again, many embodiments rely on a hinge domain derived from human IgG1 having the C220S variant). In the embodiments herein, particularly for the “1+1” format, when scFv is bound to the Fc domain, the C-terminus of the scFv construct is bound to all or part of the hinge of the Fc domain. For example, it is generally bound to the sequence EPKS (SEQ ID NO: 473), which is the start of the hinge. In some cases, for example, in the “2+1” format, the domain linker may be a combination of mobile linker amino acids as well as part or all of the hinge. For example, SEQ ID NO: 29 is such an example.

[0093] Embodiments of the present invention include at least one scFv domain, which generally includes a variable heavy-chain domain and a variable light-chain domain that do not occur naturally but are linked together by an scFv linker. As outlined herein, the scFv domain is generally oriented as VH-scFv linker-VL from the N-terminus to the C-terminus, but this can be reversed to VL-scFv linker-VH for any scFv domain (or a domain constructed using Fab-derived VH and VL sequences) using one- or both-optional linkers depending on the format (see generally Figures 13 and 42).

[0094] (a) Linker As shown herein, there are several suitable linkers (used as either domain linkers or scFv linkers) that can be used to covalently bond to the enumerated domains, including conventional peptide bonds generated by recombinant technology. In some embodiments, the linker peptide may mainly consist of the following amino acid residues, namely Gly, Ser, Ala, or Thr. The linker peptide should be long enough to link the two molecules in such a way that they assume the correct conformations relative to each other so that they retain the desired activity. In one embodiment, the linker is about 1 to 50 amino acid long, preferably about 1 to 30 amino acid long. In one embodiment, a linker of 1 to 20 amino acid long may be used, and in some embodiments, about 5 to about 10 amino acids are used. Useful linkers include, for example, glycine-serine polymers, glycine-alanine polymers, alanine-serine polymers, and other mobile linkers, including (GS)n, (GSGGS)n (SEQ ID NO: 474), (GGGGS)n (SEQ ID NO: 475), and (GGGS)n (SEQ ID NO: 476) (where n is an integer of at least 1 (and generally 3-4)), some of which are shown in Figure 5. Alternatively, a variety of non-proteinaceous polymers, including but not limited to polyethylene glycol (PEG), polypropylene glycol, polyoxyalkylene, or copolymers of polyethylene glycol and polypropylene glycol, can be used as linkers.

[0095] Other linker sequences may include, but not all, the residues of the CL / CH1 domain, any sequence of any length of the CL / CH1 domain, for example, the first 5 to 12 amino acid residues of the CL / CH1 domain. Linkers may be derived from immunoglobulin light chains, e.g., Cκ or Cλ. Linkers may be derived from immunoglobulin heavy chains of any isotype, including, for example, γ1, Cγ2, Cγ3, Cγ4, Cα1, Cα2, Cδ, Cε, and Cμ. Linker sequences may also be derived from other proteins, such as Ig-like proteins (e.g., TCR, FcR, KIR), hinge region-derived sequences, and other native sequences derived from other proteins.

[0096] In some embodiments, the linker is a “domain linker” used to link any two domains outlined herein together. For example, in Figure 42F, there may be a domain linker that links the C-terminus of the CH1 domain of Fab to the N-terminus of scFv, and another optional domain linker that links the C-terminus of scFv to the CH2 domain (although in many embodiments, a hinge is used as this domain linker). Any suitable linker can be used, but many embodiments utilize glycine-serine polymers as domain linkers, including, for example, (GS)n, (GSGGS)n (SEQ ID NO: 474), (GGGGS)n (SEQ ID NO: 475), and (GGGS)n (SEQ ID NO: 476) (where n is an integer of at least 1 (generally 3-4-5)), as well as any peptide sequence that enables recombination of two domains having sufficient length and flexibility so that each domain retains its biological function. In some cases, charged domain linkers can be used, as used in some embodiments of the scFv linker, with attention paid to the “twistiness” outlined below.

[0097] There are several domain linkers that are particularly useful, including sequence numbers 28, 29, 30, and 31, which are especially relevant to domain linkers used to join scFv domains to Fc domains in the "2+1" format.

[0098] In some embodiments, the linker is an “scFv linker” used to covalently bond the VH and VL domains discussed herein. In many cases, the scFv linker is a charged scFv linker, some of which are shown in Figure 5. Thus, the present invention further provides charged scFv linkers for facilitating the separation of pI between a first monomer and a second monomer. That is, by incorporating either a positive or negative charged scFv linker (or both in the case of a scaffold using scFvs on different monomers), this makes it possible to change the pI of monomers including the charged linker without further altering the Fc domain. These charged linkers can be substituted in any scFv including a standard linker. Also, as will be understood by those skilled in the art, the charged scFv linker is used on the correct “chain” or monomers according to the desired change in pI. For example, as discussed herein, in order to produce a 1+1Fab-scFv-Fc format heterodimer antibody, the original pI of the Fv region for each of the desired antigen-binding domains is calculated, one is selected to produce the scFv, and depending on the pI, either a positive or negative linker is selected.

[0099] Charged domain linkers can also be used to increase the pI separation of monomers according to the present invention, and thus those shown in Figure 5 can be used in any embodiment of this specification in which a linker is utilized.

[0100] Specifically, the format shown in Figure 1 is an antibody commonly referred to as a "heterodimal antibody," meaning that the protein has at least two associated Fc sequences that self-assemble into a heterodimer Fc domain, and at least two Fv regions, whether as Fab or scFv.

[0101] F. chimeric and humanized antibodies In certain embodiments, the antibody of the present invention comprises a heavy chain variable region derived from a specific germline heavy chain immunoglobulin gene and / or a light chain variable region derived from a specific germline light chain immunoglobulin gene. For example, such an antibody may comprise or consist of a human antibody comprising a heavy chain or light chain variable region that is a "product of" or "derived from" a specific germline sequence. A human antibody that is a "product of" or "derived from" a human germline immunoglobulin sequence may be identified as such by comparing the amino acid sequence of the human antibody with the amino acid sequence of the human germline immunoglobulin and selecting the human germline immunoglobulin that is the sequence closest to the sequence of the human antibody (i.e., the highest identity %) (using the method outlined herein). A human antibody that is a "product of" or "derived from" a specific human germline immunoglobulin sequence may include amino acid differences compared to the germline sequence, for example, due to naturally occurring somatic mutations or the intentional introduction of site-directed mutations. However, humanized antibodies typically contain amino acid residues that are at least 90% identical in amino acid sequence to the amino acid sequence encoded by human germline immunoglobulin genes, and that identify the antibody as derived from a human sequence when compared to germline immunoglobulin amino acid sequences of other species (e.g., mouse germline sequences). In certain cases, humanized antibodies may be at least 95, 96, 97, 98, or 99% identical in amino acid sequence to the amino acid sequence encoded by germline immunoglobulin genes, or even at least 96%, 97%, 98, or 99% identical. Typically, humanized antibodies derived from a particular human germline sequence show differences of 10 to 20 amino acids or less from the amino acid sequence encoded by human germline immunoglobulin genes (the number of variants is generally small before the introduction of any skew, pI, and elimination variants as defined herein, i.e., before the introduction of the variants of the present invention).In certain cases, humanized antibodies may differ from the amino acid sequence encoded by germline immunoglobulin genes by only 5 amino acids or less, or even 4, 3, 2, or 1 amino acid or less (again, the number of variants is generally small before the introduction of the skew, pI, and elimination variants of this specification, i.e., before the introduction of the variants of the present invention).

[0102] In one embodiment, the parent antibody is affinity-matured, as is known in the art. Structure-based methods may be used for humanization and affinity maturation, for example, as described in U.S. Patent Application No. 11 / 004,590. Selection-based methods may be used to humanize and / or affinity mature antibody variable regions, including, but not limited to, the methods described in Wu et al., 1999, J.Mol.Biol.294:151-162, Baca et al., 1997, J.Biol.Chem.272(16):10678-10684, Rosok et al., 1996, J.Biol.Chem.271(37):22611-22618, Rader et al., 1998, Proc.Natl.Acad.Sci.USA 95:8910-8915, and Krauss et al., 2003, Protein Engineering 16(10):753-759, all of which are incorporated by reference. Other humanization methods may include grafting only portions of the CDR, including, but not limited to, the methods described in U.S. Patent Application No. 09 / 810,510, Tan et al., 2002, J.Immunol. 169:1119-1125, and De Pascalis et al., 2002, J.Immunol. 169:3076-3084, which are incorporated by reference in their entirety.

[0103] G. Heterodimer antibody Therefore, in some embodiments, the target antibody is a heterodimer antibody that relies on the use of two different heavy chain variant Fc sequences. Such an antibody will self-assemble to form a heterodimer Fc domain and a heterodimer antibody.

[0104] The present invention relates to novel constructs for providing heterodimer antibodies that enable binding to one or more antigens or ligands, for example, to enable bispecific binding (e.g., anti-CLDN18.2 binding and anti-CD3 binding). Heterodimer antibody constructs are based on the self-assembly of two "monomers" that assemble into two Fc domains of the antibody's heavy chain, e.g., a "dimer". Heterodimer antibodies are made by altering the amino acid sequence of each monomer, as will be discussed more fully below. Thus, the present invention generally relates to the creation of heterodimer antibodies, which allow antigens (e.g., CLDN18.2 and CD3) to be co-linked in several ways, relying on amino acid variants in different constant regions on each chain, in order to facilitate heterodimer formation and / or to facilitate the purification of heterodimers than homodimers.

[0105] Therefore, the present invention provides a bispecific antibody. In some embodiments, the present invention provides a bispecific antibody comprising a CLDN18.2 binding domain. In some embodiments, the bispecific antibody is an anti-CLDN18.2 × anti-CD3 bispecific antibody. A continuing problem in antibody technology is the demand for "bispecific" antibodies that bind simultaneously to two different antigens, generally bringing different antigens into proximity and potentially resulting in novel functions and therapeutic approaches. Generally, these antibodies are produced by incorporating the genes for each heavy and light chain into a host cell. This generally results in the formation of not only two homodimers (AA and BB (without the issue of light chain heterodimers)) but also a desired heterodimer (AB). However, the main obstacle in bispecific antibody formation is the difficulty in purifying heterodimer antibodies separately from homodimer antibodies and / or biasing towards heterodimer formation to outweigh homodimer formation.

