HETERODIMERIC ANTIBODIES THAT BIND CD3 AND GPC3

SEQUENCE LISTING

The instant application contains a Sequence Listing which has been submitted electronically in XML file format and is hereby incorporated by reference in its entirety. Said XML copy, created on Jul. 25, 2023, is named 067461-5282-US01.xml and is 838,123 bytes in size.

SUMMARY

Provided herein are novel GPC3 binding domains, and antibodies that include such GPC3 binding domains (e.g., anti-GPC3×anti-CD3). Also provided herein are methods of using such antibodies for the treatment of GPC3-associated cancers.

In a first aspect, provided herein is a composition comprising a GPC3 antigen binding domain (ABD). The GPC3 binding domain comprises a set of 6 CDRs (vhCDR1, vhCDR2, vhCDR3, vlCDR1, vlCDR2 and vlCDR3) from a variable heavy domain (VH)/variable light domain (VL) pair selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

In some embodiments, the ABD has a VH/VL pair selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

In some embodiments, the VH/VL pair is selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69. In exemplary embodiments, the composition is a monoclonal antibody.

In another aspect, provided herein is a heterodimeric antibody comprising: a) a first monomer, b) a second monomer, and c) a light chain. The first monomer comprises: i) an anti-CD3 scFv comprising a first variable light domain, an scFv linker and a first variable heavy domain; and ii) a first Fc domain, wherein the scFv is covalently attached to the N-terminus of the first Fc domain using a domain linker. The second monomer comprises a VH2-CH1-hinge-CH2-CH3 monomer, wherein VH is a second variable heavy domain and CH2-CH3 is a second Fc domain. In this embodiment, the second variable heavy domain and the second variable light domain form a GPC3 antigen binding domain (ABD).

In some embodiments, the GPC3 binding domain comprises a set of 6 CDRs (vhCDR1, vhCDR2, vhCDR3, vlCDR1, vlCDR2 and vlCDR3) from a VH/VL pair selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

In exemplary embodiments, the GPC3 binding domain comprises a VH/VL pair selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

In exemplary embodiments, the GPC3 binding domain comprises a VH/VL pair selected from the group consisting of: [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69.

In some embodiments, the anti-CD3 scFv comprises a VH and VL pair selected from the group consisting of: H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31.

In some embodiments of the heterodimeric antibody, the scFv linker is a charged scFv linker.

In some embodiments, the first and second Fc domains are variant Fc domains. In some embodiments, the first and second Fc domains comprise a set of heterodimerization variants selected from the group consisting of those depicted in FIGS. 1A-IE. In some embodiments, the set of heterodimerization variants selected is from the group consisting of S364K/E357Q: L368D/K370S; S364K: L368D/K370S; S364K: L368E/K370S; D401K: T411E/K360E/Q362E; and T366W: T366S/L368A/Y407V, wherein numbering is according to EU numbering.

In some embodiments, the first and second monomers further comprise one or more ablation variants. In some embodiments, the one or more ablation variants are E233P/L234V/L235A/G236del/S267K, wherein numbering is according to EU numbering.

In exemplary embodiments, one of the first or second monomer comprises one or more pI variants. In some embodiments, the one or more pI variants are N208D/Q295E/N384D/Q418E/N421D, wherein numbering is according to EU numbering.

In exemplary embodiments of the heterodimeric antibody, the first monomer comprises amino acid variants S364K/E357Q/E233P/L234V/L235A/G236del/S267K, the second monomer comprises amino acid variants L368D/K370S/N208D/Q295E/N384D/Q418E/N421D/E233P/L234V/L235A/G236del/S267K, and wherein numbering is according to EU numbering.

In some embodiments, the first and second monomers each further comprise amino acid variants 428/434S, wherein numbering is according to EU numbering.

In another aspect, provided herein is a heterodimeric antibody comprising: a) a first monomer comprising, from N-terminal to C-terminal, a scFv-linker-CH2-CH3, wherein scFv is an anti-CD3 scFv and CH2-CH3 is a first Fc domain; b) a second monomer comprising, from N-terminal to C-terminal, a VH-CH1-hinge-CH2-CH3, wherein CH2-CH3 is a second Fc domain; and c) a light chain comprising VL-CL. The first variant Fc domain comprises amino acid variants S364K/E357Q, the second variant Fc domain comprises amino acid variants L368D/K370S, the first and second variant Fc domains each comprises amino acid variants E233P/L234V/L235A/G236del/S267K, and the CH1-hinge-CH2-CH3 of the second monomer comprises amino acid variants N208D/Q295E/N384D/Q418E/N421D (EU numbering). Further, the VH and VL form an GPC3 binding domain comprising the variable heavy domain and the variable light domain, respectively, of a GPC3 binding domain selected from [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69; and the anti-CD3 scFv comprises the variable heavy domain and the variable light domain of a CD3 binding domain selected from H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31.

In some embodiments, the first and second variant Fc domains each further comprise amino acid variants 428/434S, wherein numbering is according to EU numbering.

In another aspect, provided herein is a heterodimeric antibody comprising: a) a first monomer, b) a second monomer, and c) a common light chain. The first monomer comprises, from N-terminal to C-terminal, a VH1-CH1-linker 1-scFv-linker 2-CH2-CH3, wherein VH1 is a first variable heavy domain, scFv is an anti-CD3 scFv, linker 1 and linker 2 are a first domain linker and second domain linker, respectively, and CH2-CH3 is a first Fc domain. The second monomer comprises, from N-terminal to C-terminal, a VH2-CH1-hinge-CH2-CH3, wherein VH2 is a second variable heavy domain and CH2-CH3 is a second Fc domain; and c) a common light chain comprising a variable light domain. The first variable heavy domain and the variable light domain form a first CLDN6 ABD, and the second variable heavy domain and the variable light domain form a second CLDN6 ABD.

In another aspect, provided herein is a heterodimeric antibody comprising: a) a first monomer, b) a second monomer, and c) a common light chain. The first monomer comprises, from N-terminal to C-terminal, a VH1-CH1-linker 1-scFv-linker 2-CH2-CH3, wherein VH1 is a first variable heavy domain, scFv is an anti-CD3 scFv, linker 1 and linker 2 are a first domain linker and second domain linker, respectively, and CH2-CH3 is a first Fc domain. The second monomer comprises, from N-terminal to C-terminal, a VH2-CH1-hinge-CH2-CH3, wherein VH2 is a second variable heavy domain and CH2-CH3 is a second Fc domain; and c) a common light chain comprising a variable light domain. The first variable heavy domain and the variable light domain form a first GPC3 ABD, and the second variable heavy domain and the variable light domain form a second GPC3 ABD.

In some embodiments, the first and second GPC3 binding domains each comprise a set of 6 CDRs (vhCDR1, vhCDR2, vhCDR3, vlCDR1, vlCDR2 and vlCDR3) from a VH/VL pair selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

In some embodiments, each of the first and second GPC3 binding domains have a VH/VL pair selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73. In some embodiments, the VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69.

In some embodiments, the scFv comprises has a set of 6 CDRs (vhCDR1, vhCDR2, vhCDR3, vlCDR1, vlCDR2 and vlCDR3) from VH/VL pairs selected from the group consisting of: H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31. In some embodiments, the scFv comprises the variable heavy domain and variable light domain of any of the following CD3 binding domains: H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31.

In some embodiments of the heterodimeric antibody, the scFv linker is a charged scFv linker. In some embodiments, the scFv linker is a charged scFv linker having the amino acid sequence (GKPGS)₄(SEQ ID NO: 1).

In some embodiments, the first and second Fc domains are variant Fc domains. In some embodiments, the first and second Fc domains comprise a set of heterodimerization variants selected from the group consisting of those depicted in FIGS. 1A-1E. In some embodiments, the set of heterodimerization variants selected is from the group consisting of S364K/E357Q: L368D/K370S; S364K: L368D/K370S; S364K: L368E/K370S; D401K: T411E/K360E/Q362E; and T366W: T366S/L368A/Y407V, wherein numbering is according to EU numbering.

In some embodiments, the first and second monomers each further comprise amino acid variants 428/434S, wherein numbering is according to EU numbering.

45. A heterodimeric antibody according to 44, wherein the first and second variant Fc domains each further comprise amino acid variants 428/434S, wherein numbering is according to EU numbering.

In another aspect, provided herein is heterodimeric antibody comprising: a) a first monomer, b) a second monomer, and c) a common light chain. The first monomer comprises, from N-terminal to C-terminal, a VH1-CH1-linker 1-scFv-linker 2-CH2-CH3, wherein scFv is an anti-CD3 scFv and CH2-CH3 is a first Fc domain. The second monomer comprises, from N-terminal to C-terminal, a VH1-CH1-hinge-CH2-CH3, wherein CH2-CH3 is a second Fc domain. The common light chain comprising VL-CL. The first variant Fc domain comprises amino acid variants S364K/E357Q, the second variant Fc domain comprises amino acid variants L368D/K370S, the first and second variant Fc domains each comprises amino acid variants E233P/L234V/L235A/G236del/S267K, and the CH1-hinge-CH2-CH3 of the second monomer comprises amino acid variants N208D/Q295E/N384D/Q418E/N421D (EU numbering). The VH and VL comprise the variable heavy domain and the variable light domain of a GPC3 ABD selected from H1.9_L1.187, H1.24_L1.187, H2.91_L1.187 and H1.9_L1.187; and the anti-CD3 scFv comprises the variable heavy domain and the variable light domain of a CD3 binding domain selected from H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31. In some embodiments, the first and second variant Fc domains each further comprise amino acid variants 428/434S.

Also provided herein are nucleic acid compositions comprising nucleic acids encoding the antibodies described herein, expression vector compositions that include such nucleic acids, host cells for making the antibodies that comprise the expression vector compositions, and methods of making the antibodies.

BACKGROUND

Antibody-based therapeutics have been used successfully to treat a variety of diseases, including cancer. An increasingly prevalent avenue being explored is the engineering of single immunoglobulin molecules that co-engage two different antigens. Such alternate antibody formats that engage two different antigens are often referred to as bispecific antibodies. Because the considerable diversity of the antibody variable region (Fv) makes it possible to produce an Fv that recognizes virtually any molecule, the typical approach to bispecific antibody generation is the introduction of new variable regions into the antibody.

A particularly useful approach for bispecific antibodies is to engineer a first binding domain which engages CD3 and a second binding domain which engages an antigen associated with or upregulated on cancer cells so that the bispecific antibody redirects CD3+ T cells to destroy the cancer cells. Heparan sulfate proteoglycan family member Glypican 3 (GPC3) has been previously reported to be highly expressed in hepatocellular carcinoma (particularly of the liver), and minimally expressed in healthy tissue. In view of this, it is believed that anti-GPC3 antibodies are useful, for example, for localizing anti-tumor therapeutics (e.g., chemotherapeutic agents and T cells) to such GPC3 expressing tumors. Previous attempts to target GPC3 such as CD3 bispecific antibodies and CAR-T therapies have shown some success but have demonstrated adverse events such as cytokine release syndrome in preclinical or clinical settings. The present invention provides novel bispecific antibodies to CD3 and GPC3 that are capable of localizing CD3+ effector T cells to GPC3 expressing tumors with improved therapeutic profile.

BRIEF DESCRIPTION OF THE DRAWINGS

FIGS. 1A-1E depict useful pairs of Fc heterodimerization variant sets (including skew and pI variants) that lead to Fc heterodimerization. There are variants for which there are no corresponding “monomer 2” variants; these are pI variants which can be used alone on either monomer.

FIG. 2 depicts a list of isosteric variant antibody constant regions and their respective substitutions. pI_(−) indicates lower pI variants, while pI_(+) indicates higher pI variants. These can be optionally and independently combined with other heterodimerization variants of the inventions (and other variant types as well, as outlined herein.)

FIG. 3 depicts useful ablation variants that ablate FcγR binding (sometimes referred to as “knock outs” or “KO” variants). Generally, ablation variants are found on both monomers, although in some cases they may be on only one monomer

FIG. 4 depicts particularly useful embodiments of “non-Fv” components of the invention.

FIG. 5 depicts a number of charged scFv linkers that find use in increasing or decreasing the pI of the subject heterodimeric bsAbs that utilize one or more scFv as a component, as described herein. The (+H) positive linker finds particular use herein, particularly with anti-CD3 V_Land V_Hsequences shown herein. A single prior art scFv linker with a single charge is referenced as “Whitlow”, from Whitlow et al., Protein Engineering 6(8):989-995 (1993). It should be noted that this linker was used for reducing aggregation and enhancing proteolytic stability in scFvs. Such charged scFv linkers can be used in any of the subject antibody formats disclosed herein that include scFvs (e.g., 1+1 Fab-scFv-Fc and 2+1 Fab₂-scFv-Fc formats).

FIG. 6 depicts a number of exemplary domain linkers. In some embodiments, these linkers find use linking a single-chain Fv to an Fc chain. In some embodiments, these linkers may be combined. For example, a GGGGS linker (SEQ ID NO: 2) may be combined with a “half hinge” linker.

FIG. 7A-7D depicts the sequences of several useful 1+1 Fab-scFv-Fc bispecific antibody format heavy chain backbones based on human IgG1, without the Fv sequences (e.g. the scFv and the VH for the Fab side). Backbone 1 is based on human IgG1 (356E/358M allotype), and includes the S364K/E357Q: L368D/K370S skew variants, C220S on the chain with the S364K/E357Q skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 2 is based on human IgG1 (356E/358M allotype), and includes S364K: L368D/K370S skew variants, C220S on the chain with the S364K skew variant, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants, and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 3 is based on human IgG1 (356E/358M allotype), and includes S364K: L368E/K370S skew variants, C220S on the chain with the S364K skew variant, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368E/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 4 is based on human IgG1 (356E/358M allotype), and includes D401K: K360E/Q362E/T411E skew variants, C220S on the chain with the D401K skew variant, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with K360E/Q362E/T411E skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 5 is based on human IgG1 (356D/358L allotype), and includes S364K/E357Q: L368D/K370S skew variants, C220S on the chain with the S364K/E357Q skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 6 is based on human IgG1 (356E/358M allotype), and includes S364K/E357Q: L368D/K370S skew variants, C220S on the chain with the S364K/E357Q skew variants, N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains, as well as an N297A variant on both chains. Backbone 7 is identical to 6 except the mutation is N297S. Backbone 8 is based on human IgG4, and includes the S364K/E357Q: L368D/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants, as well as a S228P (EU numbering, this is S241P in Kabat) variant on both chains that ablates Fab arm exchange as is known in the art. Backbone 9 is based on human IgG2, and includes the S364K/E357Q: L368D/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants. Backbone 10 is based on human IgG2, and includes the S364K/E357Q: L368D/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants as well as a S267K variant on both chains. Backbone 11 is identical to backbone 1, except it includes M428L/N434S Xtend mutations. Backbone 12 is based on human IgG1 (356E/358M allotype), and includes S364K/E357Q: L368D/K370S skew variants, C220S and the P217R/P229R/N276K pI variants on the chain with S364K/E357Q skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Included within each of these backbones are sequences that are 90, 95, 98 and 99% identical (as defined herein) to the recited sequences, and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 additional amino acid substitutions (as compared to the “parent” of the Figure, which, as will be appreciated by those in the art, already contain a number of amino acid modifications as compared to the parental human IgG1 (or IgG2 or IgG4, depending on the backbone). That is, the recited backbones may contain additional amino acid modifications (generally amino acid substitutions) in addition to the skew, pI and ablation variants contained within the backbones of this figure.

FIGS. 8A-8C depict the sequences of several useful 2+1 Fab₂-scFv-Fc bispecific antibody format heavy chain backbones based on human IgG1, without the Fv sequences (e.g. the scFv and the VH for the Fab side). Backbone 1 is based on human IgG1 (356E/358M allotype), and includes the S364K/E357Q: L368D/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 2 is based on human IgG1 (356E/358M allotype), and includes S364K: L368D/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants, and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 3 is based on human IgG1 (356E/358M allotype), and includes S364K: L368E/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368E/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 4 is based on human IgG1 (356E/358M allotype), and includes D401K: K360E/Q362E/T411E skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with K360E/Q362E/T411E skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 5 is based on human IgG1 (356D/358L allotype), and includes S364K/E357Q: L368D/K370S skew variants, the N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Backbone 6 is based on human IgG1 (356E/358M allotype), and includes S364K/E357Q: L368D/K370S skew variants, N208D/Q295E/N384D/Q418E/N421D pI variants on the chain with L368D/K370S skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains, as well as an N297A variant on both chains. Backbone 7 is identical to 6 except the mutation is N297S. Backbone 8 is identical to backbone 1, except it includes M428L/N434S Xtend mutations. Backbone 9 is based on human IgG1 (356E/358M allotype), and includes S364K/E357Q: L368D/K370S skew variants, the P217R/P229R/N276K pI variants on the chain with S364K/E357Q skew variants and the E233P/L234V/L235A/G236del/S267K ablation variants on both chains. Included within each of these backbones are sequences that are 90, 95, 98 and 99% identical (as defined herein) to the recited sequences, and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 additional amino acid substitutions (as compared to the “parent” of the Figure, which, as will be appreciated by those in the art, already contain a number of amino acid modifications as compared to the parental human IgG1 (or IgG2 or IgG4, depending on the backbone). That is, the recited backbones may contain additional amino acid modifications (generally amino acid substitutions) in addition to the skew, pI and ablation variants contained within the backbones of this figure.

FIG. 9 depicts the sequences of several useful constant light domain backbones based on human IgG1, without the Fv sequences (e.g. the scFv or the Fab). Included herein are constant light backbone sequences that are 90, 95, 98 and 99% identical (as defined herein) to the recited sequences, and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 additional amino acid modifications.

FIG. 10A-10F depicts sequences for exemplary anti-CD3 scFvs suitable for use in the bispecific antibodies of the invention. The CDRs are underlined, the scFv linker is double underlined (in the sequences, the scFv linker is a positively charged scFv (GKPGS)₄linker (SEQ ID NO: 15), although as will be appreciated by those in the art, this linker can be replaced by other linkers, including uncharged or negatively charged linkers, some of which are depicted in FIG. 5), and the slashes indicate the border(s) of the variable domains. In addition, the naming convention illustrates the orientation of the scFv from N- to C-terminus. As noted herein and is true for every sequence herein containing CDRs, the exact identification of the CDR locations may be slightly different depending on the numbering used as is shown in Table 2, and thus included herein are not only the CDRs that are underlined but also CDRs included within the V_Hand V_Ldomains using other numbering systems. Furthermore, as for all the sequences in the Figures, these V_Hand V_Lsequences can be used either in a scFv format or in a Fab format.

FIG. 11A-11B depicts the antigen sequences for a number of antigens of use in the invention, including both human and cyno, to facilitate the development of antigen binding domains that bind to both for ease of clinical development.

FIG. 12 depicts the variable heavy and variable light chain sequences for humanized GPC3-A variants. CDRs are underlined and slashes indicate the border(s) between the variable regions and constant domain. As noted herein and is true for every sequence herein containing CDRs, the exact identification of the CDR locations may be slightly different depending on the numbering used as is shown in Table 2, and thus included herein are not only the CDRs that are underlined but also CDRs included within the V_Hand V_Ldomains using other numbering systems. Further, as for all the sequences in the Figures, these V_Hand V_Lsequences can be used either in a scFv format or in a Fab format. Furthermore, each of the variable heavy domains depicted herein can be paired with any other αGPC3-A variable light domain; and each of the variable light domains depicted herein can be paired with any other αGPC3-A variable heavy domain.

FIG. 13A-13G depicts the variable heavy and variable light chain sequences for GPC3-A variants engineered for reduced degradation (e.g. aspartic acid isomerization and deamidation) liability, modulated GPC3 binding affinity, and/or selectivity for high GPC3 expression cell lines. CDRs are underlined and slashes indicate the border(s) between the variable regions and constant domain. As noted herein and is true for every sequence herein containing CDRs, the exact identification of the CDR locations may be slightly different depending on the numbering used as is shown in Table 2, and thus included herein are not only the CDRs that are underlined but also CDRs included within the V_Hand V_Ldomains using other numbering systems. Further, as for all the sequences in the Figures, these V_Hand V_Lsequences can be used either in a scFv format or in a Fab format. Furthermore, each of the variable heavy domains depicted herein can be paired with any other αGPC3 variable light domain; and each of the variable light domains depicted herein can be paired with any other αGPC3 variable heavy domain.