[0106] There are several mechanisms that can be used to produce the heterodimers of the present invention. In addition, as will be understood by those skilled in the art, these mechanisms can be combined to ensure high heterodimerization. Therefore, amino acid variants that result in the production of heterodimers are referred to as "heterodimerizing variants." As will be discussed below, heterodimerizing variants may include stereovariants (e.g., the "knob-and-hole" or "skew" variants and the "charge pair" variants below) as well as "pI variants," which allow for the purification of homodimers separated from heterodimers. Useful mechanisms for heterodimerization include "knobs and holes" ("KIH," sometimes referred to herein as the "skew" variant (see the discussion in International Publication 2014 / 145806), the "electrostatic steering" or "charge pair" described in International Publication 2014 / 145806, the pI variant described in International Publication 2014 / 145806, and the general additional Fc variants outlined in International Publication 2014 / 145806 and below), as generally described in International Publication 2014 / 145806, which is incorporated herein in its entirety by reference.

[0107] In this invention, several fundamental mechanisms exist that can facilitate the purification of heterodimer antibodies. One relies on the use of pI variants such that each monomer has a different pI, thus enabling isoelectric purification of AA, AB, and BB dimer proteins. Alternatively, several scaffolding forms, such as the "1+1 Fab-scFv-Fc" format, also allow for size-based separation. It is also possible to "skew" the formation of heterodimers rather than homodimers, as will be further outlined below. Therefore, combinations of steric heterodimerization variants and pI variants or charge-pair variants are particularly used in this invention.

[0108] Generally, embodiments particularly used in the present invention rely on a set of variants comprising a scuba riant, which, in combination with a pI variant that increases the pI difference between two monomers, promotes heterodimerization in preference to homodimerization and facilitates the purification of heterodimers separated from homodimers.

[0109] In addition, as fully outlined below, depending on the format of the heterodimeric antibody, the pI variant can be contained within the constant domain and / or Fc domain of the monomer, or a charged linker, which is either a domain linker or an scFv linker, can be used. That is, scFv-utilizing scaffolds, such as the "1+1 Fab-scFv-Fc" format, can include a charged scFv linker (either positive or negative) to provide an additional pI boost for purification purposes. As will be understood by those skilled in the art, some 1+1 Fab-scFv-Fc formats are useful with only a charged scFv linker and no additional pI adjustment, but the present invention also provides pI variants present in one or both monomers, and / or charged domain linkers. In addition, additional amino acid manipulations for alternative functionality can also confer pI changes such as Fc, FcRn, and KO variants.

[0110] In this invention, which utilizes pI as a separation mechanism to enable the purification of heterodimeric proteins, amino acid variants can be introduced into one or both monomeric polypeptides. That is, the pI of one of the monomers (referred to herein as "monomer A" for simplicity) can be manipulated to separate it from monomer B, or the changes in both monomers A and B can be altered by increasing the pI of monomer A and decreasing the pI of monomer B. As will be fully outlined below, the pI changes of either or both monomers can be made by removing or adding charged residues (e.g., replacing a neutral amino acid with a positively or negatively charged amino acid residue, e.g., glycine to glutamic acid), changing a charged residue from positive or negative to the opposite charge (e.g., aspartic acid to lysine), or changing a charged residue to a neutral residue (e.g., loss of charge, lysine to serine). Some of these variants are shown in the figure.

[0111] Accordingly, this embodiment of the present invention provides a method for causing a sufficient change in pI in at least one monomer so that a heterodimer can be separated from a homodimer. As will be understood by those skilled in the art and will be further discussed below, this can be done by using a “wild-type” heavy chain constant region and a variant region manipulated to increase or decrease its pI (wt A-+B+ or wt A--B), or by increasing one region and decreasing the other (A+-B- or A-B+).

[0112] Therefore, generally, components of some embodiments of the present invention are amino acid variants in the constant region of an antibody, aimed at modifying the isoelectric point (pI) of at least one, if not both, monomers of a dimeric protein by incorporating the amino acid substitution ("pI variant" or "pI substitution") into one or both monomers, thereby forming a "pI antibody." As shown herein, the separation of heterodimers from two homodimers can be achieved when the pIs of the two monomers differ by only 0.1 pH units, and differences of 0.2, 0.3, 0.4, and 0.5 or more are all used in the present invention.

[0113] As will be understood by those skilled in the art, the number of pI variants to be present in each monomer or both monomers to obtain good separation will depend in part on the starting pIs of the constituents, e.g., in the 1+1 Fab-scFv-Fc format, on the starting pIs of the scFv and Fab of interest. That is, the Fv sequences of the two target antigens are calculated and decisions are made from there to determine which monomers to work with or in which "direction" (e.g., more positive or more negative). As is known in the art, different Fvs will have different starting pIs to be utilized in this invention. Generally, as outlined herein, the pIs are worked to result in a total pI difference of at least about 0.1 log for each monomer, and preferably between 0.2 and 0.5 as outlined herein.

[0114] Furthermore, as will be understood by those skilled in the art and outlined herein, in some embodiments, heterodimers can be separated from homodimers based on size. As shown in Figure 1, for example, some formats allow for the separation of heterodimers and homodimers based on size.

[0115] When pI variants are used to achieve heterodimerization by utilizing the constant region of the heavy chain, a more modular approach is provided for designing and purifying bispecific proteins, including antibodies. Therefore, in some embodiments, the dimerized variant (including scuba arians or purified dimerized variants) is not included in the variable region, and each individual antibody must be manipulated. In addition, in some embodiments, the potential for immunogenicity arising from pI variants is significantly reduced by introducing pI variants from different IgG isotypes so that pI is altered without introducing significant immunogenicity. Therefore, an additional problem to be addressed is the elucidation of low pI constant domains with high human sequence content, e.g., minimizing or avoiding non-human residues at any particular location.

[0116] Other potential secondary benefits of this pI manipulation include prolonged serum half-life and increased FcRn binding. Specifically, as described in U.S. Patent Application No. 13 / 194,904 (which is incorporated herein by reference in its entirety), reducing the pI of antibody constant domains (including those found in antibodies and Fc fusions) can result in longer serum retention in vivo. These pI variants for prolonging serum half-life also facilitate pI changes for purification.

[0117] Furthermore, it should be noted that the pI variant of the dimerized variant provides additional benefits to the analysis and quality control processes of bispecific antibodies because of its remarkable ability to exclude, minimize, or distinguish when homodimers are present. Similarly, the ability to reliably test the reproducibility of heterodimer antibody production is important.

[0118] Heterodimizing variant The present invention provides heterodimer proteins, including heterodimer antibodies in various formats that utilize heterodimerizing variants to enable heterodimerization and / or purification from homodimers.

[0119] There are several preferred pairs of sets of heterodimerizing scuba riants. These variants form "pairs" of "sets," where one pair from one set is incorporated into the first monomer, and the other pair from the other set is incorporated into the second monomer. It should be noted that these sets do not necessarily behave as "knob-in-hole" variants, but rather there is a one-to-one correspondence between residues on one monomer and residues on the other monomer. That is, these pairs of sets form an interface between two monomers that promotes heterodimer formation but not homodimer formation, allowing the proportion of heterodimers spontaneously formed under biological conditions to exceed 90%, rather than the expected 50% (25% homodimer A / A: 50% heterodimer A / B: 25% homodimer B / B).

[0120] 3D variant In some embodiments, heterodimer formation can be facilitated by the addition of stereovariants. That is, by altering the amino acids in each heavy chain, different heavy chains are more likely to associate to form a heterodimer structure than to form a homodimer with the same Fc amino acid sequence. Suitable stereovariants are shown in Figure 1.

[0121] Certain mechanisms, commonly referred to in the art as “knob-and-hole,” refer to amino acid manipulations that produce steric effects that promote heterodimer formation but not homodimer formation, and can be used selectively. This is also sometimes referred to as “knob-and-hole,” as described in U.S. Patent Application No. 61 / 596,846, Ridgway et al., Protein Engineering 9(7):617(1996); Atwell et al., J.Mol.Biol.1997 270:26; and U.S. Patent No. 8,216,805 (all of which are incorporated herein by reference in their entirety). These figures illustrate several “monomer A-monomer B” pairs that rely on “knob-and-hole.” In addition, as described in Merchant et al., Nature Biotech. 16:677 (1998), these "knob-and-hole" mutations can be combined with disulfide bonds to distort the formation of heterodimers.

[0122] The additional mechanism used for the formation of heterodimers, as described in Gunasekaran et al., J. Biol. Chem. 285(25):19637(2010), is sometimes referred to as “electrostatic steering,” which is also sometimes referred to herein as “charge pairing.” In this embodiment, electrostatics is used to skew the formation toward heterodimerization. As will be understood by those skilled in the art, these may also affect pI, i.e., purification, and therefore may in some cases be considered pI variants. However, since these are generated to force heterodimerization and are not used as a means of purification, they are classified as “steric variants.” These include, but are not limited to, D221E / P228E / L368E paired with D221R / P228R / K409R (for example, these are "monomer correspondence sets"), and C220E / P228E / 368E paired with C220R / E224R / P228R / K409R.

[0123] Additional monomer A and monomer B variants may be combined in any quantity, selectively and independently, with the pI variant outlined herein or with other variants, such as the other stereo variant shown in Figure 37 of U.S. Patent Application Publication No. 2012 / 0149876, whose figures, descriptions, and sequence numbers are expressly incorporated herein by reference.

[0124] In some embodiments, the stereovariants outlined herein can optionally and independently incorporate any pI variant (or other variants such as Fc variants, FcRn variants, etc.) into one or both monomers, and can be optionally and independently included in or excluded from the proteins of the present invention.

[0125] A list of preferred scubarians is found in Figure 1, and Figure 4 shows some pairs that are particularly useful in many embodiments. Pairs of sets that are particularly useful in many embodiments include, but are not limited to, S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E357L, and K370S:S364K / E357Q. With respect to nomenclature, the pair "S364K / E357Q:L368D / K370S" means that one monomer has the double variant set S364K / E357Q and the other has the double variant set L368D / K370S.

[0126] pI (isoelectric point) variant of a heterodimer Generally, as will be understood by those skilled in the art, pI variants fall into two common categories: those that increase the pI of a protein (basic changes) and those that decrease the pI of a protein (acidic changes). As described herein, all combinations of these variants are possible, with one monomer being wild-type or a variant that does not exhibit a significantly different pI from the wild-type, and the other being either more basic or more acidic. Alternatively, each monomer may change one to be more basic and the other to be more acidic.