FIG. 14A-14E depicts illustrative GPC3-A variants formatted as bivalent anti-GPC3 mAbs and IgG1 backbone with E233P/L234V/L235A/G236del/S267K ablation variant. CDRs are underlined and slashes indicate the border(s) between the variable regions and constant domain. As noted herein and is true for every sequence herein containing CDRs, the exact identification of the CDR locations may be slightly different depending on the numbering used as is shown in Table 2, and thus included herein are not only the CDRs that are underlined but also CDRs included within the VH and VL domains using other numbering systems. Furthermore, as for all the sequences in the Figures, these VH and VL sequences can be used either in a scFv format or in a Fab format.

FIG. 15A-15B depicts a couple of formats of the present invention. FIG. 15A depicts the “1+1 Fab-scFv-Fc” format, with a first Fab arm binding GPC3 and a second scFv arm binding CD3. FIG. 15B depicts the “2+1 Fab₂-scFv-Fc” format, with a first Fab arm binding GPC3 and a second Fab-scFv arm, wherein the Fab binds GPC3 and the scFv binds CD3.

FIG. 16A-16C depicts the sequences for illustrative αGPC3×αCD3 bsAbs in the 1+1 Fab-scFv-Fc format and comprising a CD3 High scFv (H1.30_L1.47 in either the VHVL orientation or the VLVH orientation). CDRs are underlined and slashes indicate the border(s) between the variable regions and other chain components (e.g. constant region and domain linkers). It should be noted that the αGPC3×αCD3 bsAbs can utilize variable region, Fc region, and constant domain sequences that are 90, 95, 98 and 99% identical (as defined herein), and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid substitutions. In addition, each sequence outlined herein can include or exclude the M428L/N434S variants in one or preferably both Fc domains, which results in longer half-life in serum.

FIG. 17A-17G depicts the sequences for illustrative αGPC3×αCD3 bsAbs in the 1+1 Fab-scFv-Fc format and comprising a CD3 High-Int #1 scFv (H1.32_L1.47 in either the VHVL orientation or the VLVH orientation). CDRs are underlined and slashes indicate the border(s) between the variable regions and other chain components (e.g. constant region and domain linkers). It should be noted that the αGPC3×αCD3 bsAbs can utilize variable region, Fc region, and constant domain sequences that are 90, 95, 98 and 99% identical (as defined herein), and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid substitutions. In addition, each sequence outlined herein can include or exclude the M428L/N434S variants in one or preferably both Fc domains, which results in longer half-life in serum.

FIG. 18A-18B depicts the sequences for illustrative αGPC3×αCD3 bsAbs in 2+1 Fab₂-scFv-Fc format and comprising a CD3 High scFv (H1.30_L1.47 in either the VHVL orientation or the VLVH orientation). CDRs are underlined and slashes indicate the border(s) between the variable regions and other chain components (e.g. constant region and domain linkers). It should be noted that the αGPC3×αCD3 bsAbs can utilize variable region, Fc region, and constant domain sequences that are 90, 95, 98 and 99% identical (as defined herein), and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid substitutions. In addition, each sequence outlined herein can include or exclude the M428L/N434S variants in one or preferably both Fc domains, which results in longer half-life in serum.

FIG. 19A-19H depicts the sequences for illustrative αGPC3×αCD3 bsAbs in the 2+1 Fab₂-scFv-Fc format and comprising a CD3 High-Int #1 scFv (H1.32_L1.47 in either the VHVL orientation or the VLVH orientation). CDRs are underlined and slashes indicate the border(s) between the variable regions and other chain components (e.g. constant region and domain linkers). It should be noted that the αGPC3×αCD3 bsAbs can utilize variable region, Fc region, and constant domain sequences that are 90, 95, 98 and 99% identical (as defined herein), and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid substitutions. In addition, each sequence outlined herein can include or exclude the M428L/N434S variants in one or preferably both Fc domains, which results in longer half-life in serum

FIG. 20A-20E depicts the sequences for illustrative αGPC3×αCD3 bsAbs in the 2+1 Fab₂-scFv-Fc format and comprising a CD3 High-Int #2 scFv (H1.89_L1.47 in either the VHVL orientation or the VLVH orientation). CDRs are underlined and slashes indicate the border(s) between the variable regions and other chain components (e.g. constant region and domain linkers). It should be noted that the αGPC3×αCD3 bsAbs can utilize variable region, Fc region, and constant domain sequences that are 90, 95, 98 and 99% identical (as defined herein), and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid substitutions. In addition, each sequence outlined herein can include or exclude the M428L/N434S variants in one or preferably both Fc domains, which results in longer half-life in serum.

FIG. 21 depicts sequences for illustrative αGPC3-A×αCD3 bsAbs in the 2+1 Fab₂-scFv-Fc format and comprising a CD3 Intermediate scFv (H1.33_L1.47 in either the VHVL orientation or the VLVH orientation). CDRs are underlined and slashes indicate the border(s) between the variable regions and other chain components (e.g. constant region and domain linkers). It should be noted that the αGPC3-A×αCD3 bsAbs can utilize variable region, Fc region, and constant domain sequences that are 90, 95, 98 and 99% identical (as defined herein), and/or contain from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid substitutions. In addition, each sequence outlined herein can include or exclude the M428L/N434S variants in one or preferably both Fc domains, which results in longer half-life in serum.

FIG. 22 depicts sequences based on a comparator αGPC3×αCD3 bispecific antibody as disclosed in WO 2016/047722.

FIG. 23 depicts binding affinity of 5 GPC-A affinity variants H1.1_L1.6, H1.1_L1.29, H1.1_L1.16, H1.1_L1.23 and H1.1_L1.31 (formatted as 1+1 Fab-scFv-Fc with CD3 High-Int #1 scFv) to human and cynomolgus GPC3 antigen. The affinity variants demonstrated a broad range of binding (4-400 nM KD) and were cross-reactive for cynomolgus GPC3.

FIG. 24 depicts binding of 3 GPC-A affinity variants H1.1_L1.29, H1.1_L1.16, and H1.1_L1.31 (formatted as 1+1 Fab-scFv-Fc with CD3 High-Int #1 scFv) to HepG2 (GPC3^high) cells.

FIG. 25 depicts binding of GPC3-A variants engineered to remove degradation liabilities to human GPC3-A. Several of the variants demonstrated reduced binding affinity to GPC3.

FIG. 26 depicts binding of GPC3-A variants engineered to remove degradation liabilities to human and cynomolgus GPC3-A. Several of the variants demonstrated reduced, biphasic, or abrogated binding to GPC3

FIG. 27 depicts A) induction of RTCC and B) activation of T cells (as indicated by IFNγ secretion) by αGPC3×αCD3 bsAb comparator XENP31308. XENP31308 induces potent RTCC and activation of T cells in the presence of each of GPC3^highHepG2, GPC3^medHuh7, and GPC3^lowHEK293 cells (10:1 effector:target ratio).

FIG. 28 depicts A) induction of RTCC and B) activation of T cells (as indicated by IFNγ secretion) by αGPC3×αCD3 bsAb comparator XENP31308 at a 1:1 effector:target ratio.

FIG. 29 depicts activation of T cells (as indicated by IFNγ release) in the presence of HepG2 (GPC3^low) cells by GPC3 binding domains binding different GPC3 epitopes formatted as 2+1 Fab₂-scFv-Fc bsAbs with CD3 High-Int #1. Each of the bsAbs having GPC3 binding domains other than GPC3-A activated T cells less potently than bsAb having GPC3-A binding domain.

FIG. 30 depicts binding of A) GPC3-A H1.1_L1.16 (100 nM) and B) GPC3-A H1.1_L1.31 (400 nM) variants to HepG2 (GPC3^high) in the context of 1+1 Fab-scFv-Fc bsAb, 2+1 Fab₂-scFv-Fc bsAb, and bivalent monospecific mAb. BsAbs in the 2+1 format demonstrated more potent binding in comparison to bsAbs in the 1+1 Format (and comparable binding to the bivalent monospecific mAb). Potency shift from 1+1 format to 2+1 format was more pronounced for the lower binding affinity GPC3-A_H1.1_L1.31 (400 nM) variant.

FIG. 31 depicts A) induction of RTCC and B) cytokine release by T cells following incubation of HepG2 (GPC3^high) cells with T cells (10:1 effector:target ratio) and αGPC3×αCD3 bsAbs in the 1+1 Fab-scFv-Fc format with different CD3 binding affinities. Detuning the CD3 binding affinity from CD3 High to CD3 High-Int #1 provide substantial reduction in potency of RTCC, and more importantly, IFNγ release to potentially mitigate cytokine release syndrome.

FIG. 32 depicts A) induction of RTCC and B) cytokine release by T cells following incubation of HepG2 (GPC3^high) cells with T cells (10:1 effector:target ratio) and αGPC3×αCD3 bsAbs in the 2+1 Fab₂-scFv-Fc format with different CD3 binding affinities. Detuning the CD3 binding affinity from CD3 High to CD3 High-Int #1 did not provide substantial reduction in potency, but further reduction to CD3 High-Int #2 did provide substantial reduction in potency.

FIG. 33 depicts A) induction of RTCC and B) cytokine secretion in the presence of HepG2 (GPC3^high) cells by 3 GPC3-A affinity variants formatted as 2+1 Fab₂-scFv-Fc with CD3 High-Int #2[VLVH] scFv. The bsAbs induced RTCC in a manner correlating with their reduced affinities, with XENP37625 having the 70 nm GPC3 affinity variant demonstrating the strongest potency, and XENP37626 having the 400 nM GPC3 affinity variant demonstrating the weakest potency. Surprisingly, the H1.1_L1.23 200 nM affinity variant did not appear to differentiate significantly from the 400 nM affinity variant.

FIG. 34 depicts induction of RTCC on A) HepG2 and B) HEK293 cells by additional αGPC3×αCD3 bsAbs of the invention (XENP36935, XENP37430, XENP34364, XENP34920, XENP35843, XENP36939, and XENP37433) at a 1:1 effector:target ratio.

FIG. 35 depicts induction of IFNγ release by T cells in the presence of A) HepG2 and B) HEK293 cells and additional αGPC3×αCD3 bsAbs of the invention (XENP36935, XENP37430, XENP34364, XENP34920, XENP35843, XENP36939, and XENP37433) at a 1:1 effector:target ratio. XENP37430 and XENP34920 demonstrated the highest potency on HepG2, while XENP36939 had the lowest potency. The potency in inducing IFNγ release by each bsAb was proportional to their potency in RTCC induction. Most of the bsAbs selectivity induced killing (and cytokine release) in the presence of HepG2 (GPC^high) over HEK293 (GPC^low).

FIG. 36 depicts induction of RTCC on HepG2 by αGPC3×αCD3 bsAbs of the invention (XENP36935, XENP37430, XENP37625, XENP35843, XENP34920, XENP37624, XENP36939, XENP37433, and XENP37626) and comparator bsAb XENP31308 at a 10:1 effector:target ratio. BsAbs XENP34920, XENP37430, XENP37625, and XENP37624 demonstrated similar potency (less than EC50 of 100 ng/ml) as comparator XENP31308. Other bsAbs XENP35843, XENP36935, XENP37433, and XENP37626 were much less potent but were able to achieve efficacious killing at higher concentrations. XENP36939 did not demonstrate any killing.

FIG. 37 depicts induction of RTCC on HepG2 by αGPC3×αCD3 bsAbs of the invention (XENP36935, XENP37430, XENP37625, XENP35843, XENP34920, XENP37624, XENP36939, XENP37433, and XENP37626) and comparator bsAb XENP31308 at a 1:1 effector:target ratio.

FIG. 38 depicts induction of RTCC on Huh7 by αGPC3×αCD3 bsAbs of the invention (XENP36935, XENP37430, XENP37625, XENP35843, XENP34920, XENP37624, XENP36939, XENP37433, and XENP37626) and comparator bsAb XENP31308 at a 10:1 effector:target ratio.

FIG. 39 depicts the growth of GPC3high HepG2 tumors over time in a 2D Incucyte model. A 10 ug/ml treatment concentration of each test article and a 10:1 E:T ratio were used. As seen in the other Incucyte studies, XENP36939 did not show any significant activity, with an efficacy similar to the PBS control. However, most other test including XENP34920, XENP37430, XENP36935, XENP37624, and XENP37625 showed a similar efficacy as the comparator XENP31308.

FIG. 40 depicts the growth of GPC3^medHuh7 tumor cells over time in a 2D Incucyte model. A 10 ug/ml treatment concentration of each test article was used. None of the bsAbs having the 400 nM GPC3 binder (XENP36939, XENP37433, or XENP37626) were able to significantly inhibit cell growth. However, 2+1 bsAbs having either the 70 nM or 100 nM GPC3 binders were all able to prevent growth with a similar efficacy as the comparator.

FIG. 41 depicts the results of a 3D spheroid RTCC model using Incucyte. 1,000 HepG2 cells (which have a doubling time of 24 hr) are seeded in a well and given 72-96 hours to grow, and then 40,000 PBMCs are added (resulting in an E:T ratio of approximately 3:1) along with the indicated bsAb at a concentration of 10 ug/ml. The intensity of each signal (tumor or immune cell) was integrated over each well. The curves for XENP34920 and XENP37624 were more similar than expected, given that XENP37625 has a lower affinity CD3 arm (High-Int #2) compared to XENP36935 (High-In #1). After 144 hours all bsAbs but XENP36939 show 100% efficacy.

FIG. 42 depicts the results of a 3D spheroid RTCC model using Incucyte. 1,000 HepG2 cells (which have a doubling time of 24 hr) are seeded in a well and given 72-96 hours to grow, and then 40,000 PBMCs are added (resulting in an E:T ratio of approximately 3:1) along with the indicated bsAb at a concentration of 10 ug/ml. The intensity of each signal (tumor or immune cell) was integrated over each well. Interestingly, one of the weaker affinity 400 nM GPC3 binders showed the highest level of proliferation. With the exception of the 1+1 constructs XENP35843 and XENP36939, the rest of the bsAbs showed a similar or higher level of immune cell proliferation than the comparator XENP31308.

FIG. 43 depicts photographs taken of the results of a 3D spheroid RTCC model using Incucyte. 1,000 HepG2 cells (which have a doubling time of 24 hr) are seeded in a well and given 72-96 hours to grow, and then 40,000 PBMCs are added (resulting in an E:T ratio of approximately 3:1) along with the indicated bsAb at a concentration of 10 ug/ml. The target tumor cell spheroid can be most clearly seen in the PBS control column, where PBMCs are clustered around the spheroid and remain unchanged over time. In comparing the PBS control column with the columns treated by XENP31308, XENP37625, XENP34920, and XENP36935, the expansion of the PBMCs as the immune cells kill the tumor cells can be visualized.

FIGS. 44A to 44J depict several formats of the present invention. The first is the 1+1 Fab-scFv-Fc format, with a first and a second anti-antigen binding domain. Additionally, mAb-Fv, mAb-scFv, Central-scFv, Central-Fv, one-armed central-scFv, one scFv-mAb, scFv-mAb and a dual scFv format are all shown. For all of the scFv domains depicted, they can be either N- to C-terminus variable heavy-(optional linker)-variable light, or the opposite. In addition, for the one-armed scFv-mAb, the scFv can be attached either to the N-terminus of a heavy chain monomer or to the N-terminus of the light chain.

DETAILED DESCRIPTION

The present invention provides heterodimeric bispecific antibodies that bind to human CD3ε and human GPC3.

A. Overview

Anti-bispecific antibodies that co-engage CD3 and a tumor antigen target are used to redirect T cells to attack and lyse targeted tumor cells. Examples include the BiTE® and DART formats, which monovalently engage CD3 and a tumor antigen. While the CD3-targeting approach has shown considerable promise, a common side effect of such therapies is the associated production of cytokines, often leading to toxic cytokine release syndrome. Because the anti-CD3 binding domain of the bispecific antibody engages all T cells, the high cytokine-producing CD4 T cell subset is recruited. Moreover, the CD4 T cell subset includes regulatory T cells, whose recruitment and expansion can potentially lead to immune suppression and have a negative impact on long-term tumor suppression. In addition, these formats do not contain Fc domains and show very short serum half-lives in patients.

Provided herein are novel anti-CD3×anti-GPC3 (also referred to as anti-GPC3×anti-CD3, αCD3×αGPC3, αGPC3×αCD3 or sometimes just GPC3×CD3) heterodimeric bispecific antibodies and methods of using such antibodies for the treatment of cancers. In particular, provided herein are anti-CD3, anti-GPC3 bispecific antibodies in a variety of formats. These bispecific antibodies are useful for the treatment of cancers, particularly those with increased GPC3 expression such as renal cell carcinoma. Such antibodies are used to direct CD3+ effector T cells to GPC3+ tumors, thereby allowing the CD3+ effector T cells to attack and lyse the GPC3+ tumors.

Additionally, in some embodiments, the disclosure provides bispecific antibodies that have different binding affinities to human CD3 that can alter or reduce the potential side effects of anti-CD3 therapy. That is, in some embodiments the antibodies described herein provide antibody constructs comprising anti-CD3 antigen binding domains that are “strong” or “high affinity” binders to CD3 (e.g. one example are heavy and light variable domains depicted as H1.30_L1.47 (optionally including a charged linker as appropriate)) and also bind to GPC3. In other embodiments, the antibodies described herein provide antibody constructs comprising anti-CD3 antigen binding domains that are “lite” or “lower affinity” binders to CD3. Additional embodiments provides antibody constructs comprising anti-CD3 antigen binding domains that have intermediate or “medium” affinity to CD3 that also bind to GPC3. While a very large number of anti-CD3 antigen binding domains (ABDs) can be used, particularly useful embodiments use 6 different anti-CD3 ABDs, although they can be used in two scFv orientations as discussed herein. Affinity is generally measured using a Biacore assay.

It should be appreciated that the “high, medium, low” anti-CD3 sequences provided herein can be used in a variety of heterodimerization formats as discussed herein. In general, due to the potential side effects of T cell recruitment, exemplary embodiments utilize formats that only bind CD3 monovalently, such as depicted in FIGS. 15A and 15B, and in the formats depicted herein, it is the CD3 ABD that is a scFv as more fully described herein. In contrast, the subject bispecific antibodies can bind GPC3 either monovalently (e.g. FIG. 15A) or bivalently (e.g. FIG. 15B).

Provided herein are compositions that include GPC3 binding domains, including antibodies with such GPC3 binding domains (e.g., GPC3×CD3 bispecific antibodies). Subject antibodies that include such GPC3 binding domains advantageously elicit a range of different immune responses, depending on the particular GPC3 binding domain used. For example, the subject antibodies exhibit differences in selectivity for cells with different GPC3 expression, potencies for GPC3 expressing cells, ability to elicit cytokine release, and sensitivity to soluble GPC3. Such GPC3 binding domains and related antibodies find use, for example, in the treatment of GPC3 associated cancers.

Accordingly, in one aspect, provided herein are heterodimeric antibodies that bind to two different antigens, e.g. the antibodies are “bispecific”, in that they bind two different target antigens, generally GPC3 and CD3 as described herein. These heterodimeric antibodies can bind these target antigens either monovalently (e.g. there is a single antigen binding domain such as a variable heavy and variable light domain pair) or bivalently (there are two antigen binding domains that each independently bind the antigen). In some embodiments, the heterodimeric antibody provided herein includes one CD3 binding domain and one GPC3 binding domain (e.g., heterodimeric antibodies in the “1+1 Fab-scFv-Fc” format described herein). In other embodiments, the heterodimeric antibody provided herein includes one CD3 binding domain and two GPC3 binding domains (e.g., heterodimeric antibodies in the “2+1 Fab2-scFv-Fc” formats described herein). The heterodimeric antibodies provided herein are based on the use different monomers which contain amino acid substitutions that “skew” formation of heterodimers over homodimers, as is more fully outlined below, coupled with “pI variants” that allow simple purification of the heterodimers away from the homodimers, as is similarly outlined below. The heterodimeric bispecific antibodies provided generally rely on the use of engineered or variant Fc domains that can self-assemble in production cells to produce heterodimeric proteins, and methods to generate and purify such heterodimeric proteins.

B. Nomenclature

The antibodies provided herein are listed in several different formats. In some instances, each monomer of a particular antibody is given a unique “XENP” number, although as will be appreciated in the art, a longer sequence might contain a shorter one. For example, a “scFv-Fc” monomer of a 1+1 Fab-scFv-Fc format antibody may have a first XENP number, while the scFv domain itself will have a different XENP number. Some molecules have three polypeptides, so the XENP number, with the components, is used as a name. Thus, the molecule XENP33744, which is in 2+1 Fab2-scFv-Fc format, comprises three sequences (see FIG. 18) a “Fab-Fc Heavy Chain” monomer; 2) a “Fab-scFv-Fc Heavy Chain” monomer; and 3) a “Light Chain” monomer or equivalents, although one of skill in the art would be able to identify these easily through sequence alignment. These XENP numbers are in the sequence listing as well as identifiers, and used in the Figures. In addition, one molecule, comprising the three components, gives rise to multiple sequence identifiers. For example, the listing of the Fab includes, the full heavy chain sequence, the variable heavy domain sequence and the three CDRs of the variable heavy domain sequence, the full light chain sequence, a variable light domain sequence and the three CDRs of the variable light domain sequence. A Fab-scFv-Fc monomer includes a full length sequence, a variable heavy domain sequence, 3 heavy CDR sequences, and an scFv sequence (include scFv variable heavy domain sequence, scFv variable light domain sequence and scFv linker). Note that some molecules herein with a scFv domain use a single charged scFv linker (+H), although others can be used. In addition, the naming nomenclature of particular antigen binding domains (e.g., GPC3 and CD3 binding domains) use a “Hx.xx_Ly.yy” type of format, with the numbers being unique identifiers to particular variable chain sequences. Thus, an Fv domain of the antigen binding domain is “H1 L1”, which indicates that the variable heavy domain, H1, was combined with the light domain L1. In the case that these sequences are used as scFvs, the designation “H1 L1”, indicates that the variable heavy domain, H1 is combined with the light domain, L1, and is in VH-linker-VL orientation, from N- to C-terminus. This molecule with the identical sequences of the heavy and light variable domains but in the reverse order (VL-linker-VH orientation, from N- to C-terminus) would be designated “L1_H1.1”. Similarly, different constructs may “mix and match” the heavy and light chains as will be evident from the sequence listing and the figures.

Additionally, the bispecific antibodies of the invention are referred to herein as “anti-CD3×anti-GPC3”, “αCD3×αGPC3”, “αGPC3×αCD3” or sometimes just “GPC3×CD3”. The order of the antigens is not determinative as will be discussed below, although the majority of the formats that utilize as scFv have the an anti-CD3 ABD as the scFv.

C. Definitions

In order that the application may be more completely understood, several definitions are set forth below. Such definitions are meant to encompass grammatical equivalents.

By “GPC3” herein is meant a protein belonging to the claudin family. GPC3 sequences are depicted, for example, in FIG. 11. The ABDs of the invention bind to human GPC3.

By “ablation” herein is meant a decrease or removal of activity. Thus for example, “ablating FcγR binding” means the Fc region amino acid variant has less than 50% starting binding as compared to an Fc region not containing the specific variant, with more than 70-80-90-95-98% loss of activity being preferred, and in general, with the activity being below the level of detectable binding in a Biacore, SPR or BLI assay. Of particular use in the ablation of FcγR binding are those shown in FIG. 3, which generally are added to both monomers.

By “ADCC” or “antibody dependent cell-mediated cytotoxicity” as used herein is meant the cell-mediated reaction wherein nonspecific cytotoxic cells that express FcγRs recognize bound antibody on a target cell and subsequently cause lysis of the target cell. ADCC is correlated with binding to FcγRIIIa; increased binding to FcγRIIIa leads to an increase in ADCC activity.

By “ADCP” or antibody dependent cell-mediated phagocytosis as used herein is meant the cell-mediated reaction wherein nonspecific phagocytic cells that express FcγRs recognize bound antibody on a target cell and subsequently cause phagocytosis of the target cell.

As used herein, term “antibody” is used generally. Antibodies described herein can take on a number of formats as described herein, including traditional antibodies as well as antibody derivatives, fragments and mimetics, including a number of bispecific formats described herein.

Traditional immunoglobulin (Ig) antibodies are “Y” shaped tetramers. Each tetramer is typically composed of two identical pairs of polypeptide chains, each pair having one “light chain” monomer (typically having a molecular weight of about 25 kDa) and one “heavy chain” monomer (typically having a molecular weight of about 50-70 kDa).

Other useful antibody formats include, but are not limited to, the 1+1 Fab-scFv-Fc format and 2+1 Fab-scFv-Fc antibody formats described herein and depicted in FIG. 15, as well as “mAb-Fv,” “mAb-scFv,” “central-Fv”, “one-armed scFv-mAb,” “scFv-mAb,” “dual scFv,” and “trident” format antibodies, as discussed below and shown in FIG. 44.

Antibody heavy chains typically include a variable heavy (VH) domain, which includes vhCDR1-3, and an Fc domain, which includes a CH2-CH3 monomer. In some embodiments, antibody heavy chains include a hinge and CH1 domain. Traditional antibody heavy chains are monomers that are organized, from N- to C-terminus: VH-CH1-hinge-CH2-CH3. The CH1-hinge-CH2-CH3 is collectively referred to as the heavy chain “constant domain” or “constant region” of the antibody, of which there are five different categories or “isotypes”: IgA, IgD, IgG, IgE and IgM. Thus, “isotype” as used herein is meant any of the subclasses of immunoglobulins defined by the chemical and antigenic characteristics of their constant regions. It should be understood that therapeutic antibodies can also comprise hybrids of isotypes and/or subclasses. For example, as shown in US Publication 2009/0163699, incorporated by reference, the antibodies described herein include the use of human IgG1/G2 hybrids.

In some embodiments, the antibodies provided herein include IgG isotype constant domains, which has several subclasses, including, but not limited to IgG1, IgG2, IgG3, and IgG4. In the IgG subclass of immunoglobulins, there are several immunoglobulin domains in the heavy chain. By “immunoglobulin (Ig) domain” herein is meant a region of an immunoglobulin having a distinct tertiary structure. Of interest in the antibodies described herein are the heavy chain domains, including, the constant heavy (CH) domains and the hinge domains. In the context of IgG antibodies, the IgG isotypes each have three CH regions. Accordingly, “CH” domains in the context of IgG are as follows: “CH1” refers to positions 118-220 according to the EU index as in Kabat. “CH2” refers to positions 237-340 according to the EU index as in Kabat, and “CH3” refers to positions 341-447 according to the EU index as in Kabat. As shown herein and described below, the pI variants can be in one or more of the CH regions, as well as the hinge region, discussed below.

It should be noted that IgG1 has different allotypes with polymorphisms at 356 (D or E) and 358 (L or M). The sequences depicted herein use the 356D/358M allotype, however the other allotype is included herein. That is, any sequence inclusive of an IgG1 Fc domain included herein can have 356E/358L replacing the 356D/358M allotype. It should be understood that therapeutic antibodies can also comprise hybrids of isotypes and/or subclasses. For example, as shown in US Publication 2009/0163699, incorporated by reference, the present antibodies, in some embodiments, include IgG1/IgG2 hybrids.

By “Fc” or “Fc region” or “Fc domain” as used herein is meant the polypeptide comprising the constant region of an antibody, in some instances, excluding all of the first constant region immunoglobulin domain (e.g., CH1) or a portion thereof, and in some cases, optionally including all or part of the hinge. For IgG, the Fc domain comprises immunoglobulin domains CH2 and CH3 (Cγ2 and Cγ3), and optionally all or a portion of the hinge region between CH1 (Cγ1) and CH2 (Cγ2). Thus, in some cases, the Fc domain includes, from N- to C-terminal, CH2-CH3 and hinge-CH2-CH3. In some embodiments, the Fc domain is that from human IgG1, IgG2, IgG3 or IgG4, with human IgG1 hinge-CH2-CH3 and IgG4 hinge-CH2-CH3 finding particular use in many embodiments. Additionally, in the case of human IgG1 Fc domains, frequently the hinge includes a C220S amino acid substitution. Furthermore, in the case of human IgG4 Fc domains, frequently the hinge includes a S228P amino acid substitution. Although the boundaries of the Fc region may vary, the human IgG heavy chain Fc region is usually defined to include residues E216, C226, or A231 to its carboxyl-terminal, wherein the numbering is according to the EU index as in Kabat. In some embodiments, as is more fully described below, amino acid modifications are made to the Fc region, for example to alter binding to one or more FcγR or to the FcRn.

By “heavy chain constant region” herein is meant the CH1-hinge-CH2-CH3 portion of an antibody (or fragments thereof), excluding the variable heavy domain; in EU numbering of human IgG1 this is amino acids 118-447 By “heavy chain constant region fragment” herein is meant a heavy chain constant region that contains fewer amino acids from either or both of the N- and C-termini but still retains the ability to form a dimer with another heavy chain constant region.

Another type of Ig domain of the heavy chain is the hinge region. By “hinge” or “hinge region” or “antibody hinge region” or “hinge domain” herein is meant the flexible polypeptide comprising the amino acids between the first and second constant domains of an antibody. Structurally, the IgG CH1 domain ends at EU position 215, and the IgG CH2 domain begins at residue EU position 231. Thus for IgG the antibody hinge is herein defined to include positions 216 (E216 in IgG1) to 230 (p230 in IgG1), wherein the numbering is according to the EU index as in Kabat. In some cases, a “hinge fragment” is used, which contains fewer amino acids at either or both of the N- and C-termini of the hinge domain. As noted herein, pI variants can be made in the hinge region as well. Many of the antibodies herein have at least one the cysteines at position 220 according to EU numbering (hinge region) replaced by a serine. Generally, this modification is on the “scFv monomer” side for most of the sequences depicted herein, although it can also be on the “Fab monomer” side, or both, to reduce disulfide formation. Specifically included within the sequences herein are one or both of these cysteines replaced (C220S).

As will be appreciated by those in the art, the exact numbering and placement of the heavy constant region domains can be different among different numbering systems. A useful comparison of heavy constant region numbering according to EU and Kabat is as below, see Edelman et al., 1969, Proc Natl Acad Sci USA 63:78-85 and Kabat et al., 1991, Sequences of Proteins of Immunological Interest, 5th Ed., United States Public Health Service, National Institutes of Health, Bethesda, entirely incorporated by reference.

TABLE 1

EU Numbering
Kabat Numbering

CH1
118-215
114-223

Hinge
216-230
226-243

CH2
231-340
244-360

CH3
341-447
361-478

The antibody light chain generally comprises two domains: the variable light domain (VL), which includes light chain CDRs vlCDR1-3, and a constant light chain region (often referred to as CL or Cκ). The antibody light chain is typically organized from N- to C-terminus: VL-CL.

By “antigen binding domain” or “ABD” herein is meant a set of six Complementary Determining Regions (CDRs) that, when present as part of a polypeptide sequence, specifically binds a target antigen (e.g., GPC3 or CD3) as discussed herein. As is known in the art, these CDRs are generally present as a first set of variable heavy CDRs (vhCDRs or VHCDRs) and a second set of variable light CDRs (vlCDRs or VLCDRs), each comprising three CDRs: vhCDR1, vhCDR2, vhCDR3 variable heavy CDRs and vlCDR1, vlCDR2 and vlCDR3 vhCDR3 variable light CDRs. The CDRs are present in the variable heavy domain (vhCDR1-3) and variable light domain (vlCDR1-3). The variable heavy domain and variable light domain form an Fv region.

The antibodies described herein provide a large number of different CDR sets. In this case, a “full CDR set” comprises the three variable light and three variable heavy CDRs, e.g., a vlCDR1, vlCDR2, vlCDR3, vhCDR1, vhCDR2 and vhCDR3. These can be part of a larger variable light or variable heavy domain, respectfully. In addition, as more fully outlined herein, the variable heavy and variable light domains can be on separate polypeptide chains, when a heavy and light chain is used (for example when Fabs are used), or on a single polypeptide chain in the case of scFv sequences.

As will be appreciated by those in the art, the exact numbering and placement of the CDRs can be different among different numbering systems. However, it should be understood that the disclosure of a variable heavy and/or variable light sequence includes the disclosure of the associated (inherent) CDRs. Accordingly, the disclosure of each variable heavy region is a disclosure of the vhCDRs (e.g., vhCDR1, vhCDR2 and vhCDR3) and the disclosure of each variable light region is a disclosure of the vlCDRs (e.g., vlCDR1, vlCDR2 and vlCDR3). A useful comparison of CDR numbering is as below, see Lafranc et al., Dev. Comp. Immunol. 27(1):55-77 (2003):

TABLE 2

Kabat +

Chothia
IMGT
Kabat
AbM
Chothia
Contact
Xencor

vhCDR1
26-35
27-38
31-35
26-35
26-32
30-35
27-35

vhCDR2
50-65
56-65
50-65
50-58
52-56
47-58
54-61

vhCDR3
95-102
105-117
95-102
95-102
95-102
93-101
103-116

vlCDR1
24-34
27-38
24-34
24-34
24-34
30-36
27-38

vlCDR2
50-56
56-65
50-56
50-56
50-56
46-55
56-62

vlCDR3
89-97
105-117
89-97
89-97
89-97
89-96
97-105

Throughout the present specification, the Kabat numbering system is generally used when referring to a residue in the variable domain (approximately, residues 1-107 of the light chain variable region and residues 1-113 of the heavy chain variable region) and the EU numbering system for Fc regions (e.g., Kabat et al., supra (1991)).

The CDRs contribute to the formation of the antigen-binding, or more specifically, epitope binding site of the antigen binding domains and antibodies. “Epitope” refers to a determinant that interacts with a specific antigen binding site in the variable region of an antibody molecule known as a paratope. Epitopes are groupings of molecules such as amino acids or sugar side chains and usually have specific structural characteristics, as well as specific charge characteristics. A single antigen may have more than one epitope.

The epitope may comprise amino acid residues directly involved in the binding (also called immunodominant component of the epitope) and other amino acid residues, which are not directly involved in the binding, such as amino acid residues which are effectively blocked by the specifically antigen binding peptide; in other words, the amino acid residue is within the footprint of the specifically antigen binding peptide.

Epitopes may be either conformational or linear. A conformational epitope is produced by spatially juxtaposed amino acids from different segments of the linear polypeptide chain. A linear epitope is one produced by adjacent amino acid residues in a polypeptide chain. Conformational and nonconformational epitopes may be distinguished in that the binding to the former but not the latter is lost in the presence of denaturing solvents.

An epitope typically includes at least 3, and more usually, at least 5 or 8-10 amino acids in a unique spatial conformation. Antibodies that recognize the same epitope can be verified in a simple immunoassay showing the ability of one antibody to block the binding of another antibody to a target antigen, for example “binning.” As outlined below, the disclosure not only includes the enumerated antigen binding domains and antibodies herein, but those that compete for binding with the epitopes bound by the enumerated antigen binding domains.

In some embodiments, the six CDRs of the antigen binding domain are contributed by a variable heavy and a variable light domain. In a “Fab” format, the set of 6 CDRs are contributed by two different polypeptide sequences, the variable heavy domain (vh or VH; containing the vhCDR1, vhCDR2 and vhCDR3) and the variable light domain (vl or VL; containing the vlCDR1, vlCDR2 and vlCDR3), with the C-terminus of the vh domain being attached to the N-terminus of the CH1 domain of the heavy chain and the C-terminus of the vl domain being attached to the N-terminus of the constant light domain (and thus forming the light chain). In a scFv format, the vh and vl domains are covalently attached, generally through the use of a linker (a “scFv linker”) as outlined herein, into a single polypeptide sequence, which can be either (starting from the N-terminus) vh-linker-vl or vl-linker-vh, with the former being generally preferred (including optional domain linkers on each side, depending on the format used (e.g., from FIG. 44). In general, the C-terminus of the scFv domain is attached to the N-terminus of the hinge in the second monomer.

By “variable region” or “variable domain” as used herein is meant the region of an immunoglobulin that comprises one or more Ig domains substantially encoded by any of the Vκ, Vλ, and/or VH genes that make up the kappa, lambda, and heavy chain immunoglobulin genetic loci respectively, and contains the CDRs that confer antigen specificity. Thus, a “variable heavy domain” pairs with a “variable light domain” to form an antigen binding domain (“ABD”). In addition, each variable domain comprises three hypervariable regions (“complementary determining regions,” “CDRs”) (VHCDR1, VHCDR2 and VHCDR3 for the variable heavy domain and VLCDR1, VLCDR2 and VLCDR3 for the variable light domain) and four framework (FR) regions, arranged from amino-terminus to carboxy-terminus in the following order: FR1-CDR1-FR2-CDR2-FR3-CDR3-FR4. The hypervariable region generally encompasses amino acid residues from about amino acid residues 24-34 (LCDR1; “L” denotes light chain), 50-56 (LCDR2) and 89-97 (LCDR3) in the light chain variable region and around about 31-35B (HCDR1; “H” denotes 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 Health, Bethesda, Md. (1991) and/or those residues forming a hypervariable loop (e.g. residues 26-32 (LCDR1), 50-52 (LCDR2) and 91-96 (LCDR3) in the light chain variable region and 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 invention are described in Table 2.

By “Fab” or “Fab region” as used herein is meant the polypeptide that comprises the VH, CH1, VL, and CL immunoglobulin domains, generally on two different polypeptide chains (e.g. VH-CH1 on one chain and VL-CL on the other). Fab may refer to this region in isolation, or this region in the context of a bispecific antibody described herein. In the context of a Fab, the Fab comprises an Fv region in addition to the CH1 and CL domains.

By “Fv” or “Fv fragment” or “Fv region” as used herein is meant a polypeptide that comprises the VL and VH domains of an ABD. Fv regions can be formatted as both Fabs (as discussed above, generally two different polypeptides that also include the constant regions as outlined above) and scFvs, where the VL and VH domains are combined (generally with a linker as discussed herein) to form an scFv.

By “single chain Fv” or “scFv” herein is meant a variable heavy domain covalently attached to a variable light domain, generally using a scFv linker as discussed herein, to form a scFv or scFv domain. A scFv domain can be in either orientation from N- to C-terminus (VH-linker-VL or VL-linker-VH). In the sequences depicted in the sequence listing and in the figures, the order of the VH and VL domain is indicated in the name, e.g. H.X_L.Y means N- to C-terminal is VH-linker-VL, and L.Y_H.X is VL-linker-VH. However, the disclosure of any “H_L” pairs is meant to include them in either order.

Some embodiments of the subject antibodies provided herein comprise at least one scFv domain, which, while not naturally occurring, generally includes a variable heavy domain and a variable light domain, linked together by a scFv linker. As outlined herein, while the scFv domain is generally from N- to C-terminus oriented as VH-scFv linker-VL, this can be reversed for any of the scFv domains (or those constructed using vh and vl sequences from Fabs), to VL-scFv linker-VH, with optional linkers at one or both ends depending on the format.

By “modification” herein is meant an amino acid substitution, insertion, and/or deletion in a polypeptide sequence or an alteration to a moiety chemically linked to a protein. For example, a modification may be an altered carbohydrate or PEG structure attached to a protein. By “amino acid modification” herein is meant an amino acid substitution, insertion, and/or deletion in a polypeptide sequence. For clarity, unless otherwise noted, the amino acid modification is always to an amino acid coded for by DNA, e.g. the 20 amino acids that have codons in DNA and RNA.