[0127] Preferred combinations of pI variants are shown in Figures 1 and 2. These modifications are shown in comparison to IgG1, as outlined herein and shown in the figures, but all isotypes can be modified in this way, as can isotype hybrids. R133E and R133Q can also be used when the heavy chain constant domains are derived from IgG2-4.

[0128] In one embodiment, for example, in the format of Figures 42A, E, F, G, H, and I, a preferred combination of pI variants has one monomer (negative Fab side) containing the 208D / 295E / 384D / 418E / 421D variant (N208D / Q295E / N384D / Q418E / N421D for human IgG1), and a second monomer (positive scFv side) containing a positively charged scFv linker containing (GKPGS)4 (SEQ ID NO: 10). However, as will be understood by those skilled in the art, the first monomer contains a CH1 domain including position 208. Therefore, in constructs that do not contain the CH1 domain (for example, for antibodies that do not utilize the CH1 domain on one of the domains, this is a dual scFv format or "one-arm" format, such as those shown in Figures 42B, C, or D), the preferred negative pI variant Fc set includes the 295E / 384D / 418E / 421D variants (Q295E / N384D / Q418E / N421D for human IgG1).

[0129] Therefore, in some embodiments, one monomer has a set of substitutions from Figure 2, and the other monomer has a charged linker (either in the format of a charged scFv linker, as the monomer includes an scFv or a charged domain linker, as the format indicates, which can be selected from those shown in Figure 5).

[0130] Isotype variant In addition, many embodiments of the present invention rely on the “transfer” of pI amino acids at specific positions from one IgG isotype to another, thus reducing or eliminating the possibility of introducing undesirable immunogenicity to the variant. Some of these are shown in Figure 21 of U.S. Patent Application Publication 2014 / 0370013, which are incorporated herein by reference. Specifically, IgG1 is a common isotype for therapeutic antibodies for various reasons, including high effector function. However, the polyconstant region of IgG1 has a higher pI than that of IgG2 (8.10 vs. 7.31). By introducing IgG2 residues into the IgG1 backbone at specific positions, the pI of the resulting monomer is reduced (or increased), and in addition, it exhibits a longer serum half-life. For example, IgG1 has glycine at position 137 (pI 5.97), while IgG2 has glutamic acid (pI 3.22), and the transfer of glutamic acid affects the pI of the resulting protein. As described below, several amino acid substitutions are generally required to have a significant effect on the pI of variant antibodies. However, it should be noted that even changes in the IgG2 molecule can lead to an increase in serum half-life, as discussed below.

[0131] In other embodiments, non-isotype amino acid changes are performed (for example, by changing high-pI amino acids to low-pI amino acids) to reduce the overall charge state of the resulting protein or to allow for structural modifications for purposes such as stability, which are described in more detail below.

[0132] In addition, significant changes can be observed in each monomer of the heterodimer by manipulating the pI of both the heavy chain and light chain constant domains. As discussed herein, a difference of at least 0.5 in the pI of two monomers may enable separation by ion exchange chromatography, isoelectric focusing, or other isoelectric-sensitive methods.

[0133] Calculate pI The pI of each monomer depends on the pI of the variant heavy chain steady domain and the pI of the whole monomer, and may include the variant heavy chain steady domain and fusion partners. Therefore, in some embodiments, the pI change is calculated based on the variant heavy chain steady domain using the chart in Figure 19 of U.S. Patent Application Publication 2014 / 0370013. As discussed herein, which monomer to work with is generally determined by the intrinsic pI of the Fv region and the scaffold region. Alternatively, the pI of each monomer can be compared.

[0134] A pI variant that also confers better FcRn in in vivo binding. If pI variants reduce monomeric pI, they may have the additional benefit of improving serum retention in vivo.

[0135] Although still under investigation, it is thought that the Fc region has a longer half-life in vivo because Fc is sequestered when it binds to FcRn at pH 6 within the endosome (Ghetie and Ward, 1997 Immunol Today. 18(12):592-598, whole text incorporated by reference). The endosomal compartment then recycles Fc to the cell surface. When the compartment opens to the extracellular space, a higher pH of approximately 7.4 induces the release of Fc back into the bloodstream. In mice, Dall'Acqua et al. showed that Fc mutants with increased FcRn binding at pH 6 and pH 7.4 actually had reduced serum concentrations and the same half-life as wild-type Fc (Dall'Acqua et al. 2002, J.Immunol. 169:5171-5180, whole text incorporated by reference). The increased affinity of Fc to FcRn at pH 7.4 is thought to prevent the release of Fc back into the bloodstream. Therefore, Fc mutations that increase the half-life of Fc in vivo ideally increase FcRn binding at lower pH levels while still allowing Fc release at higher pH levels. The amino acid histidine changes its charge state in the pH range of 6.0–7.4. Therefore, it is not surprising to find His residues at key positions in the Fc / FcRn complex.

[0136] Recently, it has been suggested that antibodies with variable regions having lower isoelectric points may also have longer serum half-lives (Igawa et al., 2010 PEDS.23(5):385-392, whole incorporated by reference). However, the mechanism is still not well understood. Furthermore, the variable region differs from antibody to antibody. As described herein, constant-region variants with reduced pI and extended half-lives would offer a more modular approach to improving the pharmacokinetic properties of antibodies.

[0137] Additional Fc variants for additional functionality In addition to pI amino acid variants, there are several useful Fc amino acid modifications that can be performed for a variety of reasons, including but not limited to modifications of binding to one or more FcγR receptors and modifications of binding to FcRn receptors.

[0138] Accordingly, the proteins of the present invention may include amino acid modifications comprising the heterodimerization variants outlined herein, including pI variants and stereovariants. Each set of variants may, independently and selectively, include or exclude any particular heterodimer protein.

[0139] FcγR variant Therefore, several useful Fc substitutions exist that can be performed to modify binding to one or more of the FcγR receptors. Substitutions that result in increased or decreased binding may be useful. For example, increased binding to FcγRIIIa is generally known to increase ADCC (antibody-dependent cell-mediated cytotoxicity; a cell-mediated reaction in which nonspecific cytotoxic cells expressing FcγR recognize a bound antibody on a target cell, subsequently causing lysis of the target cell). Similarly, under some circumstances, decreased binding to FcγRIIb (an inhibitory receptor) may also be beneficial. The amino acid substitutions used in the present invention include those enumerated in U.S. Patent Application No. 11 / 124,620 (particularly Figure 41), U.S. Patent Application No. 11 / 174,287, U.S. Patent Application No. 11 / 396,495, and U.S. Patent Application No. 11 / 538,406, all of which are expressly incorporated herein by reference in whole, and specifically with respect to the variants disclosed therein. Specific variants used include, but are not limited to, 236A, 239D, 239E, 332E, 332D, 239D / 332E, 267D, 267E, 328F, 267E / 328F, 236A / 332E, 239D / 332E / 330Y, 239D / 332E / 330L, 243A, 243L, 264A, 264V, and 299T.

[0140] In addition, there are additional Fc substitutions used to increase binding to the FcRn receptor and increase serum half-life, including but not limited to 434S, 434A, 428L, 308F, 259I, 428L / 434S, 259I / 308F, 436I / 428L, 436I or V / 434S, 436V / 428L, and 259I / 308F / 428L, as specifically disclosed in U.S. Patent Application No. 12 / 341,769, which is incorporated herein by reference in its entirety.

[0141] Removal variant Similarly, another category of functional variants is “FcγR elimination variant” or “Fc knockout (FcKO or KO)” variant. In these embodiments, for some therapeutic applications, it is desirable to reduce or eliminate the normal binding of the Fc domain to one or more or all Fcγ receptors (e.g., FcγR1, FcγRIIa, FcγRIIb, FcγRIIIa, etc.) in order to avoid additional mechanisms of action. That is, for example, in many embodiments, particularly in the use of bispecific antibodies that monovalently bind to CD3, it is generally desirable to eliminate FcγRIIIa binding in order to eliminate or significantly reduce ADCC activity, and one of the Fc domains comprises one or more Fcγ receptor elimination variants. These removal variants are shown in Figure 3, and each can be included or excluded independently and optionally. A preferred embodiment utilizes removal variants selected from the group consisting of G236R / L328R, E233P / L234V / L235A / G236del / S239K, E233P / L234V / L235A / G236del / S267K, E233P / L234V / L235A / G236del / S239K / A327G, E233P / L234V / L235A / G236del / S267K / A327G, and E233P / L234V / L235A / G236del. It should be noted that the removal variants referred to herein remove FcγR bonds but generally do not remove FcRn bonds.

[0142] As is known in the art, the Fc domain of human IgG1 has the highest binding to the Fcγ receptor, and therefore, if the constant domain (or Fc domain) of the heterodimeric antibody backbone is IgG1, the removal variant can be used. Alternatively, or in addition to the removal variant of the IgG1 background, mutations at glycosylation site 297 (generally to A or S) can significantly remove binding to FcγRIIIa, for example. Human IgG2 and IgG4 have naturally reduced binding to the Fcγ receptor, so these backbones can be used with or without the removal variant.

[0143] Combinations of heterodimers and Fc variants As will be understood by those skilled in the art, all listed heterodimerization variants (including scuba riants and / or pI variants) can be combined selectively and independently, as long as their "twist" or "monomer division" is preserved. In addition, all of these variants can be combined with any of the heterodimerization formats.

[0144] While the figure shows specific embodiments that may be used for pI variants, other combinations can be generated by following the basic rule of varying the pI difference between the two monomers to facilitate purification.

[0145] In addition, heterodimerization variants, skew, and pI can be independently and optionally combined with Fc-removal variants, Fc variants, and FcRn variants, as generally outlined herein.

[0146] H. Useful Format of the Invention As will be understood by those skilled in the art and will be discussed more fully below, the heterodimer fusion proteins of the present invention can generally take on a wide variety of configurations, as shown in Figures 13 and 42. Some figures show a “single-ended” configuration, where one “arm” of the molecule has one type of specificity and the other “arm” has a different specificity. Other figures show a “dual-ended” configuration, where the “top” of the molecule has at least one type of specificity and the “bottom” of the molecule has one or more different specificities. Accordingly, the present invention relates to novel immunoglobulin compositions for the simultaneous linkage of different first or second antigens.