By “amino acid substitution” or “substitution” herein is meant the replacement of an amino acid at a particular position in a parent polypeptide sequence with a different amino acid. In particular, in some embodiments, the substitution is to an amino acid that is not naturally occurring at the particular position, either not naturally occurring within the organism or in any organism. For example, the substitution E272Y refers to a variant polypeptide, in this case an Fc variant, in which the glutamic acid at position 272 is replaced with tyrosine. For clarity, a protein which has been engineered to change the nucleic acid coding sequence but not change the starting amino acid (for example exchanging CGG (encoding arginine) to CGA (still encoding arginine) to increase host organism expression levels) is not an “amino acid substitution”; that is, despite the creation of a new gene encoding the same protein, if the protein has the same amino acid at the particular position that it started with, it is not an amino acid substitution.

By “amino acid insertion” or “insertion” as used herein is meant the addition of an amino acid sequence at a particular position in a parent polypeptide sequence. For example, −233E or 233E designates an insertion of glutamic acid after position 233 and before position 234. Additionally, −233ADE or A233ADE designates an insertion of AlaAspGlu after position 233 and before position 234.

By “amino acid deletion” or “deletion” as used herein is meant the removal of an amino acid sequence at a particular position in a parent polypeptide sequence. For example, E233− or E233 #, E233( ) or E233del designates a deletion of glutamic acid at position 233. Additionally, EDA233− or EDA233 #designates a deletion of the sequence GluAspAla that begins at position 233.

By “variant protein” or “protein variant”, or “variant” as used herein is meant a protein that differs from that of a parent protein by virtue of at least one amino acid modification. The protein variant has at least one amino acid modification compared to the parent protein, yet not so many that the variant protein will not align with the parental protein using an alignment program such as that described below. In general, variant proteins (such as variant Fc domains, etc., outlined herein, are generally at least 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98 or 99% identical to the parent protein, using the alignment programs described below, such as BLAST. “Variant” as used herein also refers to particular amino acid modifications that confer particular function (e.g., a “heterodimerization variant,” “pI variant,” “ablation variant,” etc.).

As described below, in some embodiments the parent polypeptide, for example an Fc parent polypeptide, is a human wild-type sequence, such as the heavy constant domain or Fc region from IgG1, IgG2, IgG3 or IgG4, although human sequences with variants can also serve as “parent polypeptides”, for example the IgG1/2 hybrid of US Publication 2006/0134105 can be included. The protein variant sequence herein will preferably possess at least about 80% identity with a parent protein sequence, and most preferably at least about 90% identity, more preferably at least about 95-98-99% identity. Accordingly, by “antibody variant” or “variant antibody” as used herein is meant an antibody that differs from a parent antibody by virtue of at least one amino acid modification, “IgG variant” or “variant IgG” as used herein is meant an antibody that differs from a parent IgG (again, in many cases, from a human IgG sequence) by virtue of at least one amino acid modification, and “immunoglobulin variant” or “variant immunoglobulin” as used herein is meant an immunoglobulin sequence that differs from that of a parent immunoglobulin sequence by virtue of at least one amino acid modification.

“Fc variant” or “variant Fc” as used herein is meant a protein comprising an amino acid modification in an Fc domain. The modification can be an addition, deletion, or substitution. The Fc variants are defined according to the amino acid modifications that compose them. Thus, for example, N434S or 434S is an Fc variant with the substitution for serine at position 434 relative to the parent Fc polypeptide, wherein the numbering is according to the EU index. Likewise, M428L/N434S defines an Fc variant with the substitutions M428L and N434S relative to the parent Fc polypeptide. The identity of the WT amino acid may be unspecified, in which case the aforementioned variant is referred to as 428/434S. It is noted that the order in which substitutions are provided is arbitrary, that is to say that, for example, 428L/434S is the same Fc variant as 434S/428L, and so on. For all positions discussed herein that relate to antibodies or derivatives and fragments thereof (e.g., Fc domains), unless otherwise noted, amino acid position numbering is according to the EU index. The “EU index” or “EU index as in Kabat” or “EU numbering” scheme refers to the numbering of the EU antibody (Edelman et al., 1969, Proc Natl Acad Sci USA 63:78-85, hereby entirely incorporated by reference).

In general, variant Fc domains have at least about 80, 85, 90, 95, 97, 98 or 99 percent identity to the corresponding parental human IgG Fc domain (using the identity algorithms discussed below, with one embodiment utilizing the BLAST algorithm as is known in the art, using default parameters). Alternatively, the variant Fc domains can have from 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 as compared to the parental Fc domain. Alternatively, the variant Fc domains can 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 as compared to the parental Fc domain. Additionally, as discussed herein, the variant Fc domains described herein still retain the ability to form a dimer with another Fc domain as measured using known techniques as described herein, such as non-denaturing gel electrophoresis.

By “protein” herein is meant at least two covalently attached amino acids, which includes proteins, polypeptides, oligopeptides and peptides. In addition, polypeptides that make up the antibodies described herein may include synthetic derivatization of one or more side chains or termini, glycosylation, PEGylation, circular permutation, cyclization, linkers to other molecules, fusion to proteins or protein domains, and addition of peptide tags or labels.

By “residue” as used herein is meant a position in a protein and its associated amino acid identity. For example, Asparagine 297 (also referred to as Asn297 or N297) is a residue at position 297 in the human antibody IgG1.

By “IgG subclass modification” or “isotype modification” as used herein is meant an amino acid modification that converts one amino acid of one IgG isotype to the corresponding amino acid in a different, aligned IgG isotype. For example, because IgG1 comprises a tyrosine and IgG2 a phenylalanine at EU position 296, a F296Y substitution in IgG2 is considered an IgG subclass modification.

By “non-naturally occurring modification” as used herein is meant an amino acid modification that is not isotypic. For example, because none of the human IgGs comprise a serine at position 434, the substitution 434S in IgG1, IgG2, IgG3, or IgG4 (or hybrids thereof) is considered a non-naturally occurring modification.

By “amino acid” and “amino acid identity” as used herein is meant one of the 20 naturally occurring amino acids that are coded for by DNA and RNA.

By “effector function” as used herein is meant a biochemical event that results from the interaction of an antibody Fc region with an Fc receptor or ligand. Effector functions include but are not limited to ADCC, ADCP, and CDC.

By “IgG Fc ligand” as used herein is meant a molecule, preferably a polypeptide, from any organism 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 FcγRIs, FcγRIIs, FcγRIIIs, FcRn, C1q, C3, mannan binding lectin, mannose receptor, staphylococcal protein A, streptococcal protein G, and viral FcγR. Fc ligands also include Fc receptor homologs (FcRH), which are a family of Fc receptors that are homologous to the FcγRs (Davis et al., 2002, Immunological Reviews 190:123-136, entirely incorporated by reference). Fc ligands may include undiscovered molecules that bind Fc. Particular IgG Fc ligands are FcRn and Fc gamma receptors. By “Fc ligand” as used herein is meant a molecule, preferably a polypeptide, from any organism that binds to the Fc region of an antibody to form an Fc/Fc ligand complex.

By “Fc gamma receptor”, “FcγR” or “FcgammaR” as used herein is meant any member of the family of proteins that bind the IgG antibody Fc region and is encoded by an FcγR gene. In humans this family includes but is not limited to FcγRI (CD64), including isoforms FcγRIa, FcγRIb, and FcγRIc; FcγRII (CD32), including isoforms FcγRIIa (including allotypes H131 and R131), FcγRIIb (including FcγRIIb-1 and FcγRIIb-2), and FcγRIIc; and FcγRIII (CD16), including isoforms FcγRIIIa (including allotypes V158 and F158) and FcγRIIIb (including allotypes FcγRIIb-NA1 and FcγRIIb-NA2) (Jefferis et al., 2002, Immunol Lett 82:57-65, entirely incorporated by reference), as well as any undiscovered human FcγRs or FcγR isoforms or allotypes. An FcγR may be from any organism, including but not limited to humans, mice, rats, rabbits, and monkeys. Mouse FcγRs include but are not limited to FcγRI (CD64), FcγRII (CD32), FcγRIII (CD16), and FcγRIII-2 (CD16-2), as well as any undiscovered mouse FcγRs or FcγR isoforms or allotypes.

By “FcRn” or “neonatal Fc Receptor” as used herein is meant a protein that binds the IgG antibody Fc region and is encoded at least in part by an FcRn gene. The FcRn may be from any organism, including but not limited to humans, mice, rats, rabbits, and monkeys. As is known in the art, the functional FcRn protein comprises two polypeptides, often referred to as the heavy chain and light chain. The light chain is beta-2-microglobulin and the heavy chain is encoded by the FcRn gene. Unless otherwise noted herein, FcRn or an FcRn protein refers to the complex of FcRn heavy chain with beta-2-microglobulin. A variety of FcRn variants used to increase binding to the FcRn receptor, and in some cases, to increase serum half-life. An “FcRn variant” is one that increases binding to the FcRn receptor, and suitable FcRn variants are shown below.

By “parent polypeptide” as used herein is meant a starting polypeptide that is subsequently modified to generate a variant. The parent polypeptide may be a naturally occurring polypeptide, or a variant or engineered version of a naturally occurring polypeptide. Accordingly, by “parent immunoglobulin” as used herein is meant an unmodified immunoglobulin polypeptide that is modified to generate a variant, and by “parent antibody” as used herein is meant an unmodified antibody that is modified to generate a variant antibody. It should be noted that “parent antibody” includes known commercial, recombinantly produced antibodies as outlined below. In this context, a “parent Fc domain” will be relative to the recited variant; thus, a “variant human IgG1 Fc domain” is compared to the parent Fc domain of human IgG1, a “variant human IgG4 Fc domain” is compared to the parent Fc domain human IgG4, etc.

By “position” as used herein is meant a location in the sequence of a protein. Positions may be numbered sequentially, or according to an established format, for example the EU index for antibody numbering.

By “target antigen” as used herein is meant the molecule that is bound specifically by the antigen binding domain comprising the variable regions of a given antibody.

By “strandedness” in the context of the monomers of the heterodimeric antibodies described herein is meant that, similar to the two strands of DNA that “match”, heterodimerization variants are incorporated into each monomer so as to preserve the ability to “match” to form heterodimers. For example, if some pI variants are engineered into monomer A (e.g. making the pI higher) then steric variants that are “charge pairs” that can be utilized as well do not interfere with the pI variants, e.g. the charge variants that make a pI higher are put on the same “strand” or “monomer” to preserve both functionalities. Similarly, for “skew” variants that come in pairs of a set as more fully outlined below, the skilled artisan will consider pI in deciding into which strand or monomer one set of the pair will go, such that pI separation is maximized using the pI of the skews as well.

By “target cell” as used herein is meant a cell that expresses a target antigen.

By “host cell” in the context of producing a bispecific antibody according to the antibodies described herein is meant a cell that contains the exogeneous nucleic acids encoding the components of the bispecific antibody and is capable of expressing the bispecific antibody under suitable conditions. Suitable host cells are discussed below.

By “wild type or WT” herein is meant an amino acid sequence or a nucleotide sequence that is found in nature, including allelic variations. A WT protein has an amino acid sequence or a nucleotide sequence that has not been intentionally modified.

Provided herein are a number of antibody domains that have sequence identity to human antibody domains. Sequence identity between two similar sequences (e.g., antibody variable domains) can be measured by algorithms such as that of Smith, T. F. & Waterman, M. S. (1981) “Comparison Of Biosequences,” Adv. Appl. Math. 2:482 [local homology algorithm]; Needleman, S. B. & 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, W. R. & Lipman, D. J. (1988) “Improved Tools For Biological Sequence Comparison,” Proc. Natl. Acad. Sci. (U.S.A.) 85:2444 [search for similarity method]; or Altschul, S. F. et al, (1990) “Basic Local Alignment Search Tool,” J. Mol. Biol. 215:403-10, the “BLAST” algorithm, see blast.ncbi.nlm.nih.gov. When using any of the aforementioned algorithms, the default parameters (for Window length, gap penalty, etc.) are used. In one embodiment, sequence identity is done using the BLAST algorithm, using default parameters

The antibodies described herein are generally isolated or recombinant. “Isolated,” when used to describe the various polypeptides disclosed herein, means a polypeptide that has been identified and separated and/or recovered from a cell or cell culture from which it was expressed. Ordinarily, an isolated polypeptide will be prepared by at least one purification step. An “isolated antibody,” refers to an antibody which is substantially free of other antibodies having different antigenic specificities. “Recombinant” means the antibodies are generated using recombinant nucleic acid techniques in exogeneous host cells, and they can be isolated as well.

“Specific binding” or “specifically binds to” or is “specific for” a particular antigen or an epitope means binding that is measurably different from a non-specific interaction. Specific binding can be measured, for example, by determining binding of a molecule compared to binding of a control molecule, which generally is a molecule of similar structure that does not have binding activity. For example, specific binding can be determined by competition with a control molecule that is similar to the target.

Specific binding for a particular antigen or an epitope can be exhibited, for example, by an antibody having a KD for an antigen or epitope of at least about 104 M, at least about 10⁻⁵M, at least about 10⁻⁶M, at least about 10⁻⁷M, at least about 10⁻⁸M, at least about 10⁻⁹M, alternatively at least about 10⁻¹⁰M, at least about 10⁻¹¹M, at least about 10⁻¹²M, or greater, where KD refers to a dissociation rate of a particular antibody-antigen interaction. Typically, an antibody that specifically binds an antigen will have a KD that is 20-, 50-, 100-, 500-, 1000-, 5,000-, 10,000- or more times greater for a control molecule relative to the antigen or epitope.

Also, specific binding for a particular antigen or an epitope can be exhibited, for example, by an antibody having a KA or Ka for an antigen or epitope of at least 20-, 50-, 100-, 500-, 1000-, 5,000-, 10,000- or more times greater for the epitope relative to a control, where KA or Ka refers to an association rate of a particular antibody-antigen interaction. Binding affinity is generally measured using a Biacore, SPR or BLI assay.

D. Antibodies of the Invention

The present invention provides antibodies, including monoclonal antibodies and bispecific antibodies, that bind to human GPC3 (it should be noted that many, if not most, of the exemplified antibodies also bind to cyno GPC3 for ease of pre-clinical testing, but this is not required in all embodiments). In particularly, bispecific antibodies are provided that bind CD3 and GPC3 that make take on a variety of formats as more fully described below.

1. Antibodies

The antibodies provided herein include different antibody domains as is more fully described below. As described herein and known in the art, the antibodies described herein include different domains within the heavy and light chains, which can be overlapping as well. These domains include, but are not limited to, the Fc domain, the CH1 domain, the CH2 domain, the CH3 domain, the hinge domain, the heavy constant domain (CH1-hinge-Fc domain or CH1-hinge-CH2-CH3), the variable heavy domain, the variable light domain, the light constant domain, Fab domains and scFv domains.

In particular, the formats depicted in FIGS. 15A and 15B are antibodies, usually referred to as “heterodimeric antibodies”, meaning that the protein has at least two associated Fc sequences self-assembled into a heterodimeric Fc domain and at least two Fv regions, whether as Fabs or as scFvs.

a. Chimeric and Humanized Antibodies

In certain embodiments, the antibodies described herein comprise a heavy chain variable region from a particular germline heavy chain immunoglobulin gene and/or a light chain variable region from a particular germline light chain immunoglobulin gene. For example, such antibodies may comprise or consist of a human antibody comprising heavy or light chain variable regions that are “the product of” or “derived from” a particular germline sequence. A human antibody that is “the product of” or “derived from” a human germline immunoglobulin sequence can be identified as such by comparing the amino acid sequence of the human antibody to the amino acid sequences of human germline immunoglobulins and selecting the human germline immunoglobulin sequence that is closest in sequence (i.e., greatest % identity) to the sequence of the human antibody (using the methods outlined herein). A human antibody that is “the product of” or “derived from” a particular human germline immunoglobulin sequence may contain amino acid differences as compared to the germline sequence, due to, for example, naturally-occurring somatic mutations or intentional introduction of site-directed mutation. However, a humanized antibody typically is at least 90% identical in amino acids sequence to an amino acid sequence encoded by a human germline immunoglobulin gene and contains amino acid residues that identify the antibody as being derived from human sequences when compared to the germline immunoglobulin amino acid sequences of other species (e.g., murine germline sequences). In certain cases, a humanized antibody may be at least 95, 96, 97, 98 or 99%, or even at least 96%, 97%, 98%, or 99% identical in amino acid sequence to the amino acid sequence encoded by the germline immunoglobulin gene. Typically, a humanized antibody derived from a particular human germline sequence will display no more than 10-20 amino acid differences from the amino acid sequence encoded by the human germline immunoglobulin gene (prior to the introduction of any skew, pI and ablation variants herein; that is, the number of variants is generally low, prior to the introduction of the variants described herein). In certain cases, the humanized antibody may display no more than 5, or even no more than 4, 3, 2, or 1 amino acid difference from the amino acid sequence encoded by the germline immunoglobulin gene (again, prior to the introduction of any skew, pI and ablation variants herein; that is, the number of variants is generally low, prior to the introduction of the variants described herein).

In one embodiment, the parent antibody has been affinity matured, as is known in the art. Structure-based methods may be employed for humanization and affinity maturation, for example as described in U.S. Ser. No. 11/004,590. Selection based methods may be employed to humanize and/or affinity mature antibody variable regions, including but not limited to 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; Krauss et al., 2003, Protein Engineering 16(10):753-759, all entirely incorporated by reference. Other humanization methods may involve the grafting of only parts of the CDRs, including but not limited to methods described in U.S. Ser. No. 09/810,510; Tan et al., 2002, J. Immunol. 169:1119-1125; De Pascalis et al., 2002, J. Immunol. 169:3076-3084, all entirely incorporated by reference.

2. Heterodimeric Antibodies

In exemplary embodiments, the bispecific antibodies provided herein are heterodimeric bispecific antibodies that include two variant Fc domain sequences. Such variant Fc domains include amino acid modifications to facilitate the self-assembly and/or purification of the heterodimeric antibodies.

An ongoing problem in antibody technologies is the desire for “bispecific” antibodies that bind to two different antigens simultaneously, in general thus allowing the different antigens to be brought into proximity and resulting in new functionalities and new therapies. In general, these antibodies are made by including genes for each heavy and light chain into the host cells. This generally results in the formation of the desired heterodimer (A-B), as well as the two homodimers (A-A and B-B (not including the light chain heterodimeric issues)). However, a major obstacle in the formation of bispecific antibodies is the difficulty in biasing the formation of the desired heterodimeric antibody over the formation of the homodimers and/or purifying the heterodimeric antibody away from the homodimers.

There are a number of mechanisms that can be used to generate the subject heterodimeric antibodies. In addition, as will be appreciated by those in the art, these different mechanisms can be combined to ensure high heterodimerization. Amino acid modifications that facilitate the production and purification of heterodimers are collectively referred to generally as “heterodimerization variants.” As discussed below, heterodimerization variants include “skew” variants (e.g., the “knobs and holes” and the “charge pairs” variants described below) as well as “pI variants,” which allow purification of heterodimers from homodimers. As is generally described in U.S. Pat. No. 9,605,084, hereby incorporated by reference in its entirety and specifically as below for the discussion of heterodimerization variants, useful mechanisms for heterodimerization include “knobs and holes” (“KIH”) as described in U.S. Pat. No. 9,605,084, “electrostatic steering” or “charge pairs” as described in U.S. Pat. No. 9,605,084, pI variants as described in U.S. Pat. No. 9,605,084, and general additional Fc variants as outlined in U.S. Pat. No. 9,605,084 and below.

Heterodimerization variants that are useful for the formation and purification of the subject heterodimeric antibody (e.g., bispecific antibodies) are further discussed in detailed below.

a. Skew Variants

In some embodiments, the heterodimeric antibody includes skew variants which are one or more amino acid modifications in a first Fc domain (A) and/or a second Fc domain (B) that favor the formation of Fc heterodimers (Fc dimers that include the first and the second Fc domain; (A-B) over Fc homodimers (Fc dimers that include two of the first Fc domain or two of the second Fc domain; A-A or B-B). Suitable skew variants are included in the FIG. 29 of US Publ. App. No. 2016/0355608, hereby incorporated by reference in its entirety and specifically for its disclosure of skew variants, as well as in FIG. 1.

Thus, suitable Fc heterodimerization variant pairs that will permit the formation of heterodimeric Fc regions are shown in FIG. 1. Thus a first Fc domain has first Fc heterodimerization variants and the second Fc domain has second Fc heterodimerization variants selected from the pairs in FIG. 1.

One mechanism is generally referred to in the art as “knobs and holes”, referring to amino acid engineering that creates steric influences to favor heterodimeric formation and disfavor homodimeric formation can also optionally be used; this is sometimes referred to as “knobs and holes”, as described in U.S. Ser. No. 61/596,846, Ridgway et al., Protein Engineering 9(7):617 (1996); Atwell et al., J. Mol. Biol. 1997 270:26; U.S. Pat. No. 8,216,805, all of which are hereby incorporated by reference in their entirety. The Figures identify a number of “monomer A-monomer B” pairs that rely on “knobs and holes”. In addition, as described in Merchant et al., Nature Biotech. 16:677 (1998), these “knobs and hole” mutations can be combined with disulfide bonds to skew formation to heterodimerization.

An additional mechanism that finds use in the generation of heterodimers is sometimes referred to as “electrostatic steering” as described in Gunasekaran et al., J. Biol. Chem. 285(25):19637 (2010), hereby incorporated by reference in its entirety. This is sometimes referred to herein as “charge pairs”. In this embodiment, electrostatics are used to skew the formation towards heterodimerization. As those in the art will appreciate, these may also have an effect on pI, and thus on purification, and thus could in some cases also be considered pI variants. However, as these were generated to force heterodimerization and were not used as purification tools, they are classified as “steric variants”. These include, but are not limited to, D221E/P228E/L368E paired with D221R/P228R/K409R (e.g. these are “monomer corresponding sets) and C220E/P228E/368E paired with C220R/E224R/P228R/K409R.

In some embodiments, the skew variants advantageously and simultaneously favor heterodimerization based on both the “knobs and holes” mechanism as well as the “electrostatic steering” mechanism. In some embodiments, the heterodimeric antibody includes one or more sets of such heterodimerization skew variants. These variants come in “pairs” of “sets”. That is, one set of the pair is incorporated into the first monomer and the other set of the pair is incorporated into the second monomer. It should be noted that these sets do not necessarily behave as “knobs in holes” variants, with a one-to-one correspondence between a residue on one monomer and a residue on the other. That is, these pairs of sets may instead form an interface between the two monomers that encourages heterodimer formation and discourages homodimer formation, allowing the percentage of heterodimers that spontaneously form under biological conditions to be over 90%, rather than the expected 50% (25% homodimer A/A:50% heterodimer A/B:25% homodimer B/B). Exemplary heterodimerization “skew” variants are depicted in FIG. 1. In exemplary embodiments, the heterodimeric antibody includes Fc heterodimerization variants as sets: S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L; K370S: S364K/E357Q; or a T366S/L368A/Y407V: T366W (optionally including a bridging disulfide, T366S/L368A/Y407V/Y349C: T366W/S354C) are all “skew” variant amino acid substitution sets of Fc heterodimerization variants. In an exemplary embodiment, the heterodimeric antibody includes a “S364K/E357Q: L368D/K370S” amino acid substitution set. In terms of nomenclature, the pair “S364K/E357Q: L368D/K370S” means that one of the monomers includes an Fc domain that includes the amino acid substitutions S364K and E357Q and the other monomer includes an Fc domain that includes the amino acid substitutions L368D and K370S; as above, the “strandedness” of these pairs depends on the starting pI.

In some embodiments, the skew variants provided herein can be optionally and independently incorporated with any other modifications, including, but not limited to, other skew variants (see, e.g., in FIG. 37 of US Publ. App. No. 2012/0149876, herein incorporated by reference, particularly for its disclosure of skew variants), pI variants, isotypic variants, FcRn variants, ablation variants, etc. into one or both of the first and second Fc domains of the heterodimeric antibody. Further, individual modifications can also independently and optionally be included or excluded from the subject the heterodimeric antibody.

Additional monomer A and monomer B variants that can be combined with other variants, optionally and independently in any amount, such as pI variants outlined herein or other steric variants that are shown in FIG. 37 of US 2012/0149876, the figure and legend and SEQ ID NOs of which are incorporated expressly by reference herein.

In some embodiments, the steric variants outlined herein can be optionally and independently incorporated with any pI variant (or other variants such as Fc variants, FcRn variants, etc.) into one or both monomers, and can be independently and optionally included or excluded from the proteins of the antibodies described herein.

A list of suitable skew variants is found in FIG. 1, which shows some pairs of particular utility in many embodiments. Of particular use in many embodiments are the pairs of sets including, but 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. In terms of nomenclature, the pair “S364K/E357Q: L368D/K370S” means that one of the monomers has the double variant set S364K/E357Q and the other has the double variant set L368D/K370S.

b. pI (Isoelectric Point) Variants for Heterodimers

In some embodiments, the heterodimeric antibody includes purification variants that advantageously allow for the separation of heterodimeric antibody (e.g., anti-GPC3×anti-CD3 bispecific antibody) from homodimeric proteins.

There are several basic mechanisms that can lead to ease of purifying heterodimeric antibodies. For example, modifications to one or both of the antibody heavy chain monomers A and B such that each monomer has a different pI allows for the isoelectric purification of heterodimeric A-B antibody from monomeric A-A and B-B proteins. Alternatively, some scaffold formats, such as the “1+1 Fab-scFv-Fc” format and the “2+1 Fab₂-scFv-Fc” format, also allows separation on the basis of size. As described above, it is also possible to “skew” the formation of heterodimers over homodimers using skew variants. Thus, a combination of heterodimerization skew variants and pI variants find particular use in the heterodimeric antibodies provided herein.

Additionally, as more fully outlined below, depending on the format of the heterodimeric antibody, pI variants either contained within the constant region and/or Fc domains of a monomer, and/or domain linkers can be used. In some embodiments, the heterodimeric antibody includes additional modifications for alternative functionalities that can also create pI changes, such as Fc, FcRn and KO variants.

In some embodiments, the subject heterodimeric antibodies provided herein include at least one monomer with one or more modifications that alter the pI of the monomer (i.e., a “pI variant”). In general, as will be appreciated by those in the art, there are two general categories of pI variants: those that increase the pI of the protein (basic changes) and those that decrease the pI of the protein (acidic changes). As described herein, all combinations of these variants can be done: one monomer may be wild type, or a variant that does not display a significantly different pI from wild-type, and the other can be either more basic or more acidic. Alternatively, each monomer is changed, one to more basic and one to more acidic.

Depending on the format of the heterodimer antibody, pI variants can be either contained within the constant and/or Fc domains of a monomer, or charged linkers, either domain linkers or scFv linkers, can be used. That is, antibody formats that utilize scFv(s) such as “1+1 Fab-scFv-Fc”, format can include charged scFv linkers (either positive or negative), that give a further pI boost for purification purposes. As will be appreciated by those in the art, some 1+1 Fab-scFv-Fc formats are useful with just charged scFv linkers and no additional pI adjustments, although the antibodies described herein do provide pI variants that are on one or both of the monomers, and/or charged domain linkers as well. In addition, additional amino acid engineering for alternative functionalities may also confer pI changes, such as Fc, FcRn and KO variants.

In subject heterodimeric antibodies that utilizes pI as a separation mechanism to allow the purification of heterodimeric proteins, amino acid variants are introduced into one or both of the monomer polypeptides. That is, the pI of one of the monomers (referred to herein for simplicity as “monomer A”) can be engineered away from monomer B, or both monomer A and B change be changed, with the pI of monomer A increasing and the pI of monomer B decreasing. As is outlined more fully below, the pI changes of either or both monomers can be done by removing or adding a charged residue (e.g., a neutral amino acid is replaced by 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 (aspartic acid to lysine) or changing a charged residue to a neutral residue (e.g., loss of a charge; lysine to serine). A number of these variants are shown in the FIGS. 1 and 2.

Thus, in some embodiments, the subject heterodimeric antibody includes amino acid modifications in the constant regions that alter the isoelectric point (pI) of at least one, if not both, of the monomers of a dimeric protein to form “pI antibodies”) by incorporating amino acid substitutions (“pI variants” or “pI substitutions”) into one or both of the monomers. As shown herein, the separation of the heterodimers from the two homodimers can be accomplished if the pIs of the two monomers differ by as little as 0.1 pH unit, with 0.2, 0.3, 0.4 and 0.5 or greater all finding use in the antibodies described herein.

As will be appreciated by those in the art, the number of pI variants to be included on each or both monomer(s) to get good separation will depend in part on the starting pI of the components, for example in the 1+1 Fab-scFv-Fc and 2+1 Fab₂-scFv-Fc formats, the starting pI of the scFv and Fab(s) of interest. That is, to determine which monomer to engineer or in which “direction” (e.g., more positive or more negative), the Fv sequences of the two target antigens are calculated and a decision is made from there. As is known in the art, different Fvs will have different starting pIs which are exploited in the antibodies described herein. In general, as outlined herein, the pIs are engineered to result in a total pI difference of each monomer of at least about 0.1 logs, with 0.2 to 0.5 being preferred as outlined herein.

In the case where pI variants are used to achieve heterodimerization, by using the constant region(s) of the heavy chain(s), a more modular approach to designing and purifying bispecific proteins, including antibodies, is provided. Thus, in some embodiments, heterodimerization variants (including skew and pI heterodimerization variants) are not included in the variable regions, such that each individual antibody must be engineered. In addition, in some embodiments, the possibility of immunogenicity resulting from the pI variants is significantly reduced by importing pI variants from different IgG isotypes such that pI is changed without introducing significant immunogenicity. Thus, an additional problem to be solved is the elucidation of low pI constant domains with high human sequence content, e.g., the minimization or avoidance of non-human residues at any particular position. Alternatively or in addition to isotypic substitutions, the possibility of immunogenicity resulting from the pI variants is significantly reduced by utilizing isosteric substitutions (e.g. Asn to Asp; and Gln to Glu).

As discussed below, a side benefit that can occur with this pI engineering is also the extension of serum half-life and increased FcRn binding. That is, as described in US Publ. App. No. US 2012/0028304 (incorporated by reference in its entirety), lowering the pI of antibody constant domains (including those found in antibodies and Fc fusions) can lead to longer serum retention in vivo. These pI variants for increased serum half-life also facilitate pI changes for purification.

In addition, it should be noted that the pI variants give an additional benefit for the analytics and quality control process of bispecific antibodies, as the ability to either eliminate, minimize and distinguish when homodimers are present is significant. Similarly, the ability to reliably test the reproducibility of the heterodimeric antibody production is important.

In general, embodiments of particular use rely on sets of variants that include skew variants, which encourage heterodimerization formation over homodimerization formation, coupled with pI variants, which increase the pI difference between the two monomers to facilitate purification of heterodimers away from homodimers.

Exemplary combinations of pI variants are shown in FIGS. 1 and 2, and FIG. 30 of US Publ. App. No. 2016/0355608, all of which are herein incorporated by reference in its entirety and specifically for the disclosure of pI variants. Preferred combinations of pI variants are shown in FIGS. 1 and 2. As outlined herein and shown in the figures, these changes are shown relative to IgG1, but all isotypes can be altered this way, as well as isotype hybrids. In the case where the heavy chain constant domain is from IgG2-4, R133E and R133Q can also be used.

Accordingly, in some embodiments, one monomer has a set of substitutions from FIG. 2 and the other monomer has a charged linker (either in the form of a charged scFv linker because that monomer comprises an scFv or a charged domain linker, as the format dictates, which can be selected from those depicted in FIGS. 5 and 6).

In some embodiments, modifications are made in the hinge of the Fc domain, including positions 216, 217, 218, 219, 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, and 230 based on EU numbering. Thus, pI mutations and particularly substitutions can be made in one or more of positions 216-230, with 1, 2, 3, 4 or 5 mutations finding use. Again, all possible combinations are contemplated, alone or with other pI variants in other domains.

Specific substitutions that find use in lowering the pI of hinge domains include, but are not limited to, a deletion at position 221, a non-native valine or threonine at position 222, a deletion at position 223, a non-native glutamic acid at position 224, a deletion at position 225, a deletion at position 235 and a deletion or a non-native alanine at position 236. In some cases, only pI substitutions are done in the hinge domain, and in others, these substitution(s) are added to other pI variants in other domains in any combination.

In some embodiments, mutations can be made in the CH2 region, including positions 233, 234, 235, 236, 274, 296, 300, 309, 320, 322, 326, 327, 334 and 339, based on EU numbering. It should be noted that changes in 233-236 can be made to increase effector function (along with 327A) in the IgG2 backbone. Again, all possible combinations of these 14 positions can be made; e.g., = may include a variant Fc domain with 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 CH2 pI substitutions.

Specific substitutions that find use in lowering the pI of CH2 domains include, but are not limited to, a non-native glutamine or glutamic acid at position 274, a non-native phenylalanine at position 296, a non-native phenylalanine at position 300, a non-native valine at position 309, a non-native glutamic acid at position 320, a non-native glutamic acid at position 322, a non-native glutamic acid at position 326, a non-native glycine at position 327, a non-native glutamic acid at position 334, a non-native threonine at position 339, and all possible combinations within CH2 and with other domains.

In this embodiment, the modifications can be independently and optionally selected from position 355, 359, 362, 384, 389,392, 397, 418, 419, 444 and 447 (EU numbering) of the CH3 region. Specific substitutions that find use in lowering the pI of CH3 domains include, but are not limited to, a non-native glutamine or glutamic acid at position 355, a non-native serine at position 384, a non-native asparagine or glutamic acid at position 392, a non-native methionine at position 397, a non-native glutamic acid at position 419, a non-native glutamic acid at position 359, a non-native glutamic acid at position 362, a non-native glutamic acid at position 389, a non-native glutamic acid at position 418, a non-native glutamic acid at position 444, and a deletion or non-native aspartic acid at position 447.

In general, as will be appreciated by those in the art, there are two general categories of pI variants: those that increase the pI of the protein (basic changes) and those that decrease the pI of the protein (acidic changes). As described herein, all combinations of these variants can be done: one monomer may be wild type, or a variant that does not display a significantly different pI from wild-type, and the other can be either more basic or more acidic. Alternatively, each monomer is changed, one to more basic and one to more acidic.

Preferred combinations of pI variants are shown in FIG. 2. As outlined herein and shown in the figures, these changes are shown relative to IgG1, but all isotypes can be altered this way, as well as isotype hybrids. In the case where the heavy chain constant domain is from IgG2-4, R133E and R133Q can also be used.

In one embodiment, for example in the FIGS. 15 and 44 formats, a preferred combination of pI variants has one monomer (the negative Fab side) comprising 208D/295E/384D/418E/421D variants (N208D/Q295E/N384D/Q418E/N421D when relative to human IgG1) and a second monomer (the positive scFv side) comprising a positively charged scFv linker, including (GKPGS)₄(SEQ ID NO: 15). However, as will be appreciated by those in the art, the first monomer includes a CH1 domain, including position 208. Accordingly, in constructs that do not include a CH1 domain (for example for antibodies that do not utilize a CH1 domain on one of the domains, for example in a dual scFv format or a “one-armed” format such as those depicted in FIG. 44C or D), a preferred negative pI variant Fc set includes 295E384D/418E/421D variants (Q295E/N384D/Q418E/N421D when relative to human IgG1).

Accordingly, in some embodiments, one monomer has a set of substitutions from FIG. 4 and the other monomer has a charged linker (either in the form of a charged scFv linker because that monomer comprises an scFv or a charged domain linker, as the format dictates, which can be selected from those depicted in FIGS. 5 and 6).

c. Isotypic Variants

In addition, many embodiments of the antibodies described herein rely on the “importation” of pI amino acids at particular positions from one IgG isotype into another, thus reducing or eliminating the possibility of unwanted immunogenicity being introduced into the variants. A number of these are shown in FIG. 21 of US Publ. 2014/0370013, hereby incorporated by reference. That is, IgG1 is a common isotype for therapeutic antibodies for a variety of reasons, including high effector function. However, the heavy constant region of IgG1 has a higher pI than that of IgG2 (8.10 versus 7.31). By introducing IgG2 residues at particular positions into the IgG1 backbone, the pI of the resulting monomer is lowered (or increased) and additionally exhibits longer serum half-life. For example, IgG1 has a glycine (pI 5.97) at position 137, and IgG2 has a glutamic acid (pI 3.22); importing the glutamic acid will affect the pI of the resulting protein. As is described below, a number of amino acid substitutions are generally required to significant affect the pI of the variant antibody. However, it should be noted as discussed below that even changes in IgG2 molecules allow for increased serum half-life.

In other embodiments, non-isotypic amino acid changes are made, either to reduce the overall charge state of the resulting protein (e.g. by changing a higher pI amino acid to a lower pI amino acid), or to allow accommodations in structure for stability, etc. as is further described below.

In addition, by pI engineering both the heavy and light constant domains, significant changes in each monomer of the heterodimer can be seen. As discussed herein, having the pIs of the two monomers differ by at least 0.5 can allow separation by ion exchange chromatography or isoelectric focusing, or other methods sensitive to isoelectric point.

d. Calculating pI

The pI of each monomer can depend on the pI of the variant heavy chain constant domain and the pI of the total monomer, including the variant heavy chain constant domain and the fusion partner. Thus, in some embodiments, the change in pI is calculated on the basis of the variant heavy chain constant domain, using the chart in the FIG. 19 of US Pub. 2014/0370013. As discussed herein, which monomer to engineer is generally decided by the inherent pI of the Fv and scaffold regions. Alternatively, the pI of each monomer can be compared.

e. pI Variants that Also Confer Better FcRn In Vivo Binding

In the case where the pI variant decreases the pI of the monomer, they can have the added benefit of improving serum retention in vivo.

Although still under examination, Fc regions are believed to have longer half-lives in vivo, because binding to FcRn at pH 6 in an endosome sequesters the Fc (Ghetie and Ward, 1997 Immunol Today. 18(12): 592-598, entirely incorporated by reference). The endosomal compartment then recycles the Fc to the cell surface. Once the compartment opens to the extracellular space, the higher pH, ˜7.4, induces the release of Fc back into the blood. 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, entirely incorporated by reference). The increased affinity of Fc for FcRn at pH 7.4 is thought to forbid the release of the Fc back into the blood. Therefore, the Fc mutations that will increase Fc's half-life in vivo will ideally increase FcRn binding at the lower pH while still allowing release of Fc at higher pH. The amino acid histidine changes its charge state in the pH range of 6.0 to 7.4. Therefore, it is not surprising to find His residues at important positions in the Fc/FcRn complex.

Recently it has been suggested that antibodies with variable regions that have lower isoelectric points may also have longer serum half-lives (Igawa et al., 2010 PEDS. 23(5): 385-392, entirely incorporated by reference). However, the mechanism of this is still poorly understood. Moreover, variable regions differ from antibody to antibody. Constant region variants with reduced pI and extended half-life would provide a more modular approach to improving the pharmacokinetic properties of antibodies, as described herein.

f. Additional Fc Variants for Additional Functionality

In addition to the heterodimerization variants discussed above, there are a number of useful Fc amino acid modification that can be made for a variety of reasons, including, but not limited to, altering binding to one or more FcγR receptors, altered binding to FcRn receptors, etc., as discussed below.

Accordingly, the antibodies provided herein (heterodimeric, as well as homodimeric) can include such amino acid modifications with or without the heterodimerization variants outlined herein (e.g., the pI variants and steric variants). Each set of variants can be independently and optionally included or excluded from any particular heterodimeric protein.