[0147] As will be understood by those skilled in the art, the heterodimer format of the present invention may have different valencies and may be bispecific. That is, the heterodimer antibody of the present invention may be bivalent and bispecific, where one target tumor antigen (e.g., CD3) is bound by one binding domain and the other target tumor antigen (e.g., CLDN18.2) is bound by a second binding domain. The heterodimer antibody may also be trivalent and bispecific, where the first antigen is bound by two binding domains and the second antigen is bound by a second binding domain. As outlined herein, when CD3 is one of the target antigens, it is preferable that CD3 binds only monovalently to reduce potential side effects.

[0148] The present invention utilizes an anti-CD3 antigen-binding domain in combination with an anti-CLDN18.2 binding domain. As will be understood by those skilled in the art, any collection of anti-CD3 CDRs, anti-CD3 variable light chain domains and variable heavy chain domains, Fabs, and scFvs can be used, as shown in any of the figures. Similarly, any of the anti-CLDN18.2 antigen-binding domains can be used, and any of the CDRs, variable light chain domains and variable heavy chain domains, Fabs, and scFvs can be used selectively and independently in any combination, as shown in any of the figures (e.g., Figures 8-10).

[0149] 1+1 Fab-scFv-Fc format One heterodimer scaffold particularly used in the present invention is the “1+1 Fab-scFv-Fc” format shown in Figures 13A and 42A. In this embodiment, one heavy chain of the antibody comprises a single-stranded Fv (as defined herein as “scFv”), and the other heavy chain is a “conventional” Fab format comprising a variable heavy chain and a light chain. This structure is sometimes referred to in previous related applications as the “triple F” format (scFv-Fab-Fc) or the “bottle opener” format because it is roughly visually similar to a bottle opener. The two chains are joined together by the use of amino acid variants in constant regions (e.g., Fc domain, CH1 domain and / or hinge region) that facilitate the formation of the heterodimer antibody, as fully described below.

[0150] The current "1+1 Fab-scFv-Fc" format has several clear advantages. As is known in the art, antibody analogs relying on two scFv constructs often have stability and aggregation problems, which can be mitigated in this invention by adding "conventional" heavy-chain and light-chain pairing. In addition, in contrast to formats relying on two heavy chains and two light chains, there is no problem of incorrect pairing of heavy chains and light chains (e.g., pairing of heavy chain 1 with light chain 2).

[0151] Many of the embodiments outlined herein generally rely on a bottle-opener format antibody comprising a first monomer containing scFv, which includes variable heavy and variable light chain domains covalently linked using an scFv linker (charged in many but not all cases), where the scFv is typically covalently linked to the N-terminus of the first Fc domain via a domain linker (which may be either charged or uncharged, as outlined herein). The second monomer in bottle-opener format is a heavy chain, and the composition further comprises a light chain.

[0152] Generally, in many preferred embodiments, scFv is a domain that binds to CD3, and Fab forms a CLDN18.2 binding domain.

[0153] In addition, the Fc domain of the present invention includes a scuba riant (for example, a set of amino acid substitutions shown in Figures 1 and 4, of which particularly useful scuba riants are selected from the group consisting of S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E357L, and K370S:S364K / E357Q), optionally a removal variant (including the one shown in Figure 3), optionally a charged scFv linker (including the one shown in Figure 5), and the heavy chain includes a pI variant (including the one shown in Figure 2).

[0154] In some embodiments, the bottle opener format includes a scuba riant, a pI variant, and an excision variant. Thus, some embodiments include a) a first monomer ("scFv monomer") comprising a charged scFv linker (in some embodiments, the +H sequence in Figure 5 is preferred), a scuba riant S364K / E357Q, an excision variant E233P / L234V / L235A / G236del / S267K, and an Fv bound to CD3 as outlined herein, and b) a scuba riant L368D / K370 The bottle opener format includes a second monomer ("Fab monomer") comprising a variable heavy chain domain that constitutes Fv binding to CLDN18.2 as outlined herein, along with S, pI variants N208D / Q295E / N384D / Q418E / N421D, removal variant E233P / L234V / L235A / G236del / S267K, and a variable light chain domain, as well as c) a light chain.

[0155] Exemplary variable heavy and light chain domains of scFv that bind to CD3 are shown in Figure 12. Exemplary variable heavy and light chain domains of Fv that bind to CLDN18.2 are shown in Figure 10.

[0156] In some embodiments, the bottle opener format includes a scuba riant, a pI variant, a removal variant, and an FcRn variant. Thus, some embodiments include a) a first monomer ("scFv monomer") comprising a charged scFv linker (in some embodiments, the +H sequence in Figure 2 is preferred), a scuba riant S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, an FcRn variant M428L / N434S, and Fv bound to CD3 as outlined herein, and b) a scuba riant L368D / K370 The bottle opener format includes a second monomer ("Fab monomer") comprising a variable heavy chain domain that constitutes Fv that binds to CLDN18.2 as outlined herein, along with S, pI variant N208D / Q295E / N384D / Q418E / N421D, removal variant E233P / L234V / L235A / G236del / S267K, FcRn variant M428L / N434S, and a variable light chain domain, as well as c) a light chain.

[0157] Exemplary variable heavy and light chain domains of scFv that bind to CD3 are shown in Figure 12. Exemplary variable heavy and light chain domains of Fv that bind to CLDN18.2 are shown in Figure 10.

[0158] Figure 6 shows several exemplary bottle opener "skeleton" sequences lacking Fv sequences that can be used in the present invention. In some embodiments, any of the VH and VL sequences shown herein (including all VH and VL sequences shown in the figures and sequence listings, including those directed to CLDN18.2) can be added as a "Fab side" to the bottle opener skeleton format of Figure 6 using any of the anti-CD3 scFv sequences shown in the figures and sequence listings.

[0159] Regarding the bottle opener skeleton 1 from Figure 6 (optionally including the 428L / 434S variant), CD-binding domain sequences particularly used in these embodiments include, but are not limited to, the CD3-binding domains anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, as well as those shown in Figure 12, which are bound as the scFv side of the skeleton shown in 6.

[0160] A particularly useful combination of CLDN18.2 and CD3 sequences for use with the bottle opener skeleton 1 (optionally including the 428L / 434S variant) from Figure 6 is disclosed in Figure 31.

[0161] mAb-Fv One heterodimer scaffold particularly used in the present invention is the mAb-Fv format shown in Figure 42G. In this embodiment, the format relies on the use of C-terminal binding of an "additional" variable heavy chain domain to one monomer and C-terminal binding of an "additional" variable light chain domain to the other monomer, thereby forming a third antigen-binding domain, where the Fab portions of the two monomers bind to CLDN18.2 and the "additional" scFv domain binds to CD3.

[0162] In this embodiment, the first monomer comprises a first heavy chain comprising a first variable heavy chain domain and a first steady heavy chain domain comprising a first Fc domain, and having a first variable light chain domain covalently bonded to the C-terminus of the first Fc domain using a domain linker (VH1-CH1-hinge-CH2-CH3-[optional linker]-VL2). The second monomer comprises a second variable heavy chain domain of a second steady heavy chain domain containing a second Fc domain, and a third variable heavy chain domain covalently bonded to the C-terminus of the second Fc domain using a domain linker (VH1-CH1-hinge-CH2-CH3-[optional linker]-VH2). The two C-terminally bonded variable domains constitute an Fv that binds to CD3 (as having a divalent CD3 bond is less desirable). This embodiment further utilizes a common light chain containing a variable light chain domain and a steady light chain domain that associate with the heavy chain to form two identical Fabs that bind to CLDN18.2. With respect to many of the embodiments described herein, these constructs include scuba riants, pI variants, elimination variants, additional Fc variants, etc., as desired and described herein.

[0163] The present invention provides an mAb-Fv format in which the CD3 binding domain sequence is as shown in Figure 12. The present invention also provides an mAb-Fv format in which the CLDN18.2 binding domain sequence is as shown in Figure 10.

[0164] In addition, the Fc domain of the mAb-Fv format is a scuba rian (for example, a set of amino acid substitutions shown in Figures 1 and 4, of which particularly useful scuba rians are S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E35 The group consists of 7L, K370S:S364K / E357Q, T366S / L368A / Y407V:T366W, and T366S / L368A / Y407V / Y349C:T366W / S354C, optionally includes a removal variant (including the one shown in Figure 3), optionally includes a charged scFv linker (including the one shown in Figure 5), and the heavy chain includes a pI variant (including the one shown in Figure 2).

[0165] In some embodiments, the mAb-Fv format includes a scuba riant, a pI variant, and a removal variant. Thus, some embodiments include the mAb-Fv format, which comprises a) a first monomer including the scuba riant S364K / E357Q, the removal variant E233P / L234V / L235A / G236del / S267K, and a first variable heavy chain domain and a second variable heavy chain domain that constitute the Fv bound to CLDN18.2 together with the first variable light chain domain of the light chain, and b) the scuba riant L368D / K370S, the pI variant N208D A second monomer comprising / Q295E / N384D / Q418E / N421D, the removal variant E233P / L234V / L235A / G236del / S267K, and a first variable heavy chain domain that constitutes an Fv bound to CLDN18.2 as outlined herein, together with a first variable light chain domain, and a second variable light chain that forms an Fv(ABD) bound to CD3, together with a second variable heavy chain domain, and c) a light chain comprising a first variable light chain domain and a constant light chain domain.

[0166] In some embodiments, the mAb-Fv format includes a scuba riant, a pI variant, a removal variant, and an FcRn variant. Therefore, some embodiments include the mAb-Fv format, which comprises a) a first monomer including the scuba riant S364K / E357Q, the removal variant E233P / L234V / L235A / G236del / S267K, the FcRn variant M428L / N434S, and a first variable heavy chain domain and a second variable heavy chain domain that constitute the Fv bound to CLDN18.2 together with the first variable light chain domain of the light chain; and b) the scuba riant L368D / K370S, the pI variant N208D / Q29 A second monomer comprising a first variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, together with 5E / N384D / Q418E / N421D, removal variant E233P / L234V / L235A / G236del / S267K, FcRn variant M428L / N434S, and a first variable light chain domain, and a second variable light chain that forms an Fv(ABD) that binds to CD3, together with a second variable heavy chain domain of the first monomer, and a light chain comprising a first variable light chain domain and a constant light chain domain.