(i) FcγR Variants

Accordingly, there are a number of useful Fc substitutions that can be made to alter binding to one or more of the FcγR receptors. In certain embodiments, the subject antibody includes modifications that alter the binding to one or more FcγR receptors (i.e., “FcγR variants”). Substitutions that result in increased binding as well as decreased binding can be useful. For example, it is known that increased binding to FcγRIIIa generally results in increased ADCC (antibody dependent cell-mediated cytotoxicity; the cell-mediated reaction wherein nonspecific cytotoxic cells that express FcγRs recognize bound antibody on a target cell and subsequently cause lysis of the target cell). Similarly, decreased binding to FcγRIIb (an inhibitory receptor) can be beneficial as well in some circumstances. Amino acid substitutions that find use in the antibodies described herein include those listed in U.S. Pat. No. 8,188,321 (particularly FIG. 41) and U.S. Pat. No. 8,084,582, and US Publ. App. Nos. 20060235208 and 20070148170, all of which are expressly incorporated herein by reference in their entirety and specifically for the variants disclosed therein. Particular variants that find use 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.

In addition, there are additional Fc substitutions that find use in increased binding to the FcRn receptor and increased serum half-life, as specifically disclosed in U.S. Ser. No. 12/341,769, hereby incorporated by reference in its entirety, including, but not limited to, 434S, 434A, 428L, 308F, 259I, 428L/434S, 428I/434A, 259I/308F, 4361/428L, 4361 or V/434S, 436V/428L and 259I/308F/428L. Such modification may be included in one or both Fc domains of the subject antibody.

(ii) Ablation Variants

Similarly, another category of functional variants are “FcγR ablation variants” or “Fc knock out (FcKO or KO)” variants. In these embodiments, for some therapeutic applications, it is desirable to reduce or remove the normal binding of the Fc domain to one or more or all of the Fcγ receptors (e.g. FcγR1, FcγRIIa, FcγRIIb, FcγRIIIa, etc.) to avoid additional mechanisms of action. That is, for example, in many embodiments, particularly in the use of bispecific antibodies that bind CD3 monovalently it is generally desirable to ablate FcγRIIIa binding to eliminate or significantly reduce ADCC activity. wherein one of the Fc domains comprises one or more Fcγ receptor ablation variants. These ablation variants are depicted in FIG. 14, and each can be independently and optionally included or excluded, with preferred aspects utilizing ablation 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 ablation variants referenced herein ablate FcγR binding but generally not FcRn binding.

As is known in the art, the Fc domain of human IgG1 has the highest binding to the Fcγ receptors, and thus ablation variants can be used when the constant domain (or Fc domain) in the backbone of the heterodimeric antibody is IgG1. Alternatively, or in addition to ablation variants in an IgG1 background, mutations at the glycosylation position 297 (generally to A or S) can significantly ablate binding to FcγRIIIa, for example. Human IgG2 and IgG4 have naturally reduced binding to the Fcγ receptors, and thus those backbones can be used with or without the ablation variants.

E. Combination of Heterodimeric and Fc Variants

As will be appreciated by those in the art, all of the recited heterodimerization variants (including skew and/or pI variants) can be optionally and independently combined in any way, as long as they retain their “strandedness” or “monomer partition”. In some embodiments, the heterodimeric antibodies provided herein include the combination of heterodimerization skew variants, isosteric pI substitutions and FcKO variants as depicted in FIG. 4. In addition, all of these variants can be combined into any of the heterodimerization formats.

In the case of pI variants, while embodiments finding particular use are shown in the Figures, other combinations can be generated, following the basic rule of altering the pI difference between two monomers to facilitate purification.

In addition, any of the heterodimerization variants, skew and pI, are also independently and optionally combined with Fc ablation variants, Fc variants, FcRn variants, as generally outlined herein.

Exemplary combination of variants that are included in some embodiments of the heterodimeric 1+1 Fab-scFv-Fc and 2+1 Fab₂-scFv-Fc format antibodies are included in FIG. 4. In certain embodiments, the antibody is a heterodimeric 1+1 Fab-scFv-Fc or 2+1 Fab₂-scFv-Fc format antibody as shown in FIGS. 15A and 15B.

F. Anti-GPC3×Anti-CD3 Bispecific Antibodies

In another aspect, provided herein are anti-GPC3×anti-CD3 (also referred to herein as “αGPC3×αCD3” or sometimes just “GPC3×CD3”) bispecific antibodies. Such antibodies include at least one GPC3 binding domain and at least one CD3 binding domain. In some embodiments, bispecific αGPC3×αCD3 provided herein immune responses selectively in tumor sites that express GPC3.

Note that unless specified herein, the order of the antigen list in the name does not confer structure; that is a GPC3×CD3 1+1 Fab-scFv-Fc antibody can have the scFv bind to GPC3 or CD3, although in some cases, the order specifies structure as indicated.

As is more fully outlined herein, these combinations of ABDs can be in a variety of formats, as outlined below, generally in combinations where one ABD is in a Fab format and the other is in an scFv format. Exemplary formats that are used in the bispecific antibodies provided herein include the 1+1 Fab-scFv-Fc and 2+1 Fab2-scFv-Fv formats (see, e.g., FIGS. 15A and 15B). Other useful antibody formats include, but are not limited to, “mAb-Fv,” “mAb-scFv,” “central-Fv”, “one-armed scFv-mAb,” “scFv-mAb,” “dual scFv,” and “trident” format antibodies, as depicted in FIG. 44 and more fully described below.

In addition, in general, one of the ABDs comprises a scFv as outlined herein, in an orientation from N- to C-terminus of VH-scFv linker-VL or VL-scFv linker-VH. One or both of the other ABDs, according to the format, generally is a Fab, comprising a VH domain on one protein chain (generally as a component of a heavy chain) and a VL on another protein chain (generally as a component of a light chain).

As will be appreciated by those in the art, any set of 6 CDRs or VH and VL domains can be in the scFv format or in the Fab format, which is then added to the heavy and light constant domains, where the heavy constant domains comprise variants (including within the CH1 domain as well as the Fc domain). The scFv sequences contained in the sequence listing utilize a particular charged linker, but as outlined herein, uncharged or other charged linkers can be used, including those depicted in FIG. 5 and FIG. 6.

In addition, as discussed above, the numbering used in the Sequence Listing for the identification of the CDRs is Kabat, however, different numbering can be used, which will change the amino acid sequences of the CDRs as shown in Table 2.

For all of the variable heavy and light domains listed herein, further variants can be made. As outlined herein, in some embodiments the set of 6 CDRs can have from 0, 1, 2, 3, 4 or 5 amino acid modifications (with amino acid substitutions finding particular use), as well as changes in the framework regions of the variable heavy and light domains, as long as the frameworks (excluding the CDRs) retain at least about 80, 85 or 90% identity to a human germline sequence selected from those listed in FIG. 1 of U.S. Pat. No. 7,657,380, which Figure and Legend is incorporated by reference in its entirety herein. Thus, for example, the identical CDRs as described herein can be combined with different framework sequences from human germline sequences, as long as the framework regions retain at least 80, 85 or 90% identity to a human germline sequence selected from those listed in FIG. 1 of U.S. Pat. No. 7,657,380. Alternatively, the CDRs can have amino acid modifications (e.g., from 1, 2, 3, 4 or 5 amino acid modifications in the set of CDRs (that is, the CDRs can be modified as long as the total number of changes in the set of 6 CDRs is less than 6 amino acid modifications, with any combination of CDRs being changed; e.g., there may be one change in vlCDR1, two in vhCDR2, none in vhCDR3, etc.)), as well as having framework region changes, as long as the framework regions retain at least 80, 85 or 90% identity to a human germline sequence selected from those listed in FIG. 1 of U.S. Pat. No. 7,657,380.

As discussed herein, the subject heterodimeric antibodies include two antigen binding domains (ABDs), each of which bind to GPC3 or CD3. As outlined herein, these heterodimeric antibodies can be bispecific and bivalent (each antigen is bound by a single ABD, for example, in the format depicted in FIG. 15A), or bispecific and trivalent (one antigen is bound by a single ABD and the other is bound by two ABDs, for example as depicted in FIG. 15B).

The disclosure provides a number of ABDs as outlined below. As will be appreciated by those in the art, any set of 6 CDRs or VH and VL domains can be in the scFv format or in the Fab format, which is then added to the heavy and light constant domains, where the heavy constant domains comprise variants (including within the CH1 domain as well as the Fc domain). The scFv sequences contained in the sequence listing utilize a particular charged linker, but as outlined herein, uncharged or other charged linkers can be used, including those depicted in FIG. 5.

For all of the variable heavy and light domains listed herein, further variants can be made. As outlined herein, in some embodiments the set of 6 CDRs can have from 0, 1, 2, 3, 4 or 5 amino acid modifications (with amino acid substitutions finding particular use), as well as changes in the framework regions of the variable heavy and light domains, as long as the frameworks (excluding the CDRs) retain at least about 80, 85 or 90% identity to a human germline sequence selected from those listed in FIG. 1 of U.S. Pat. No. 7,657,380, which Figure and Legend is incorporated by reference in its entirety herein. Thus, for example, the identical CDRs as described herein can be combined with different framework sequences from human germline sequences, as long as the framework regions retain at least 80, 85 or 90% identity to a human germline sequence selected from those listed in FIG. 1 of U.S. Pat. No. 7,657,380. Alternatively, the CDRs can have amino acid modifications (e.g. from 1, 2, 3, 4 or 5 amino acid modifications in the set of CDRs (that is, the CDRs can be modified as long as the total number of changes in the set of 6 CDRs is less than 6 amino acid modifications, with any combination of CDRs being changed; e.g. there may be one change in VLCDR1, two in VHCDR2, none in VHCDR3, etc.)), as well as having framework region changes, as long as the framework regions retain at least 80, 85 or 90% identity to a human germline sequence selected from those listed in FIG. 1 of U.S. Pat. No. 7,657,380.

1. GPC3 Antigen Binding Domains

Herein is provided monoclonal and bispecific antibodies that contain antigen binding domains that bind to human GPC3. Suitable sets of 6 CDRs and/or VH and VL domains are depicted in FIGS. 12 and 13. In some embodiments, the heterodimeric antibody is a 1+1 Fab-scFv-Fc or 2+1 Fab2-scFv-Fv format antibody (see, e.g., FIGS. 15A and 15B) although any of the formats outlined below can be utilized.

In some embodiments, the anti-GPC3 ABD has a set of vhCDRs selected from the vhCDR1, vhCDR2 and vhCDR3 sequences from a VH selected from the group consisting of H1, H1, H1.1, H1.2, H1.3 and H1.4 the sequences of which are shown in FIGS. 12 and 13.

In some embodiments, the VH domain of the anti-GPC3 ABD is selected from the group consisting of H1, H1, H1.1, H1.2, H1.3 and H1.4 the sequences of which are shown in FIGS. 12 and 13.

In some embodiments, the anti-GPC3 ABD has a set of vlCDRs selected from the vlCDR1, vlCDR2 and vlCDR3 sequences from a VL selected from the group consisting of L1, L1.1, L1.2, L1.3, L1.4, L1.5, L1.6, L1.7, L1.8, L1.9, L1.10, L1.16, L1.23, L1.29, L1.31. L1.65, L1.66, L1.68, L1.69, L1.70, L1.71, L1.72 and L1.73, the sequences of which are shown in FIGS. 12 and 13.

In some embodiments, the VL of the GPC3 ABD is selected from the group consisting of L1, L1.1, L1.2, L1.3, L1.4, L1.5, L1.6, L1.7, L1.8, L1.9, L1.10, L1.16, L1.23, L1.29, L1.31. L1.65, L1.66, L1.68, L1.69, L1.70, L1.71, L1.72 and L1.73, the sequences of which are shown in FIGS. 12 and 13.

Accordingly, included herein are GPC3 ABDs that have a set of 6 CDRs (vhCDR1, vhCDR2, vhCDR3, vlCDR1, vlCDR2 and vlCDR3) from VH/VL pairs selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

Additionally, included herein are GPC3 ABDs that have VH/VL pairs selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73.

In particular embodiments, the VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69.

In particular embodiments, the VH/VL pairs are Fabs and are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69.

As will be appreciated by those in the art, suitable GPC3 binding domains can comprise a set of 6 CDRs as depicted in the Figures, either as they are underlined or, in the case where a different numbering scheme is used as described herein and as shown in Table 2, as the CDRs that are identified using other alignments within the VH and VL sequences of those depicted in FIGS. 12 and 13. Suitable ABDs can also include the entire VH and VL sequences as depicted in these sequences and Figures, used as scFvs or as Fabs. In many of the embodiments herein that contain an Fv to GPC3, it is the Fab monomer that binds GPC3.

In addition to the parental CDR sets disclosed in the figures and sequence listing that form an ABD to GPC3, provided herein are variant GPC3 ABDs having CDRs that include at least one modification of the GPC3 ABD CDRs disclosed herein (e.g., (FIGS. 12 and 13 and the sequence listing). In one embodiment, the GPC3 ABD of the subject heterodimeric antibody includes a set of 6 CDRs with 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 amino acid modifications as compared to the 6 CDRs of a GPC3 binding domain VH/VL pair as described herein, including the figures and sequence listing. In exemplary embodiments, the GPC3 ABD of the subject heterodimeric antibody includes a set of 6 CDRs with 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 amino acid modifications as compared to the 6 CDRs of one of the following GPC3 binding domain [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69. In certain embodiments, the GPC3 ABD of the subject antibody is capable of binding to GPC3, as measured at least one of a Biacore, surface plasmon resonance (SPR), BLI (biolayer interferometry, e.g., Octet assay) assay, and/or flow cytometry, with the latter finding particular use in many embodiments. In particular embodiments, the GPC3 ABD is capable of binding human GPC3 (see FIG. 11). In some cases, each variant CDR has no more than 1 or 2 amino acid changes, with no more than 1 per CDR being particularly useful.

In some embodiments, the GPC3 ABD of the subject antibody includes 6 CDRs that are at least 90, 95, 97, 98 or 99% identical to the 6 CDRs of a GPC3 ABD as described herein, including the figures and sequence listing. In exemplary embodiments, the GPC3 ABD of the subject antibody includes 6 CDRs that are at least 90, 95, 97, 98 or 99% identical to the 6 CDRs of one of the following GPC3 binding domain VH/VL pairs: GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69. In certain embodiments, the GPC3 ABD of the subject antibody is capable of binding to GPC3, as measured at least one of a Biacore, surface plasmon resonance (SPR), BLI (biolayer interferometry, e.g., Octet assay) assay, and/or flow cytometry. In particular embodiments, the GPC3 ABD is capable of binding human GPC3 antigen (see FIG. 11).

In another exemplary embodiment, the GPC3 ABD of the subject antibody includes the variable heavy (VH) domain and variable light (VL) domain of any one of the GPC3 binding domain VH/VL pairs described herein, including the figures and sequence listing (e.g., FIGS. 12 and 13).

In some embodiments, the subject antibody includes a GPC3 ABD that includes a variable heavy domain and/or a variable light domain that are variants of a GPC3 ABD VH and VL domain disclosed herein. In one embodiment, the variant VH domain and/or VL domain has from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid changes from a VH and/or VL domain of a GPC3 ABD described herein, including the figures and sequence listing. In exemplary embodiments, the variant VH domain and/or VL domain has from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid changes from a VH and/or VL domain of one of the following GPC3 binding domain VH/VL pairs: GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69. In some embodiments, the changes are in a VH domain depicted in FIGS. 12 and 13. In some embodiments, the changes are in a VL domain are depicted in FIGS. 12 and 13. In some embodiments, the changes are in a VH and VL domain are depicted in FIGS. 12 and 13. In some embodiments, one or more amino acid changes are in the VH and/or VL framework regions (FR1, FR2, FR3, and/or FR4). In some embodiments, one or more amino acid changes are in one or more CDRs. In certain embodiments, the GPC3 ABD of the subject antibody is capable of binding to GPC3, as measured at least one of a Biacore, surface plasmon resonance (SPR), BLI (biolayer interferometry, e.g., Octet assay) assay, and/or flow cytometry. In particular embodiments, the GPC3 ABD is capable of binding human GPC3 antigen (see FIG. 11).

In one embodiment, the variant VH and/or VL domain is at least 90, 95, 97, 98 or 99% identical to the VH and/or VL of a GPC3 ABD as described herein, including the figures and sequence listing. In exemplary embodiments, the variant VH and/or VL domain is at least 90, 95, 97, 98 or 99% identical to the VH and/or VL of one of the following GPC3 binding domain VH/VL pairs: GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69. In some embodiments, the GPC3 ABD includes a VH that is at least 90, 95, 97, 98 or 99% identical to VH domain depicted in FIGS. 12 and 13. In some embodiments, the GPC3 ABD includes a VL that is at least 90, 95, 97, 98 or 99% identical to VL domain depicted in FIGS. 12 and 13. In some embodiments, the GPC3 ABD includes a VH and a VL that is at least 90, 95, 97, 98 or 99% identical to a VH domain and a VL domain depicted in FIGS. 12 and 13. In certain embodiments, the GPC3 ABD of the subject antibody is capable of binding to GPC3, as measured at least one of a Biacore, surface plasmon resonance (SPR), BLI (biolayer interferometry, e.g., Octet assay) assay, and/or flow cytometry. In particular embodiments, the GPC3 ABD is capable of binding human GPC3 antigen (see FIG. 11).

2. CD3 Antigen Binding Domains

The heterodimeric bispecific of the invention (e.g., anti-GPC3×anti-CD3 antibodies) also include an ABD that binds to human epsilon CD3 (CD3E).

Suitable sets of 6 CDRs and/or VH and VL domains, as well as scFv sequences, are depicted in FIG. 10. CD3 binding domain sequences that are of particular use 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31 as depicted in FIG. 10. As shown in FIG. 10, when the anti-CD3 ABD is a scFv domain, the VH and VL domains can be in either orientation.

As will be appreciated by those in the art, suitable CD3 binding domains can comprise a set of 6 CDRs as depicted in FIG. 10, either as they are underlined or, in the case where a different numbering scheme is used as described herein and as shown in Table 2, as the CDRs that are identified using other alignments within the VH and VL sequences of those depicted in FIGS. 10A-10F. Suitable ABDs can also include the entire VH and VL sequences as depicted in these sequences and Figures, used as scFvs or as Fabs. In many of the embodiments herein that contain an Fv to CD3, it is the scFv monomer that binds CD3.

In addition to the parental CDR sets disclosed in the figures and sequence listing that form an ABD to CD3, provided herein are variant CD3 ABDS having CDRs that include at least one modification of the CD3 ABD CDRs disclosed herein (e.g., (FIG. 10 and the sequence listing). In one embodiment, the CD3 ABD of the subject heterodimeric antibody (e.g., anti-GPC3×anti-CD3 antibody) includes a set of 6 CDRs with 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 amino acid modifications as compared to the 6 CDRs of a CD3 ABD as described herein, including the figures and sequence listing. In exemplary embodiments, the CD3 ABD of the subject heterodimeric antibody includes a set of 6 CDRs with 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 amino acid modifications as compared to the 6 CDRs of one of the following 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31 (FIG. 10). In certain embodiments, the CD3 ABD of the subject antibody is capable of binding CD3 antigen, as measured by at least one of a Biacore, surface plasmon resonance (SPR), flow cytometry, and/or BLI (biolayer interferometry, e.g., Octet assay) assay, with the latter finding particular use in many embodiments. In particular embodiments, the CD3ABD is capable of binding human CD3.

In some embodiments, the CD3 ABD of the subject antibody includes 6 CDRs that are at least 90, 95, 97, 98 or 99% identical to the 6 CDRs of a CD3 ABD as described herein, including the figures and sequence listing. In exemplary embodiments, the CD3 ABD of the subject antibody includes 6 CDRs that are at least 90, 95, 97, 98 or 99% identical to the 6 CDRs of one of the following 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31 (FIG. 10). In certain embodiments, the CD3 ABD is capable of binding to the CD3, as measured by at least one of a Biacore, surface plasmon resonance (SPR), flow cytometry, and/or BLI (biolayer interferometry, e.g., Octet assay) assay, with the latter finding particular use in many embodiments. In particular embodiments, the CD3 ABD is capable of binding human CD3 antigen.

In another exemplary embodiment, the CD3 ABD of the subject antibody includes the variable heavy (VH) domain and variable light (VL) domain of any one of the CD3 binding domains described herein, including the figures and sequence listing.