[0167] mAb-scFv A heterodimer scaffold particularly used in the present invention is the mAb-scFv format shown in Figure 42H. In this embodiment, the format relies on the use of C-terminal binding of scFv to one of the monomers, thereby forming a third antigen-binding domain, where the Fab portions of the two monomers bind to CLDN18.2, and the "additional" scFv domain binds to CD3. Thus, the first monomer comprises a first heavy chain (including a variable heavy chain domain and a constant domain) and has a C-terminally covalently bound scFv containing the scFv variable light chain domain, scFv linker, and scFv variable heavy chain domain in either direction (VH1-CH1-hinge-CH2-CH3-[optional linker]-VH2-scFv linker-VL2 or VH1-CH1-hinge-CH2-CH3-[optional linker]-VL2-scFv linker-VH2). This embodiment further utilizes a common light chain including a variable light chain domain and a constant light chain domain that associate with the heavy chain to form two identical Fabs that bind to CLDN18.2. With respect to many of the embodiments described herein, these constructs include scuba variants, pI variants, removal variants, additional Fc variants, etc., as desired and described herein.

[0168] The present invention provides an mAb-scFv format in which the CD-binding domain sequence is as shown in Figure 12 and the CLDN18.2-binding domain sequence is as shown in Figure 10.

[0169] In addition, the Fc domain of the mAb-scFv format is a scuba rian (for example, a set of amino acid substitutions shown in Figure 1, of which particularly useful scuba rians are S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E357 The linker is selected from the group consisting of L, K370S:S364K / E357Q, T366S / L368A / Y407V:T366W, and T366S / L368A / Y407V / Y349C:T366W / S354C), optionally includes a removal variant (including the one shown in Figure 3), optionally includes a charged scFv linker (including the one shown in Figure 5), and the heavy chain includes a pI variant (including the one shown in Figure 2).

[0170] In some embodiments, the mAb-scFv format includes a scubarian, a pI variant, and a removal variant. Thus, some embodiments include the mAb-scFv format, which comprises a) a first monomer including a scubarian S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, and a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with a variable light chain domain of the common light chain, and an scFv domain that binds to CD3, and b) A second monomer comprising a cubariant L368D / K370S, pI variants N208D / Q295E / N384D / Q418E / N421D, removal variants E233P / L234V / L235A / G236del / S267K, and a variable heavy chain domain constituting Fv that binds to CLDN18.2 as outlined herein, as well as a common light chain comprising a variable light chain domain and a constant light chain domain.

[0171] In some embodiments, the mAb-scFv format includes a scubarian, a pI variant, a removal variant, and an FcRn variant. Thus, some embodiments include the mAb-scFv format, which comprises a) a first monomer including a scubarian S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, an FcRn variant M428L / N434S, and a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with a variable light chain domain of a common light chain, and an scFv domain that binds to CD3, b) A second monomer comprising the scuba variant L368D / K370S, the pI variant N208D / Q295E / N384D / Q418E / N421D, the removal variant E233P / L234V / L235A / G236del / S267K, the FcRn variant M428L / N434S, and a variable heavy chain domain constituting Fv that binds to CLDN18.2 as outlined herein, along with the variable light chain of the common light chain, and a common light chain comprising a variable light chain domain and a constant light chain domain.

[0172] 2+1 Fab2-scFv-Fc format One heterodimer scaffold particularly used in the present invention is the “2+1 Fab2-scFv-Fc” format (also referred to as the “central scFv format” in a previous related application), shown in Figures 13B and 42F. In this embodiment, the format relies on the use of an inserted scFv domain, which forms a third antigen-binding domain, with the Fab portions of the two monomers binding to CLDN18.2 and the “additional” scFv domain binding to CD3. The scFv domain is inserted between the Fc domain of one of the monomers and the CH1-Fv region, thereby providing a third antigen-binding domain.

[0173] In this embodiment, one monomer comprises a first heavy chain including a first variable heavy chain domain, a CH1 domain (and an optional hinge), and an Fc domain, and has an scFv including an scFv variable light chain domain, an scFv linker, and an scFv variable heavy chain domain. The scFv is covalently bonded between the C-terminus of the CH1 domain of the heavy chain steady domain and the N-terminus of the first Fc domain using an optional domain linker (VH1-CH1-[optional linker]-VH2-scFv linker-VL2-[optional linker with hinge]-CH2-CH3, or VH1-CH1-[optional linker]-VL2-scFv linker-VH2-[optional linker with hinge]-CH2-CH3, in the opposite direction to the scFv). The other monomer is the standard Fab side. This embodiment further utilizes a common light chain including a variable light chain domain and a constant light chain domain that associate with the heavy chain to form two identical Fabs that bind to CLDN18.2. With respect to many of the embodiments described herein, these constructs include scuba variants, pI variants, removal variants, additional Fc variants, etc., as desired and described herein.

[0174] The present invention provides a "2+1 Fab2-scFv-Fc" format in which the CD3 binding domain sequence is as shown in Figure 12 and the anti-CLDN 18.2 sequence is as shown in Figure 10.

[0175] In addition, the Fc domain of the central scFv format is a scuba rian (for example, a set of amino acid substitutions shown in Figure 1, of which particularly useful scuba rians are S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E35 The group consists of 7L, K370S:S364K / E357Q, T366S / L368A / Y407V:T366W, and T366S / L368A / Y407V / Y349C:T366W / S354C, optionally includes a removal variant (including the one shown in Figure 3), optionally includes a charged scFv linker (including the one shown in Figure 5), and the heavy chain includes a pI variant (including the one shown in Figure 2).

[0176] In some embodiments, the central scFv format includes a scuba riant, a pI variant, and an excision variant. Thus, some embodiments include a) a first monomer comprising a scuba riant S364K / E357Q, an excision variant E233P / L234V / L235A / G236del / S267K, and a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with a variable light chain domain of the light chain, and an scFv domain that binds to CD3; and b) a scuba riant L368D / K370S, a pI variant The present invention comprises a second monomer comprising a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with the riant N208D / Q295E / N384D / Q418E / N421D, the removal variant E233P / L234V / L235A / G236del / S267K, and a variable light chain domain of the light chain, as well as a central scFv format comprising a light chain containing a variable light chain domain and a constant light chain domain.

[0177] In some embodiments, the central scFv format includes a scuba riant, a pI variant, an excision variant, and an FcRn variant. Thus, some embodiments include a) a first monomer comprising a scuba riant S364K / E357Q, an excision variant E233P / L234V / L235A / G236del / S267K, an FcRn variant M428L / N434S, and a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with a variable light chain domain of the light chain, and an scFv domain that binds to CD3; b) a scuba riant L368D / K370S, A second monomer comprising a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with the pI variant N208D / Q295E / N384D / Q418E / N421D, the removal variant E233P / L234V / L235A / G236del / S267K, the FcRn variant M428L / N434S, and a variable light chain of the light chain, and a central scFv format comprising a light chain comprising a variable light chain domain and a constant light chain domain.

[0178] Central Fv One heterodimer scaffold particularly used in the present invention is the central Fv format shown in Figure 42I. In this embodiment, the format relies on the use of an inserted Fv domain (i.e., a central Fv domain), so that the Fab portions of the two monomers bind to CLDN18.2 and the "central Fv domain" binds to CD3, forming a third antigen-binding domain. The scFv domain is inserted between the Fc domain and the CH1-Fv region of the monomer, so that each monomer contains components of the scFv (for example, one monomer contains a variable heavy chain domain and the other contains a variable light chain domain).

[0179] In this embodiment, one monomer comprises a first heavy chain including a first variable heavy chain domain, a CH1 domain, and an Fc domain, as well as an additional variable light chain domain. The light chain domain is covalently bonded between the C-terminus of the CH1 domain of the heavy chain steady domain and the N-terminus of the first Fc domain using a domain linker (VH1-CH1-[optional linker]-VL2-hinge-CH2-CH3). The other monomer comprises a first heavy chain including a first variable heavy chain domain, a CH1 domain, and an Fc domain, as well as an additional variable heavy chain domain (VH1-CH1-[optional linker]-VH2-hinge-CH2-CH3). The light chain domain is covalently bonded between the C-terminus of the CH1 domain of the heavy chain steady domain and the N-terminus of the first Fc domain using a domain linker.

[0180] This embodiment further utilizes a common light chain including a variable light chain domain and a constant light chain domain that associate with the heavy chain to form two identical Fabs that bind to CLDN18.2. With respect to many of the embodiments described herein, these constructs include scuba variants, pI variants, removal variants, additional Fc variants, etc., as desired and described herein.

[0181] The present invention provides a central Fv format in which the CD3 binding domain sequence is as shown in Figure 12 and the CLDN18.2 binding domain sequence is as shown in Figure 10.

[0182] Regarding the central Fv format, examples of CD3-binding domain sequences particularly used in these embodiments include, but are not limited to, anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, as shown in Figure 12.

[0183] One-arm central scFv One heterodimer scaffold particularly used in the present invention is the one-arm central scFv format shown in FIG. 42C. In this embodiment, one monomer contains only the Fc domain, while the other monomer uses the inserted scFv domain and thus forms a second antigen-binding domain. In this format, the Fab portion binds to CLDN18.2 and the scFv binds to CD3, or vice versa. The scFv domain is inserted between the CH1-Fv region and one of the Fc domains of the monomer.

[0184] In this embodiment, one monomer contains a first heavy chain comprising a first variable heavy chain domain, a CH1 domain, and an Fc domain, and the scFv contains an scFv variable light chain domain, an scFv linker, and an scFv variable heavy chain domain. The scFv is covalently bound between the C-terminus of the CH1 domain of the heavy chain constant domain and the N-terminus of the first Fc domain using a domain linker. The second monomer contains an Fc domain. This embodiment further utilizes a light chain comprising a variable light chain domain and a constant light chain domain that associates with the heavy chain to form a Fab. For many of the embodiments described herein, these constructs include scuba variants, pI variants, removal variants, additional Fc variants, etc., as desired and described herein.

[0185] The present invention provides a central Fv format in which the CD3-binding domain sequence is as shown in FIG. 12 and the CLDN18.2-binding domain sequence is as shown in FIG. 10.

[0186] In addition, the Fc domain of the one-arm central scFv format generally includes a skew variant (e.g., a set of amino acid substitutions shown in FIG. 1, and particularly useful skew variants are S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E357L, K37OS:S364K / E357Q, T366S / L368A / Y407V:T366W, and T366S / L368A / Y407V / Y349C:T366W / S354C, selected from the group consisting of), an optionally removed variant (including those shown in FIG. 3), an optionally charged scFv linker (including those shown in FIG. 5), and the heavy chain includes a pI variant (including those shown in FIG. 2).

[0187] In some embodiments, the one-arm central scFv format includes a skew variant, a pI variant, and a removed variant. Thus, some embodiments of the one-arm central scFv format include: a) a first monomer including a variable heavy chain domain that, together with a variable light chain domain of a light chain, forms an Fv that binds to CLDN18.2 as outlined herein, and an scFv domain that binds to CD3, with a skew variant SQ4K / E357Q, a removed variant E233P / L234V / L235A / G236del / S267K; b) a second monomer including an Fc domain having a skew variant L368D / K370S, a pI variant N208D / Q295E / N384D / Q418E / N421D, and a removed variant E233P / L234V / L235A / G236del / S267K; and c) a light chain including a variable light chain domain and a constant light chain domain.

[0188] In some embodiments, the one-arm central scFv format includes a scuba riant, a pI variant, a removal variant, and an FcRn variant. Thus, some embodiments of the one-arm central scFv format include a) a scuba riant S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, an FcRn variant M428L / N434S, and a variable heavy chain domain that constitutes the Fv bound to CLDN18.2 as outlined herein, along with a variable light chain domain of the light chain, and s bound to CD3. a) a first monomer containing a cFv domain, b) a second monomer containing an Fc domain having scuba riant L368D / K370S, pI variants N208D / Q295E / N384D / Q418E / N421D, removal variants E233P / L234V / L235A / G236del / S267K, and FcRn variant M428L / N434S, and c) a light chain containing a variable light chain domain and a constant light chain domain.

[0189] Regarding the one-arm central scFv format, the CD3-binding domain sequences that are particularly used include, but are not limited to, anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, as shown in Figure 12.

[0190] One-arm scFv-mAb One heterodimer scaffold particularly used in the present invention is the one-arm scFv-mAb format shown in Figure 42D. In this embodiment, one monomer contains only an Fc domain, while the other monomer uses an scFv domain bound to the N-terminus of the heavy chain, generally through the use of a linker: VH-scFv linker-VL-[optional domain linker]-CH1-hinge-CH2-CH3 or (in the opposite direction) VL-scFv linker-VH-[optional domain linker]-CH1-hinge-CH2-CH3. In this format, each Fab portion is bound to CLDN18.2, and the scFv is bound to CD3. This embodiment further utilizes a light chain containing a variable light chain domain and a constant light chain domain that associate with the heavy chain to form a Fab. With respect to many of the embodiments described herein, these constructs include scuba riants, pI variants, elimination variants, additional Fc variants, etc., as desired and described herein.

[0191] The present invention provides a one-arm scFv-mAb format in which the CD3 binding domain sequence is as shown in Figure 12 and the CLDN18.2 binding domain sequence is as shown in Figure 10.

[0192] In addition, the Fc domain of the one-arm scFv-mAb format is generally a scuba rian (for example, a set of amino acid substitutions shown in Figures 1 and 4, with particularly useful scuba rians being S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364 The linker is selected from the group consisting of K / E357L, K370S:S364K / E357Q, T366S / L368A / Y407V:T366W, and T366S / L368A / Y407V / Y349C:T366W / S354C), optionally includes a removal variant (including the one shown in Figure 3), optionally includes a charged scFv linker (including the one shown in Figure 5), and the heavy chain includes a pI variant (including the one shown in Figure 2).

[0193] In some embodiments, the one-arm scFv-mAb format includes a scuba variant, a pI variant, and a elimination variant. Accordingly, some embodiments of the one-arm scFv-mAb format include: a) a first monomer comprising a variable heavy chain domain constituting an Fv that binds to CLDN18.2 as outlined herein, along with a scuba riant S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, and a variable light chain domain of the light chain, and an scFv domain that binds to CD3; b) a second monomer comprising an Fc domain having a scuba riant L368D / K370S, a pI variant N208D / Q295E / N384D / Q418E / N421D, a removal variant E233P / L234V / L235A / G236del / S267K; and c) a light chain comprising a variable light chain domain and a constant light chain domain.

[0194] In some embodiments, the one-arm scFv-mAb format includes a scuba riant, a pI variant, a removal variant, and an FcRn variant. Thus, some embodiments of the one-arm scFv-mAb format include a) a scuba riant S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, an FcRn variant M428L / N434S, and a variable heavy chain domain constituting the Fv that binds to CLDN18.2 as outlined herein, along with a variable light chain domain of the light chain, and s that binds to CD3. a) a first monomer containing a cFv domain, b) a second monomer containing an Fc domain having scuba riant L368D / K370S, pI variants N208D / Q295E / N384D / Q418E / N421D, removal variants E233P / L234V / L235A / G236del / S267K, and FcRn variant M428L / N434S, and c) a light chain containing a variable light chain domain and a constant light chain domain.

[0195] Regarding the one-arm scFv-mAb format, the CD3-binding domain sequences that are particularly used include, but are not limited to, anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, as shown in Figure 12.

[0196] scFv-mAb A heterodimer scaffold particularly used in the present invention is the mAb-scFv format shown in Figure 42E. In this embodiment, the format relies on the use of N-terminal binding of scFv to one of the monomers, thereby forming a third antigen-binding domain, where the Fab portions of the two monomers bind to CLDN18.2, and the "additional" scFv domain binds to CD3.

[0197] In this embodiment, the first monomer includes a first heavy chain (including the variable heavy chain domain and the steady domain) having an N-terminal covalently bonded scFv containing an scFv variable light chain domain, an scFv linker, and an scFv variable heavy chain domain in either direction ((VH1-scFv linker-VL1-[optional domain linker]-VH2-CH1-hinge-CH2-CH3) or (with the scFv in the opposite direction) ((VL1-scFv linker-VH1-[optional domain linker]-VH2-CH1-hinge-CH2-CH3)). This embodiment further utilizes a common light chain including the variable light chain domain and the steady light chain domain that associate with the heavy chain to form two identical Fabs that bind to CLDN18.2. With respect to many of the embodiments described herein, these constructs include scubarians, pI variants, removal variants, additional Fc variants, etc., as desired and described herein.

[0198] The present invention provides an scFv-mAb format in which the CD3 binding domain sequence is as shown in Figure 12 and the CLDN18.2 binding domain sequence is as shown in Figure 10.

[0199] In addition, the Fc domain of the scFv-mAb format is a scuba rian (for example, a set of amino acid substitutions shown in Figure 1, of which particularly useful scuba rians are S364K / E357Q:L368D / K370S, L368D / K370S:S364K, L368E / K370S:S364K, T411T / E360E / Q362E:D401K, L368D / K370S:S364K / E357 The linker is selected from the group consisting of L, K370S:S364K / E357Q, T366S / L368A / Y407V:T366W, and T366S / L368A / Y407V / Y349C:T366W / S354C), optionally includes a removal variant (including the one shown in Figure 3), optionally includes a charged scFv linker (including the one shown in Figure 5), and the heavy chain includes a pI variant (including the one shown in Figure 2).

[0200] In some embodiments, the scFv-mAb format includes a scubarian, a pI variant, and a removal variant. Accordingly, some embodiments include the scFv-mAb format, which comprises a) a first monomer including a scubarian S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, and a variable heavy chain domain constituting Fv bound to CLDN18.2 as outlined herein, along with a variable light chain domain of the common light chain, and an scFv domain bound to CD3, and b) A second monomer comprising a cubariant L368D / K370S, pI variants N208D / Q295E / N384D / Q418E / N421D, removal variants E233P / L234V / L235A / G236del / S267K, and a variable heavy chain domain constituting Fv that binds to CLDN18.2 as outlined herein, as well as a common light chain comprising a variable light chain domain and a constant light chain domain.

[0201] In some embodiments, the scFv-mAb format includes a scubarian, a pI variant, a removal variant, and an FcRn variant. Thus, some embodiments include the scFv-mAb format, which comprises a) a first monomer including a scubarian S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, an FcRn variant M428L / N434S, and a variable heavy chain domain constituting Fv that binds to CLDN18.2 as outlined herein, along with a variable light chain domain of a common light chain, and an scFv domain that binds to CD3, b A second monomer comprising a variable heavy chain domain constituting Fv that binds to CLDN18.2 as outlined herein, along with the scuba variant L368D / K370S, pI variant N208D / Q295E / N384D / Q418E / N421D, removal variant E233P / L234V / L235A / G236del / S267K, FcRn variant M428L / N434S, and a variable light chain of the common light chain, as well as a common light chain comprising a variable heavy chain domain constituting Fv that binds to CLDN18.2 as outlined herein, and a common light chain comprising a variable light chain domain and a constant light chain domain.

[0202] For mAb-scFv format skeleton 1 from Figure 10 (optionally including M428L / N434S), the CD3 binding domain sequences particularly used include, but are not limited to, anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, as shown in Figure 12.

[0203] Dual scFv format The present invention also provides a dual scFv format, which is known in the art and is shown in Figure 42B. In this embodiment, the CLDN18.2×CD3 heterodimer bispecific antibody is composed of two scFv-Fc monomers (both in either the (VH-scFv linker-VL-[optional domain linker]-CH2-CH3) format or the (VL-scFv linker-VH-[optional domain linker]-CH2-CH3) format, or one monomer in one direction and the other in the other direction).

[0204] The present invention provides a dual scFv format in which the CD3 binding domain sequence is as shown in Figure 12 and the CLDN18.2 binding domain sequence is as shown in Figure 10.

[0205] In some embodiments, the dual scFv format includes a scubarian, a pI variant, and a removal variant. Accordingly, some embodiments include a dual scFv format that includes a) a first monomer comprising a first scFv bound to either CD3 or CLDN18.2, a scubarian S364K / E357Q, a removal variant E233P / L234V / L235A / G236del / S267K, and b) a second monomer comprising a second scFv bound to either CD3 or CLDN18.2, a scubarian L368D / K370S, a pI variant N208D / Q295E / N384D / Q418E / N421D, a removal variant E233P / L234V / L235A / G236del / S267K, and b) a second scFv bound to either CD3 or CLDN18.2.

[0206] In some embodiments, the dual scFv format includes a skew variant, a pI variant, a clearance variant, and an FcRn variant. In some embodiments, the dual scFv format includes a skew variant, a pI variant, and a clearance variant. Thus, some embodiments include a) a first monomer comprising a first scFv that binds to either CD3 or CLDN18.2, the skew variant S364K / E357Q, the clearance variant E233P / L234V / L235A / G236del / S267K, the FcRn variant M428L / N434S, and b) a second monomer comprising a second scFv that binds to either CD3 or CLDN18.2, the skew variant L368D / K370S, the pI variant N208D / Q295E / N384D / Q418E / N421D, the clearance variant E233P / L234V / L235A / G236del / S267K, the FcRn variant M428L / N434S, including a dual scFv format.