In some embodiments, the subject antibody includes a CD3 ABD that includes a variable heavy domain and/or a variable light domain that are variants of a CD3 ABD VH and VL domain disclosed herein. In one embodiment, the variant VH domain and/or VL domain has from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid changes from a VH and/or VL domain of a CD3 ABD described herein, including the figures and sequence listing. In exemplary embodiments, the variant VH domain and/or VL domain has from 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid changes from a VH and/or VL domain of one of the following 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31 (FIG. 10). In some embodiments, the changes are in a VH domain depicted in FIG. 10. In some embodiments, the changes are in a VL domain are depicted in FIG. 10. In some embodiments, the changes are in a VH and VL domain are depicted in FIG. 10. In some embodiments, one or more amino acid changes are in the VH and/or VL framework regions (FR1, FR2, FR3, and/or FR4). In some embodiments, one or more amino acid changes are in one or more CDRs. In certain embodiments, the CD3 ABD of the subject antibody is capable of binding to CD3, as measured at least one of a Biacore, surface plasmon resonance (SPR), flow cytometry, and/or BLI (biolayer interferometry, e.g., Octet assay) assay, with the latter finding particular use in many embodiments. In particular embodiments, the CD3 ABD is capable of binding human CD3 antigen.

In one embodiment, the variant VH and/or VL domain is at least 90, 95, 97, 98 or 99% identical to the VH and/or VL of a CD3 ABD as described herein, including the figures and sequence listing. In exemplary embodiments, the variant VH and/or VL domain is at least 90, 95, 97, 98 or 99% identical to the VH and/or VL of one of the following 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31 (FIG. 10). In some embodiments, the CD3 ABD includes a VH that is at least 90, 95, 97, 98 or 99% identical to VH domain depicted in FIG. 10. In some embodiments, the CD3 ABD includes a VL that is at least 90, 95, 97, 98 or 99% identical to VL domain depicted in FIG. 10. In some embodiments, the CD3 ABD includes a VH and a VL that is at least 90, 95, 97, 98 or 99% identical to a VH domain and a VL domain depicted in FIG. 10. In certain embodiments, the CD3 ABD is capable of binding to CD3, as measured by at least one of a Biacore, surface plasmon resonance (SPR), flow cytometry, and/or BLI (biolayer interferometry, e.g., Octet assay) assay, with the latter finding particular use in many embodiments. In particular embodiments, the CD3 ABD is capable of binding human CD3 antigen.

In addition to the αCD3 ABDs of FIG. 10, additional ABDs of use in the invention include those depicted in FIGS. 14 and 15 of WO2014/145806, hereby expressly incorporated herein in their entirety including the Figures and Legends therein.

3. Linkers

As shown herein, there are a number of suitable linkers (for use as either domain linkers or scFv linkers) that can be used to covalently attach the recited domains (e.g., scFvs, Fabs, Fc domains, etc.), including traditional peptide bonds, generated by recombinant techniques. Exemplary linkers to attach domains of the subject antibody to each other are depicted in FIG. 6. In some embodiments, the linker peptide may predominantly include the following amino acid residues: Gly, Ser, Ala, or Thr. The linker peptide should have a length that is adequate to link two molecules in such a way that they assume the correct conformation relative to one another so that they retain the desired activity. In one embodiment, the linker is from about 1 to 50 amino acids in length, preferably about 1 to 30 amino acids in length. In one embodiment, linkers of 1 to 20 amino acids in length may be used, with from about 5 to about 10 amino acids finding use in some embodiments. Useful linkers include glycine-serine polymers, including for example (GS)n, (GSGGS)n (SEQ ID NO: 3), (GGGGS)n (SEQ ID NO: 2), and (GGGS)n (SEQ ID NO: 4), where n is an integer of at least one (and generally from 3 to 4), glycine-alanine polymers, alanine-serine polymers, and other flexible linkers, some of which are shown in FIG. 5 and FIG. 6. Alternatively, a variety of nonproteinaceous polymers, including but not limited to polyethylene glycol (PEG), polypropylene glycol, polyoxyalkylenes, or copolymers of polyethylene glycol and polypropylene glycol, may find use as linkers.

Other linker sequences may include any sequence of any length of CL/CH1 domain but not all residues of CL/CH1 domain; for example the first 5-12 amino acid residues of the CL/CH1 domains. Linkers can be derived from immunoglobulin light chain, for example Cκ or Cλ. Linkers can be derived from immunoglobulin heavy chains of any isotype, including for example Cγ1, 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 natural sequences from other proteins.

In some embodiments, the linker is a “domain linker”, used to link any two domains as outlined herein together. For example, in FIG. 15B, there may be a domain linker that attaches the C-terminus of the CH1 domain of the Fab to the N-terminus of the scFv, with another optional domain linker attaching the C-terminus of the scFv to the CH2 domain (although in many embodiments the hinge is used as this domain linker). While any suitable linker can be used, many embodiments utilize a glycine-serine polymer as the domain linker, including for example (GS)n, (GSGGS)n (SEQ ID NO: 3), (GGGGS)n (SEQ ID NO: 2), and (GGGS)n (SEQ ID NO: 4), where n is an integer of at least one (and generally from 3 to 4 to 5) as well as any peptide sequence that allows for recombinant attachment of the two domains with sufficient length and flexibility to allow each domain to retain its biological function. In some cases, and with attention being paid to “strandedness”, as outlined below, charged domain linkers, as used in some embodiments of scFv linkers can be used. Exemplary useful domain linkers are depicted in FIG. 6.

With particular reference to the domain linker used to attach the scFv domain to the Fc domain in the “2+1” format, there are several domain linkers that find particular use, including “full hinge C220S variant,” “flex half hinge,” “charged half hinge 1,” and “charged half hinge 2” as shown in FIG. 6.

In some embodiments, the linker is a “scFv linker”, used to covalently attach the VH and VL domains as discussed herein. In many cases, the scFv linker is a charged scFv linker, a number of which are shown in FIG. 5. Accordingly, in some embodiments, the antibodies described herein further provide charged scFv linkers, to facilitate the separation in pI between a first and a second monomer. That is, by incorporating a charged scFv linker, either positive or negative (or both, in the case of scaffolds that use scFvs on different monomers), this allows the monomer comprising the charged linker to alter the pI without making further changes in the Fc domains. These charged linkers can be substituted into any scFv containing standard linkers. Again, as will be appreciated by those in the art, charged scFv linkers are used on the correct “strand” or monomer, according to the desired changes in pI. For example, as discussed herein, to make 1+1 Fab-scFv-Fc format heterodimeric antibody, the original pI of the Fv region for each of the desired antigen binding domains are calculated, and one is chosen to make an scFv, and depending on the pI, either positive or negative linkers are chosen.

Charged domain linkers can also be used to increase the pI separation of the monomers of the antibodies described herein as well, and thus those included in FIG. 5 can be used in any embodiment herein where a linker is utilized.

G. Useful Formats of the Invention

As will be appreciated by those in the art and discussed more fully below, the heterodimeric bispecific antibodies provided herein can take on a wide variety of configurations, as are generally depicted in FIG. 15 as well as FIG. 44. Some figures depict “single ended” configurations, where there is one type of specificity on one “arm” of the molecule and a different specificity on the other “arm”. Other figures depict “dual ended” configurations, where there is at least one type of specificity at the “top” of the molecule and one or more different specificities at the “bottom” of the molecule. Thus, in some embodiments, the antibodies described herein are directed to novel immunoglobulin compositions that co-engage a different first and a second antigen.

As will be appreciated by those in the art, the heterodimeric formats of the antibodies described herein can have different valencies as well as be bispecific. That is, heterodimeric antibodies of the antibodies described herein can be bivalent and bispecific, wherein one target tumor antigen (e.g. CD3) is bound by one binding domain and the other target tumor antigen (e.g. GPC3) is bound by a second binding domain. The heterodimeric antibodies can also be trivalent and bispecific, wherein the first antigen is bound by two binding domains and the second antigen by a second binding domain. As is outlined herein, when CD3 is one of the target antigens, it is preferable that the CD3 is bound only monovalently, to reduce potential side effects.

The antibodies described herein utilize anti-CD3 antigen binding domains in combination with anti-GPC3 binding domains. As will be appreciated by those in the art, any collection of anti-CD3 CDRs, anti-CD3 variable light and variable heavy domains, Fabs and scFvs as depicted in any of the Figures can be used. Similarly, any of the anti-GPC3 antigen binding domains can be used, whether CDRs, variable light and variable heavy domains, Fabs and scFvs as depicted in any of the Figures can be used, optionally and independently combined in any combination.

1. 1+1 Fab-scFv-Fc Format

One heterodimeric scaffold that finds particular use in the antibodies described herein is the “1+1 Fab-scFv-Fc” or “bottle-opener” format as shown in FIG. 15A with an exemplary combination of a CD3 binding domain and a tumor target antigen (GPC3) binding domain. In this embodiment, one heavy chain monomer of the antibody contains a single chain Fv (“scFv”, as defined below) and an Fc domain. The scFv includes a variable heavy domain (VH1) and a variable light domain (VL1), wherein the VH1 is attached to the VL1 using an scFv linker that can be charged (see, e.g., FIG. 5). The scFv is attached to the heavy chain using a domain linker (see, e.g., FIG. 6). The other heavy chain monomer is a “regular” heavy chain (VH-CH1-hinge-CH2-CH3). The 1+1 Fab-scFv-Fc also includes a light chain that interacts with the VH-CH1 to form a Fab. This structure is sometimes referred to herein as the “bottle-opener” format, due to a rough visual similarity to a bottle-opener. The two heavy chain monomers are brought together by the use of amino acid variants (e.g., heterodimerization variants, discussed above) in the constant regions (e.g., the Fc domain, the CH1 domain and/or the hinge region) that promote the formation of heterodimeric antibodies as is described more fully below.

There are several distinct advantages to the present “1+1 Fab-scFv-Fc” format. As is known in the art, antibody analogs relying on two scFv constructs often have stability and aggregation problems, which can be alleviated in the antibodies described herein by the addition of a “regular” heavy and light chain pairing. In addition, as opposed to formats that rely on two heavy chains and two light chains, there is no issue with the incorrect pairing of heavy and light chains (e.g. heavy 1 pairing with light 2, etc.).

Many of the embodiments outlined herein rely in general on the 1+1 Fab-scFv-Fc or “bottle opener” format antibody that comprises a first monomer comprising an scFv, comprising a variable heavy and a variable light domain, covalently attached using an scFv linker (charged, in many but not all instances), where the scFv is covalently attached to the N-terminus of a first Fc domain usually through a domain linker The domain linker can be either charged or uncharged and exogenous or endogenous (e.g., all or part of the native hinge domain). Any suitable linker can be used to attach the scFv to the N-terminus of the first Fc domain. In some embodiments, the domain linker is chosen from the domain linkers in FIG. 6. The second monomer of the 1+1 Fab-scFv-Fc format or “bottle opener” format is a heavy chain, and the composition further comprises a light chain.

In general, in many preferred embodiments, the scFv is the domain that binds to the CD3, and the Fab forms an GPC3 binding domain. An exemplary anti-GPC3×anti-CD3 bispecific antibody in the 1+1 Fab-scFv-Fc format is depicted in FIG. 15A. Exemplary anti-GPC3×anti-CD3 bispecific antibodies in the 1+1 Fab-scFv-Fc format are depicted in FIGS. 16 and 17.

In addition, the Fc domains of the antibodies described herein generally include skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L; K370S: S364K/E357Q; T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In certain embodiments, the 1+1 Fab-scFv-Fc scaffold format includes a first monomer that includes a scFv-domain linker-CH2-CH3 monomer, a second monomer that includes a first variable heavy domain-CH1-hinge-CH2-CH3 monomer and a third monomer that includes a first variable light domain. In some embodiments, the CH2-CH3 of the first monomer is a first variant Fc domain and the CH2-CH3 of the second monomer is a second variant Fc domain. In some embodiments, the scFv includes a scFv variable heavy domain and a scFv variable light domain that form a CD3 binding moiety. In certain embodiments, the scFv variable heavy domain and scFv variable light domain are covalently attached using an scFv linker (charged, in many but not all instances. See, e.g., FIG. 5). In some embodiments, the first variable heavy domain and first variable light domain form a GPC3 binding domain.

In some embodiments, the 1+1 Fab-scFv-Fc format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include 1+1 Fab-scFv-Fc formats that comprise: a) a first monomer (the “scFv monomer”) that comprises a charged scFv linker (with the +H sequence of FIG. 5 being preferred in some embodiments), the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and an scFv that binds to CD3 as outlined herein; b) a second monomer (the “Fab monomer”) that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain; and c) a light chain that includes a variable light domain light domain (VL) and a constant light domain (CL), wherein numbering is according to EU numbering. The variable heavy domain and variable light domain make up an GPC3 binding moiety.

Any suitable CD3 ABD can be included in the 1+1 Fab-scFv-Fc format antibody, included those provided herein. CD3 binding domain sequences finding particular use in these embodiments include, but are not limited to, H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31 or a variant thereof, as well as those depicted in FIG. 10 and those depicted in FIGS. 14 and 15 of WO2014/145806, hereby incorporated by reference including the Legends.

Any suitable GPC3 ABD can be included in the 1+1 Fab-scFv-Fc format antibody, included those provided herein. GPC3 ABDs that are of particular use in these embodiments include, but are not limited to, VH and VL domains selected from have VH/VL pairs selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73 or a variant thereof.

In particular embodiments, the VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69 or a variant thereof.

In some embodiments, the 1+1 Fab-scFv-Fc format includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments include 1+1 Fab-scFv-Fc formats that comprise: a) a first monomer (the “scFv monomer”) that comprises a charged scFv linker (with the +H sequence of FIG. 5 being preferred in some embodiments), the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and an scFv that binds to CD3 as outlined herein; b) a second monomer (the “Fab monomer”) that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S, and a variable heavy domain; and c) a light chain that includes a variable light domain (VL) and a constant light domain (CL), wherein numbering is according to EU numbering. The variable heavy domain and variable light domain make up a GPC3 binding domain. CD3 binding domain sequences finding particular use in these embodiments include, but are not limited to, H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31 or a variant thereof, as well as those depicted in FIGS. 10. GPC3 binding domain sequences that are of particular use in these embodiments include, but are not limited to, the αGPC3 ABD VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69.

Particularly useful GPC3 and CD3 sequence combinations for use with the 1+1 format antibody include, for example, are disclosed in FIGS. 16 and 17.

FIGS. 7A-7D show some exemplary Fc domain sequences that are useful in the 1+1 Fab-scFv-Fc format antibodies. The “monomer 1” sequences depicted in FIGS. 7A-7D typically refer to the Fc domain of the “Fab-Fc heavy chain” and the “monomer 2” sequences refer to the Fc domain of the “scFv-Fc heavy chain.” Further, FIG. 9 provides useful CL sequences that can be used with this format.

In some embodiments, any of the VH and VL sequences depicted herein (including all VH and VL sequences depicted in the Figures and Sequence Listings, including those directed to GPC3) can be added to the bottle opener backbone formats of FIG. 7A-7D as the “Fab side”, using any of the anti-CD3 scFv sequences shown in the Figures and Sequence Listings.

For bottle opener backbone 1 from FIG. 7A, (optionally including the 428L/434S variants), CD binding domain sequences finding particular use in these embodiments include, but are not limited to, CD3 binding domain 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 attached as the scFv side of the backbones shown in FIG. 7.

2. mAb-Fv

One heterodimeric scaffold that finds particular use in the antibodies described herein is the mAb-Fv format (FIG. 44G). In this embodiment, the format relies on the use of a C-terminal attachment of an “extra” variable heavy domain to one monomer and the C-terminal attachment of an “extra” variable light domain to the other monomer, thus forming a third antigen binding domain, wherein the Fab portions of the two monomers bind a GPC3 and the “extra” scFv domain binds CD3.

In this embodiment, the first monomer comprises a first heavy chain, comprising a first variable heavy domain and a first constant heavy domain comprising a first Fc domain, with a first variable light domain covalently attached 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 domain of the second constant heavy domain comprising a second Fc domain, and a third variable heavy domain covalently attached 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 attached variable domains make up a Fv that binds CD3 (as it is less preferred to have bivalent CD3 binding). This embodiment further utilizes a common light chain comprising a variable light domain and a constant light domain that associates with the heavy chains to form two identical Fabs that bind a GPC3. As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

The antibodies described herein provide mAb-Fv formats where the CD3 binding domain sequences are as shown in FIG. 10 or a variant thereof. The antibodies described herein provide mAb-Fv formats wherein the GPC3 binding domain sequences are as shown in FIGS. 12 and 13 or a variant thereof.

In addition, the Fc domains of the mAb-Fv format comprise skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L, K370S: S364K/E357Q, T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In some embodiments, the mAb-Fv format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include mAb-Fv formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a first variable heavy domain that, with the first variable light domain of the light chain, makes up an Fv that binds to GPC3, and a second variable heavy domain; b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and a first variable heavy domain that, with the first variable light domain, makes up the Fv that binds to GPC3 as outlined herein, and a second variable light chain, that together with the second variable heavy domain forms an Fv (ABD) that binds to CD3; and c) a light chain comprising a first variable light domain and a constant light domain.

In some embodiments, the mAb-Fv format includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments include mAb-Fv formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a first variable heavy domain that, with the first variable light domain of the light chain, makes up an Fv that binds to GPC3, and a second variable heavy domain; b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a first variable heavy domain that, with the first variable light domain, makes up the Fv that binds to GPC3 as outlined herein, and a second variable light chain, that together with the second variable heavy domain of the first monomer forms an Fv (ABD) that binds CD3; and c) a light chain comprising a first variable light domain and a constant light domain.

3. mAb-scFv

One heterodimeric scaffold that finds particular use in the antibodies described herein is the mAb-scFv format (FIG. 44H). In this embodiment, the format relies on the use of a C-terminal attachment of a scFv to one of the monomers, thus forming a third antigen binding domain, wherein the Fab portions of the two monomers bind GPC3 and the “extra” scFv domain binds CD3. Thus, the first monomer comprises a first heavy chain (comprising a variable heavy domain and a constant domain), with a C-terminally covalently attached scFv comprising a scFv variable light domain, an scFv linker and a scFv variable heavy domain in either orientation (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 comprising a variable light domain and a constant light domain, that associates with the heavy chains to form two identical Fabs that bind GPC3. As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

The antibodies described herein provide mAb-scFv formats where the CD binding domain sequences are as shown in FIG. 10A-10F or a variant thereof, and the GPC3 binding domain sequences are as shown in FIGS. 12 and 13 or a variant thereof.

In addition, the Fc domains of the mAb-scFv format comprise skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L, K370S: S364K/E357Q, T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In some embodiments, the mAb-scFv format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include mAb-scFv formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein; and c) a common light chain comprising a variable light domain and a constant light domain.

In some embodiments, the mAb-scFv format includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments include mAb-scFv formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein; and c) a common light chain comprising a variable light domain and a constant light domain.

4. 2+1 Fab₂-scFv-Fc Format

One heterodimeric scaffold that finds particular use in the antibodies described herein is the “2+1 Fab₂-scFv-Fc” format (also referred to in previous related filings as “Central-scFv format”) shown in FIG. 15B with an exemplary combination of a CD3 binding domain and two tumor target antigen (GPC3) binding domains. In this embodiment, the format relies on the use of an inserted scFv domain thus forming a third antigen binding domain, wherein the Fab portions of the two monomers bind GPC3 and the “extra” scFv domain binds CD3. The scFv domain is inserted between the Fc domain and the CH1-Fv region of one of the monomers, thus providing a third antigen binding domain. As described, GPC3×CD3 bispecific antibodies having the 2+1 Fab2-scFv-Fc format are potent in inducing redirected T cell cytotoxicity in cellular environments that express low levels of GPC3. Moreover, as shown in the examples, GPC3×CD3 bispecific antibodies having the 2+1 Fab2-scFv-Fc format allow for the “fine tuning” of immune responses as such antibodies exhibit a wide variety of different properties, depending on the GPC3 and/or CD3 binding domains used. For example, such antibodies exhibit differences in selectivity for cells with different GPC3 expression, potencies for GPC3 expressing cells, ability to elicit cytokine release, and sensitivity to soluble GPC3. These GPC3 antibodies find use, for example, in the treatment of GPC3 associated cancers.