[0207] Regarding the dual scFv format, CD3 binding domain sequences particularly used include anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, and those shown in FIG. 12, but are not limited thereto.

[0208] I. Antigen-binding domain to the target antigen The bispecific antibodies of the present invention have two distinct antigen-binding domains (ABDs) that bind to two different target checkpoint antigens ("target pairs") in either a bivalent or trivalent bispecific format, as generally shown in Figure 1. It should be noted that these bispecific antibodies are generally named "anti-CLDN18.2 × anti-CD3," or, more commonly, simplified or for ease of use (and therefore interchangeable), "CLDN18.2 × CD3" for each pair. It should be noted that, unless otherwise specified herein, the order of the antigen list in the name does not indicate the structure. That is, a CLDN18.2 × CD3 bottle-opener antibody can bind scFv to CLDN18.2 or CD3, but in some cases, the order indicates the structure.

[0209] As will be outlined more fully herein, these combinations of ABDs can be in various formats, generally consisting of combinations where one ABD is in Fab format and the other is in scFv format, as outlined below. Some formats used in this specification and shown in Figure 42 use a single Fab and a single scFv (Figures 42A, C, and D), while others use two Fabs and a single scFv (Figures 42E, F, and I).

[0210] antigen-binding domain As discussed herein, the heterodimer antibodies in question contain two antigen-binding domains (ABDs), each bound to CLDN18.2 or CD3. As outlined herein, these heterodimer antibodies may be bispecific and bivalent (each antigen is bound by a single ABD in the format shown, for example, in Figure 42A) or bispecific and trivalent (for example, one antigen is bound by a single ABD and the other by two ABDs, as shown in Figure 42F).

[0211] In addition, generally, one of the ABDs contains scFv as outlined herein, in a direction from the N-terminus to the C-terminus of a VH-scFv linker-VL or a VL-scFv linker-VH. According to the format, one or both of the other ABDs are generally Fabs containing a VH domain on one protein chain (generally as a component of the heavy chain) and a VL domain on the other protein chain (generally as a component of the light chain).

[0212] The present invention provides several ABDs that bind to several different checkpoint proteins, as outlined below. As will be understood by those skilled in the art, any set of six CDRs or the VH and VL domains may be in scFv format or Fab format, which are then attached to the heavy chain constant domain and the light chain constant domain, the heavy chain constant domain containing variants (including within the CH1 and Fc domains). The scFv sequences included in the sequence listing utilize a specific charged linker, but uncharged or other charged linkers may be used, including those shown in Figure 5, as outlined herein.

[0213] In addition, as discussed above, the numbering used in sequence listings for CDR identification is Kabat, but different numbering can be used, which will alter the amino acid sequence of the CDR, as shown in Table 1.

[0214] Further variants can be created for all variable heavy and light chain domains described herein. As outlined herein, in some embodiments, a set of six CDRs may have 0, 1, 2, 3, 4, or 5 amino acid modifications (including amino acid substitutions used in particular), and variations in the framework region of the variable heavy and light chain domains may be present, provided that this framework (excluding the CDRs) maintains at least about 80, 85, or 90% identity with a human germline sequence selected from those enumerated in Figure 1 of U.S. Patent No. 7,657,380 (its figures and descriptions are incorporated herein by reference in their entirety). Thus, for example, the same CDR described herein can be combined with different framework sequences derived from human germline sequences, provided that the framework region maintains at least 80, 85, or 90% identity with a human germline sequence selected from those enumerated in Figure 1 of U.S. Patent No. 7,657,380. Alternatively, CDRs may have amino acid modifications (for example, one, two, three, four, or five amino acid modifications in a set of CDRs (i.e., CDRs may be modified as long as the total number of changes in a set of six CDRs is less than six amino acid modifications, any combination of CDRs may be changed, e.g., one change in VLCDR1, two changes in VHCDR2, no change in VHCDR3, etc.)), and variations in the framework region may exist as long as the framework region maintains at least 80, 85, or 90% identity with a human germline sequence selected from those enumerated in Figure 1 of U.S. Patent No. 7,657,380.

[0215] CLDN18.2 antigen-binding domain In some embodiments, one of the ABDs binds to CLDN18.2. A preferred set of six CDRs and / or VH and VL domains is shown in Figure 10.

[0216] As will be understood by those skilled in the art, preferred CLDN18.2 binding domains may include a set of six CDRs as shown in the figure, either as those underlined or as CDRs identified using other alignments in the VH and VL sequences of those shown in Figure 10, when different numbering schemes are used as described herein and shown in Table 1. Preferred ABDs may also include these sequences and the entire VH and VL sequences shown in the figure, used as scFv or Fab. In many of the embodiments herein involving Fv to CLDN18.2, it is the Fab monomer that binds to CLDN18.2.

[0217] In addition to the parent CDR sets disclosed in the Figures and Sequence Listings that form ABDs for CLDN18.2, the present invention provides variant CDR sets. In one embodiment, a set of six CDRs may have one, two, three, four, or five amino acid changes from the parent CDRs, as long as the CLDN18.2 ABD can still bind to the target antigen when measured by at least one of the following assays (the latter being particularly used in many embodiments): Biacore, surface plasmon resonance (SPR), and / or BLI (biolayer interferometry, e.g., Octet assay).

[0218] In addition to the parent variable heavy and variable light chain domains disclosed herein that form ABD for CLDN18.2, the present invention provides variant VH and VL domains. In one embodiment, the variant VH and VL domains may each have 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 amino acid changes from the parent VH and VL domains, as long as the ABD can still bind to the target antigen when measured by at least one of Biacore, surface plasmon resonance (SPR), and / or BLI (biolayer interferometry, e.g., Octet assay) assays (the latter being particularly used in many embodiments). In another embodiment, variants VH and VL are at least 90, 95, 97, 98, or 99% identical to their respective parental VH and VL domains, insofar as ABD can still bind to the target antigen, when measured by at least one of Biacore, surface plasmon resonance (SPR), and / or BLI (biolayer interferometry, e.g., Octet assay) assays (the latter being particularly used in many embodiments).

[0219] A specific preferred embodiment includes H1L1 and H2L1 CLDN18.2 antigen-binding domains as "Fab" contained within one of the bottle opener skeletons shown in Figure 6.

[0220] A specific preferred embodiment includes H1L1 and H2L1 CLDN18.2 antigen-binding domains as "Fab" contained within one of the 2+1 format skeletons of Figure 41.

[0221] CD3 antigen-binding domain In some embodiments, one of the ABDs binds to CD3. A preferred set of six CDR and / or VH and VL domains, as well as scFv sequences, are shown in Figures 12 and 13 and the sequence listing. Particularly useful CD3-binding domain sequences include, but are not limited to, anti-CD3 H1.30_L1.47, anti-CD3 H1.32_L1.47, anti-CD3 H1.89_L1.47, anti-CD3 H1.90_L1.47, anti-CD3 H1.33_L1.47, and anti-CD3 H1.31_L1.47, as shown in Figure 12.

[0222] As will be understood by those skilled in the art, preferred CD3-binding domains may include a set of six CDRs, as shown in Figure 12, either as those underlined or as CDRs identified using other alignments in the VH and VL sequences of those shown in Figure 12, when different numbering schemes are used as described herein and shown in Table 1. Preferred ABDs may also include these sequences and the entire VH and VL sequences shown in the figure, used as scFv or Fab. In many of the embodiments herein involving Fv to CD3, it is the scFv monomer that binds to CD3.

[0223] In addition to the parent CDR sets disclosed in the Figures and Sequence Listings that form ABDs for CD3, the present invention provides variant CDR sets. In one embodiment, a set of six CDRs may have one, two, three, four, or five amino acid changes from the parent CDRs, as long as the CD3 ABD can still bind to the target antigen when measured by at least one of the following assays (the latter being particularly used in many embodiments): Biacore, surface plasmon resonance (SPR), and / or BLI (biolayer interferometry, e.g., Octet assay).

[0224] In addition to the parent variable heavy and variable light chain domains disclosed herein that form ABD to CD3, the present invention provides variant VH and VL domains. In one embodiment, each of the variant VH and VL domains may have 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 amino acid changes from the parent VH and VL domains, insofar as the ABD can still bind to the target antigen when measured by at least one of Biacore, surface plasmon resonance (SPR), and / or BLI (biolayer interferometry, e.g., Octet assay) assays (the latter being particularly used in many embodiments). In another embodiment, the variant VH and VL are at least 90, 95, 97, 98, or 99% identical to their respective parent VH and VL domains, insofar as the ABD can still bind to the target antigen when measured by at least one of Biacore, surface plasmon resonance (SPR), and / or BLI (biolayer interferometry, e.g., Octet assay) assays (the latter being particularly used in many embodiments).

[0225] J. Useful Embodiments In one embodiment, a particular combination of scubarian and pI variant used in the present invention is T366S / L368A / Y407V:T366W (optionally containing a crosslinked disulfide, T366S / L368A / Y407V / Y349C:T366W / S354C), where one monomer contains Q295E / N384D / Q418E / N481D and the other is a positively charged domain linker (with a format containing scFv domains). As understood in the art, the "knob-in-hole" variant does not alter the pI and therefore can be used with any monomer.

[0226] K. Nucleic acid of the present invention The present invention further provides nucleic acid compositions encoding anti-CLDN18.2 antibodies provided herein, including but not limited to anti-CLDN18.2 × anti-CD3 bispecific antibodies and CLDN18.2 monospecific antibodies.

[0227] As will be understood by those skilled in the art, nucleic acid compositions depend on the format and scaffold of the heterodimeric protein. Therefore, if three amino acid sequences are required for a format, such as the 1+1 Fab-scFv-Fc format (e.g., a first amino acid monomer containing the Fc domain and scFv, and a second amino acid monomer containing the heavy and light chains), then three nucleic acid sequences can be incorporated into one or more expression vectors for expression. Similarly, in some formats (e.g., the dual scFv format disclosed in Figure 1), only two nucleic acids are required, and again, they can be incorporated into one or two expression vectors.

[0228] As is known in the art, the nucleic acids encoding the components of the present invention can be incorporated into an expression vector, as is known in the art, depending on the host cell used to produce the heterodimeric antibody of the present invention. Generally, the nucleic acids are operably ligated to any number of regulatory elements (promoters, origins of replication, selectable markers, ribosome binding sites, inducers, etc.). The expression vector may be an extrachromosomal vector or an embedded vector.

[0229] Subsequently, the nucleic acids and / or expression vectors of the present invention are transformed into any number of different types of host cells known in the art, including mammalian, bacterial, yeast, insect, and / or fungal cells, with mammalian cells (e.g., CHO cells) being used in many embodiments.

[0230] In some embodiments, the nucleic acids encoding each monomer, and an optional nucleic acid encoding the light chain, applicable depending on the format, are generally contained within a single expression vector under different or the same promoter control. In embodiments particularly useful in the present invention, each of these two or three nucleic acids is contained on a different expression vector. Different vector ratios can be used to promote heterodimerization, as shown herein and incorporated herein by reference in 62 / 025,931. That is, surprisingly, the protein contains the first monomer:second monomer:light chain in a ratio of 1:1:2 (in many embodiments herein having three polypeptides containing heterodimeric antibodies), but these are not the ratios that yield the best results.

[0231] The heterodimer antibodies of the present invention are produced by culturing host cells containing an expression vector, as is well known in the art. Once produced, conventional antibody purification processes, including ion-exchange chromatography, are carried out. As discussed herein, a difference of at least 0.5 pI between the two monomers can enable separation by ion-exchange chromatography, isoelectric focusing, or other isoelectric-sensitive methods. That is, isoelectric focusing purification of "1+1Fab-scFv-Fc" and "2+1" heterodimers (e.g., anion-exchange columns, cation-exchange columns) is facilitated by pI substitutions that alter the pI of each monomer so that each monomer has a different isoelectric point (pI) and the heterodimer also has a distinctly different pI. These substitutions are also useful for determining and monitoring contaminated dual scFv-Fc and mAb homodimers after purification (e.g., IEF gels, cIEF, and analytical IEX columns).

[0232] Biological and biochemical functions of L. heterodimer bispecific antibodies Generally, the bispecific CLDN18.2×CD3 antibody of the present invention is administered to patients with cancer, and its efficacy is evaluated in several ways, as described herein. For this reason, standard efficacy assays may be performed, such as assessment of tumor volume, tumor size, and the presence or extent of metastasis, but immuno-oncological treatment may also be evaluated based on an assessment of the immune status. This can be done in several ways, including both in vitro and in vivo assays.

[0233] M.Treatment Once prepared, the antibody composition of the present invention can be used in several applications. CLDN18.2 is highly expressed in gastric tumors. Therefore, the heterodimer composition of the present invention can be used to treat such CLDN18.2-positive cancers.

[0234] Antibody composition for in vivo administration Antibody formulations used in accordance with the present invention are prepared for storage by mixing an antibody of desired purity with an optional pharmaceutically acceptable carrier, excipient, or stabilizer, in the form of a lyophilized formulation or aqueous solution (Remington's Pharmaceutical Sciences 16th edition, Osol, A. Ed.

[1980] ). Acceptable carriers, excipients, or stabilizers are nontoxic to the recipient at the dose and concentration used and include buffers such as phosphates, citrates, and other organic acids, antioxidants including ascorbic acid and methionine, preservatives (e.g., octadecyldimethylbenzylammonium chloride, hexamethonium chloride, benzalkonium chloride, benzethonium chloride, phenol, butyl or benzyl alcohol, alkylparabens such as methyl or propylparaben, catechol, resorcinol, cyclohexanol, 3-pentanol, and m-cresol), low molecular weight (less than about 10 residues) polypeptides, serum albumin This material contains proteins such as cellulose, gelatin, or immunoglobulins; hydrophilic polymers such as polyvinylpyrrolidone; amino acids such as glycine, glutamine, asparagine, histidine, arginine, or lysine; monosaccharides, disaccharides, and other carbohydrates including glucose, mannose, or dextrin; chelating agents such as EDTA; sugars such as sucrose, mannitol, trehalose, and sorbitol; salt-forming counterions such as sodium; metal complexes (e.g., Zn-protein complexes); and / or nonionic surfactants such as TWEEN®, PLURONICS®, or polyethylene glycol (PEG).

[0235] Administration modality The antibodies and chemotherapeutic agents of the present invention are administered to subjects by known methods, such as intravenous administration as a bolus or by continuous infusion over a period of time.

[0236] Therapeutic modalities In the method of the present invention, the treatment is used to provide a positive therapeutic response with respect to a disease or condition. A “positive therapeutic response” is intended to be an improvement in the disease or condition and / or an improvement in symptoms associated with the disease or condition. For example, a positive therapeutic response would refer to one or more of the following improvements in the disease: (1) a decrease in the number of tumor cells, (2) an increase in tumor cell death, (3) an inhibition of tumor cell survival, (5) an inhibition of tumor growth (i.e., a slowdown, preferably a cessation), (6) an increase in patient survival rate, and (7) some relief from one or more symptoms associated with the disease or condition.

[0237] A positive treatment response in any given disease or condition may be determined by standardized response criteria specific to that disease or condition. Tumor response may be assessed for changes in tumor morphology (i.e., systemic tumor tissue volume, tumor size, etc.) using screening techniques such as magnetic resonance imaging (MRI) scans, radiography, computed tomographic (CT) scans, bone scans, endoscopy, and tumor biopsy sampling, including bone marrow aspiration (BMA) and counting of circulating tumor cells.

[0238] In addition to these positive treatment responses, patients receiving treatment may experience beneficial effects such as improvement in disease-related symptoms.

[0239] The treatment according to the present invention includes a pharmaceutical drug used in a "therapeutic dose." A "therapeutic dose" refers to an amount that is effective in the required dosage and duration to achieve the desired treatment outcome.

[0240] The therapeutically effective dose may vary depending on factors such as the individual's disease state, age, sex, and weight, as well as the drug's ability to induce the desired response in the individual. The therapeutically effective dose is also the amount in which the therapeutically beneficial effects outweigh any toxic or harmful effects of either the antibody or its moiety.

[0241] The "therapeutic effective dose" for tumor therapy can also be measured by its ability to stabilize disease progression. The ability of cancer-inhibiting compounds can be evaluated in animal model systems to predict efficacy in human tumors.

[0242] Alternatively, this property of a composition can be evaluated by testing the compound's ability to inhibit cell proliferation or induce apoptosis using in vitro assays known to those skilled in the art. A therapeutically effective amount of a therapeutic compound may reduce tumor size or otherwise alleviate the symptoms of the target. Those skilled in the art will be able to determine such an amount based on factors such as the size of the target, the severity of the symptoms of the target, and the specific composition or route of administration selected.

[0243] The drug regimen is adjusted to provide the optimal desired response (e.g., therapeutic response). For example, a single bolus may be administered, several divided doses may be administered over time, or the dose may be increased or decreased proportionally as indicated by the urgency of the treatment situation. Parenteral compositions may be formulated in unit dosage forms for ease of administration and uniformity of dosage. As used herein, unit dosage forms refer to physically distinct units suitable as a single dose for the target of treatment. Each unit contains a predetermined amount of the active compound calculated to produce the desired therapeutic effect in relation to the required pharmaceutical carrier.

[0244] The specifications of the unit dosage form of the present invention are determined by and directly depend on (a) the inherent characteristics of the active compound and the specific therapeutic effect to be achieved, and (b) the limitations that accompany the field in which such active compounds are formulated for the treatment of susceptibility in an individual.

[0245] The efficient dosage and administration regimen of the bispecific antibody used in this invention depend on the disease or condition being treated and can be determined by those skilled in the art.

[0246] The exemplary and non-limiting therapeutic dose range of the bispecific antibody used in this invention is approximately 0.1 to 100 mg / kg.

[0247] All cited references are explicitly incorporated herein by reference in their entirety.

[0248] While specific embodiments of the present invention have been described above for illustrative purposes, it will be understood by those skilled in the art that numerous modifications of detail can be made without departing from the present invention as described in the appended claims.

[0249] 2. Examples Examples are provided below to illustrate the present invention. These examples are not intended to limit the present invention to any particular use or operating theory. For all constant region positions considered in the present invention, the numbering follows the EU index, as in the case of Kabat (Kabat et al., 1991, Sequences of Proteins of Immunological Interest, 5th Ed., United States Public Health Service, National Institutes of Health, Bethesda, incorporated in whole by reference). Those skilled in the art of antibodies will understand that this rule consists of discontinuous numbering in specific regions of immunoglobulin sequences, enabling a normalization criterion to conserved positions in the immunoglobulin family. Thus, the position of any give...

Claims

1. A claudin 18 isoform A2 splice variant (CLDN18.2) antigen-binding domain comprising a variable heavy chain domain having the amino acid sequence of SEQ ID NO: 81 and a variable light chain domain having the amino acid sequence of SEQ ID NO:

84.

2. A nucleic acid composition, a) A first nucleic acid encoding the variable heavy chain domain as described in claim 1, b) A nucleic acid composition comprising a second nucleic acid encoding the variable light chain domain described in claim 1.

3. Expression vector composition, a) A first expression vector comprising the first nucleic acid described in claim 2, b) An expression vector composition comprising a second expression vector comprising the second nucleic acid described in claim 2.

4. A host cell comprising the nucleic acid composition described in claim 2 or the expression vector composition described in claim 3.

5. A method for producing a CLDN18.2 antigen-binding domain, comprising: culturing the host cells described in claim 4 under conditions in which the CLDN18.2 antigen-binding domain is expressed; and recovering the CLDN18.2 antigen-binding domain.

6. An antibody comprising a claudin 18 isoform A2 splice variant (CLDN18.2) antigen-binding domain, which includes a variable heavy chain domain having the amino acid sequence of SEQ ID NO: 81 and a variable light chain domain having the amino acid sequence of SEQ ID NO:

84.

7. A nucleic acid composition, a) A first nucleic acid encoding the variable heavy chain domain as described in claim 6, b) A nucleic acid composition comprising a second nucleic acid encoding the variable light chain domain described in claim 6.

8. Expression vector composition, a) A first expression vector comprising the first nucleic acid described in claim 7, b) An expression vector composition comprising a second expression vector comprising the second nucleic acid described in claim 7.

9. A host cell comprising the nucleic acid composition according to claim 7 or the expression vector composition according to claim 8.

10. A method for producing an antibody containing a CLDN18.2 antigen-binding domain, comprising: culturing host cells according to claim 9 under conditions in which the antibody containing the CLDN18.2 antigen-binding domain is expressed; and recovering the antibody containing the CLDN18.2 antigen-binding domain.

Images