In this embodiment, one monomer comprises a first heavy chain comprising a first variable heavy domain, a CH1 domain (and optional hinge) and Fc domain, with a scFv comprising a scFv variable light domain, an scFv linker and a scFv variable heavy domain. The scFv is covalently attached between the C-terminus of the CH1 domain of the heavy constant domain and the N-terminus of the first Fc domain using optional domain linkers (VH1-CH1-[optional linker]-VH2-scFv linker-VL2-[optional linker including the hinge]-CH2-CH3, or the opposite orientation for the scFv, VH1-CH1-[optional linker]-VL2-scFv linker-VH2-[optional linker including the hinge]-CH2-CH3). The optional linkers can be any suitable peptide linkers, including, for example, the domain linkers included in FIG. 6. In some embodiments, the optional linker is a hinge or a fragment thereof. The other monomer is a standard Fab side (i.e., VH1-CH1-hinge-CH2-CH3). This embodiment further utilizes a common light chain comprising a variable light domain and a constant light domain, that associates with the heavy chains to form two identical Fabs that bind GPC3. As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

In one embodiment, the 2+1 Fab2-scFv-Fc format antibody includes an scFv with the VH and VL of a CD3 binding domain sequence depicted in FIG. 10. In one embodiment, the 2+1 Fab2-scFv-Fc format antibody includes two Fabs having the VH and VL of a GPC3 binding domain as shown in FIGS. 12 and 13.

In exemplary embodiments, the GPC3 binding domain of the 2+1 Fab2-scFv-Fc GPC3×CD3 bispecific antibody includes the VH and VL CD3 binding domain sequences finding particular use in these embodiments include, but are not limited to, H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31 or a variant thereof, as well as those depicted in FIG. 10 and those depicted in FIGS. 14 and 15 of WO2014/145806, hereby incorporated by reference including the Legends.

Any suitable GPC3 ABD can be included in the 2+1 Fab₂-scFv-Fc format antibody, included those provided herein. GPC3 ABDs that are of particular use in these embodiments include, but are not limited to, VH and VL domains selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73 or a variant thereof.

In particular embodiments, the VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69 or a variant thereof.

In addition, the Fc domains of the 2+1 Fab₂-scFv-Fc format comprise skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L, K370S: S364K/E357Q, T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In some embodiments, the 2+1 Fab₂-scFv-Fc format antibody includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include 2+1 Fab₂-scFv-Fc formats that comprise: a) a first monomer (the Fab-scFv-Fc side) that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein, and an scFv domain that binds to CD3; b) a second monomer (the Fab-Fc side) that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein; and c) a common light chain comprising the variable light domain and a constant light domain, where numbering is according to EU numbering. In some embodiments, the αGPC3 VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69.

In some embodiments, the 2+1 Fab₂-scFv-Fc format antibody includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments include 2+1 Fab₂-scFv-Fc formats that comprise: a) a first monomer (the Fab-scFv-Fc side) that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein, and an scFv domain that binds to CD3; b) a second monomer (the Fab-Fc side) that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein; and c) a common light chain comprising a variable light domain and a constant light domain, where numbering is according to EU numbering. In some embodiments, the αGPC3 VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69. CD3 binding domain sequences finding particular use in these embodiments include, but are not limited to, H1.30_L1.47, H1.32_L1.47, H1.89_L1.47, H1.90_L1.47, H1.33_L1.47, H1.31_L1.47, L1.47_H1.30, L1.47_H1.30, L1.47_H1.32, L1.47_H1.89, L1.47_H1.90, L1.47_H1.33, and L1.47_H1.31 or a variant thereof.

FIGS. 8A-8C shows some exemplary Fc domain sequences that are useful with the 2+1 Fab₂-scFv-Fc format. The “monomer 1” sequences depicted in FIGS. 8A-8C typically refer to the Fc domain of the “Fab-Fc heavy chain” and the “monomer 2” sequences refer to the Fc domain of the “Fab-scFv-Fc heavy chain.” Further, FIG. 9 provides useful CL sequences that can be used with this format.

Exemplary anti-GPC3×anti-CD3 2+1 Fab₂-scFv-Fc format antibodies are depicted in FIGS. 18-21.

5. Central-Fv

One heterodimeric scaffold that finds particular use in the antibodies described herein is the Central-Fv format (FIG. 44I). In this embodiment, the format relies on the use of an inserted Fv domain (i.e., the central Fv domain) thus forming an “extra” third antigen binding domain, wherein the Fab portions of the two monomers bind a GPC3 and the “extra” central Fv domain binds CD3. The “extra” central Fv domain is inserted between the Fc domain and the CH1-Fv region of the monomers, thus providing a third antigen binding domain (i.e., the “extra” central Fv domain), wherein each monomer contains a component of the “extra” central Fv domain (i.e., one monomer comprises the variable heavy domain and the other a variable light domain of the “extra” central Fv domain).

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

This embodiment further utilizes a common light chain comprising a variable light domain and a constant light domain, that associates with the heavy chains to form two identical Fabs that each bind an GPC3. As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

The antibodies described herein provide central-Fv formats where the CD3 binding domain sequences are as shown in FIG. 10 or a variant thereof, and the GPC3 binding domain sequences are as shown in FIGS. 12 and 13 or a variant thereof.

6. One-Armed Central-scFv

One heterodimeric scaffold that finds particular use in the antibodies described herein is the one-armed central-scFv format (FIG. 44C). In this embodiment, one monomer comprises just an Fc domain, while the other monomer includes a Fab domain (a first antigen binding domain), a scFv domain (a second antigen binding domain) and an Fc domain, where the scFv domain is inserted between the Fc domain and the Fc domain. In this format, the Fab portion binds one receptor target and the scFv binds another. In this format, either the Fab portion binds a GPC3 and the scFv binds CD3 or vice versa.

In this embodiment, one monomer comprises a first heavy chain comprising a first variable heavy domain, a CH1 domain and Fc domain, with a scFv comprising a scFv variable light domain, an scFv linker and a scFv variable heavy domain. The scFv is covalently attached between the C-terminus of the CH1 domain of the heavy constant domain and the N-terminus of the first Fc domain using domain linkers, in either orientation, VH1-CH1-[optional domain linker]-VH2-scFv linker-VL2-[optional domain linker]-CH2-CH3 or VH1-CH1-[optional domain linker]-VL2-scFv linker-VH2-[optional domain linker]-CH2-CH3. The second monomer comprises an Fc domain (CH2-CH3). This embodiment further utilizes a light chain comprising a variable light domain and a constant light domain that associates with the heavy chain to form a Fab.

As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

In addition, the Fc domains of the one-armed central-scFv format generally include skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L, K370S: S364K/E357Q, T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In some embodiments, the one-armed central-scFv format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments of the one-armed central-scFv formats comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that includes an Fc domain having the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K; and c) a light chain comprising a variable light domain and a constant light domain.

In some embodiments, the one-armed central-scFv format includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments of the one-armed central-scFv formats comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that includes an Fc domain having the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and the FcRn variants M428L/N434S; and c) a light chain comprising a variable light domain and a constant light domain.

7. One-Armed scFv-mAb

One heterodimeric scaffold that finds particular use in the antibodies described herein is the one-armed scFv-mAb format (FIG. 44D). In this embodiment, one monomer comprises just an Fc domain, while the other monomer uses a scFv domain attached at 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 orientation) VL-scFv linker-VH-[optional domain linker]-CH1-hinge-CH2-CH3. In this format, the Fab portions each bind GPC3 and the scFv binds CD3. This embodiment further utilizes a light chain comprising a variable light domain and a constant light domain, that associates with the heavy chain to form a Fab. As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

The antibodies described herein provide one-armed scFv-mAb formats where the CD3 binding domain sequences are as shown in FIG. 10 or a variant thereof, and wherein the GPC3 binding domain sequences are as shown in FIGS. 12 and 13 or a variant thereof.

In addition, the Fc domains of the one-armed scFv-mAb format generally include skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L, K370S: S364K/E357Q, T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In some embodiments, the one-armed scFv-mAb format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments of the one-armed scFv-mAb formats comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that includes an Fc domain having the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K; and c) a light chain comprising a variable light domain and a constant light domain.

In some embodiments, the one-armed scFv-mAb format includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments one-armed scFv-mAb formats comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that includes an Fc domain having the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and the FcRn variants M428L/N434S; and c) a light chain comprising a variable light domain and a constant light domain.

8. scFv-mAb

One heterodimeric scaffold that finds particular use in the antibodies described herein is the mAb-scFv format (FIG. 44E). In this embodiment, the format relies on the use of a N-terminal attachment of a scFv to one of the monomers, thus forming a third antigen binding domain, wherein the Fab portions of the two monomers bind GPC3 and the “extra” scFv domain binds CD3.

In this embodiment, the first monomer comprises a first heavy chain (comprising a variable heavy domain and a constant domain), with a N-terminally covalently attached scFv comprising a scFv variable light domain, an scFv linker and a scFv variable heavy domain in either orientation ((VH1-scFv linker-VL1-[optional domain linker]-VH2-CH1-hinge-CH2-CH3) or (with the scFv in the opposite orientation) ((VL1-scFv linker-VH1-[optional domain linker]-VH2-CH1-hinge-CH2-CH3)). This embodiment further utilizes a common light chain comprising a variable light domain and a constant light domain that associates with the heavy chains to form two identical Fabs that bind GPC3. As for many of the embodiments herein, these constructs include skew variants, pI variants, ablation variants, additional Fc variants, etc. as desired and described herein.

The antibodies described herein provide scFv-mAb formats where the CD3 binding domain sequences are as shown in FIG. 10 or a variant thereof, and wherein the GPC3 binding domain sequences are as shown in FIGS. 12 and 13 or a variant thereof.

In addition, the Fc domains of the scFv-mAb format generally include skew variants (e.g. a set of amino acid substitutions as shown in FIG. 1, with particularly useful skew variants being selected from the group consisting of S364K/E357Q: L368D/K370S; L368D/K370S: S364K; L368E/K370S: S364K; T411T/E360E/Q362E: D401K; L368D/K370S: S364K/E357L, K370S: S364K/E357Q, T366S/L368A/Y407V: T366W and T366S/L368A/Y407V/Y349C: T366W/S354C), optionally ablation variants (including those shown in FIG. 3), optionally charged scFv linkers (including those shown in FIG. 5) and the heavy chain comprises pI variants (including those shown in FIG. 2).

In some embodiments, the scFv-mAb format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include scFv-mAb formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein; and c) a common light chain comprising a variable light domain and a constant light domain.

In some embodiments, the scFv-mAb format includes skew variants, pI variants, ablation variants and FcRn variants. Accordingly, some embodiments include scFv-mAb formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein, and a scFv domain that binds to CD3; b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a variable heavy domain that, with the variable light domain of the common light chain, makes up an Fv that binds to GPC3 as outlined herein; and c) a common light chain comprising a variable light domain and a constant light domain.

9. Dual scFv Formats

The antibodies described herein also provide dual scFv formats as are known in the art (FIG. 44B). In this embodiment, the GPC3×CD3 heterodimeric bispecific antibody is made up of two scFv-Fc monomers (both in either (VH-scFv linker-VL-[optional domain linker]-CH2-CH3) format or (VL-scFv linker-VH-[optional domain linker]-CH2-CH3) format, or with one monomer in one orientation and the other in the other orientation.

The antibodies described herein provide dual scFv formats where the CD3 binding domain sequences are as shown in FIG. 10 or a variant thereof, and wherein the GPC3 binding domain sequences are as shown in FIGS. 12 and 13 or a variant thereof.

In some embodiments, the dual scFv format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include dual scFv formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, and a first scFv that binds either CD3 or GPC3; and b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants

E233P/L234V/L235A/G236del/S267K, and a second scFv that binds either CD3 or GPC3. In some embodiments, the dual scFv format includes skew variants, pI variants, ablation variants and FcRn variants. In some embodiments, the dual scFv format includes skew variants, pI variants, and ablation variants. Accordingly, some embodiments include dual scFv formats that comprise: a) a first monomer that comprises the skew variants S364K/E357Q, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a first scFv that binds either CD3 or GPC3; and b) a second monomer that comprises the skew variants L368D/K370S, the pI variants N208D/Q295E/N384D/Q418E/N421D, the ablation variants E233P/L234V/L235A/G236del/S267K, the FcRn variants M428L/N434S and a second scFv that binds either CD3 or GPC3.

10. Non-Heterodimeric Bispecific Antibodies

As will be appreciated by those in the art, the anti-GPC3×anti-CD3 antibodies provided herein can also be included in non-heterodimeric bispecific formats (see FIG. 44J). In this format, the anti-GPC3×anti-CD3 includes: 1) a first monomer comprising a VH1-CH1-hinge-CH2-CH3; 2) a second monomer comprising a VH2-CH1-hinge-CH2-CH3; 3) a first light chain comprising a VL1-CL; and 4) a second light chain comprising a VL2-CL. In such embodiments, the VH1 and VL1 form a first antigen binding domain and VH2 and VL2 form a second antigen binding domain. One of the first or second antigen binding domains binds GPC3 and the other antigen binding domain binds CD3.

Any suitable GPC3 binding domain and CD3 binding domain can be included in the anti-GPC3×anti-CD3 antibody in the non-heterodimeric bispecific antibody format, including any of the GPC3 binding domains and CD3 binding domains and related VHs and VLs provided herein or a variant thereof (see, e.g., FIGS. 10, 12 and 13).

11. Trident Format

In some embodiments, the bispecific antibodies described herein are in the “Trident” format as generally described in WO2015/184203, hereby expressly incorporated by reference in its entirety and in particular for the Figures, Legends, definitions and sequences of “Heterodimer-Promoting Domains” or “HPDs”, including “K-coil” and “E-coil” sequences. Tridents rely on using two different HPDs that associate to form a heterodimeric structure as a component of the structure. In this embodiment, the Trident format include a “traditional” heavy and light chain (e.g., VH1-CH1-hinge-CH2-CH3 and VL1-CL), a third chain comprising a first “diabody-type binding domain” or “DART”, VH2-(linker)-VL3-HPD1 and a fourth chain comprising a second DART®, VH3-(linker)-(inker)-VL2-HPD2. The VH1 and VL1 form a first ABD, the VH2 and VL2 form a second ABD, and the VH3 and VL3 form a third ABD. In some cases, as is shown in FIG. 1K, the second and third ABDs bind the same antigen, in this instance generally GPC3, e.g., bivalently, with the first ABD binding a CD3 monovalently.

Any suitable GPC3 binding domain and CD3 binding domain can be included in the anti-GPC3×anti-CD3 antibody in the Trident bispecific antibody format, including any of the GPC3 binding domains and CD3 binding domains and related VHs and VLs provided herein or a variant thereof (see, e.g., FIGS. 10, 12 and 13).

12. Monospecific, Monoclonal Antibodies

As will be appreciated by those in the art, the novel Fv sequences outlined herein can also be used in both monospecific antibodies (e.g., “traditional monoclonal antibodies”) or non-heterodimeric bispecific formats. Accordingly, in some embodiments, the antibodies described herein provide monoclonal (monospecific) antibodies comprising the 6 CDRs and/or the vh and v sequences from the figures, generally with IgG1, IgG2, IgG3 or IgG4 constant regions, with IgG1, IgG2 and IgG4 (including IgG4 constant regions comprising a S228P amino acid substitution) finding particular use in some embodiments. That is, any sequence herein with a “H_L” designation can be linked to the constant region of a human IgG1 antibody.

In some embodiments, the monospecific antibody is an GPC3 monospecific antibody that has a VH/VL pairs selected from the group consisting of: [GPC3]H1_L1, [GPC3]H1_L1.1, [GPC3]H1_L1.2, [GPC3]H1_L1.3, [GPC3]H1_L1.4, [GPC3]H1_L1.5, [GPC3]H1_L1.6, [GPC3]H1_L1.7, [GPC3]H1_L1.8, [GPC3]H1_L1.9, [GPC3]H1_L1.10, [GPC3]H1_L1.16, [GPC3]H1_L1.23, [GPC3]H1_L1.29, [GPC3]H1_L1.31, [GPC3]H1_L1.65, [GPC3]H1_L1.66, [GPC3]H1_L1.67, [GPC3]H1_L1.68, [GPC3]H1_L1.70, [GPC3]H1_L1.71, [GPC3]H1_L1.72, [GPC3]H1_L1.73, [GPC3]H1.1_L1, [GPC3]H1.1_L1.1, [GPC3]H1.1_L1.2, [GPC3]H1.1_L1.3, [GPC3]H1.1_L1.4, [GPC3]H1.1_L1.5, [GPC3]H1.1_L1.6, [GPC3]H1.1_L1.7, [GPC3]H1.1_L1.8, [GPC3]H1.1_L1.9, [GPC3]H1.1_L1.10, [GPC3]H1.1_L1.16, [GPC3]H1.1_L1.23, [GPC3]H1.1_L1.29, [GPC3]H1.1_L1.31, [GPC3]H1.1_L1.65, [GPC3]H1.1_L1.66, L1.67, [GPC3]H1.1_L1.68, [GPC3]H1.1_L1.70, [GPC3]H1.1_L1.71, [GPC3]H1.1_L1.72, [GPC3]H1.1_L1.73, [GPC3]H1.2_L1, [GPC3]H1.2_L1.1, [GPC3]H1.2_L1.2, [GPC3]H1.2_L1.3, [GPC3]H1.2_L1.4, [GPC3]H1.2_L1.5, [GPC3]H1.2_L1.6, [GPC3]H1.2_L1.7, [GPC3]H1.2_L1.8, [GPC3]H1.2_L1.9, [GPC3]H1.2_L1.10, [GPC3]H1.2_L1.16, [GPC3]H1.2_L1.23, [GPC3]H1.2_L1.29, [GPC3]H1.2_L1.31, [GPC3]H1.2_L1.65, [GPC3]H1.2_L1.66, [GPC3]H1.2_L1.67, [GPC3]H1.2_L1.68, [GPC3]H1.2_L1.70, [GPC3]H1.2_L1.71, [GPC3]H1.2_L1.72, [GPC3]H1.2_L1.73, [GPC3]H1.3_L1, [GPC3]H1.3_L1.1, [GPC3]H1.3_L1.2, [GPC3]H1.3_L1.3, [GPC3]H1.3_L1.4, [GPC3]H1.3_L1.5, [GPC3]H1.3_L1.6, [GPC3]H1.3_L1.7, [GPC3]H1.3_L1.8, [GPC3]H1.3_L1.9, [GPC3]H1.3_L1.10, [GPC3]H1.3_L1.16, [GPC3]H1.3_L1.23, [GPC3]H1.3_L1.29, [GPC3]H1.3_L1.31, [GPC3]H1.3_L1.65, [GPC3]H1.3_L1.66, [GPC3]H1.3_L1.67, [GPC3]H1.3_L1.68, [GPC3]H1.3_L1.70, [GPC3]H1.3_L1.71, [GPC3]H1.3_L1.72, [GPC3]H1.3_L1.73, [GPC3]H1.4_L1, [GPC3]H1.4_L1.1, [GPC3]H1.4_L1.2, [GPC3]H1.4_L1.3, [GPC3]H1.4_L1.4, [GPC3]H1.4_L1.5, [GPC3]H1.4_L1.6, [GPC3]H1.4_L1.7, [GPC3]H1.4_L1.8, [GPC3]H1.4_L1.9, [GPC3]H1.4_L1.10, [GPC3]H1.4_L1.16, [GPC3]H1.4_L1.23, [GPC3]H1.4_L1.29, [GPC3]H1.4_L1.31, [GPC3]H1.4_L1.65, [GPC3]H1.4_L1.66, [GPC3]H1.4_L1.67, [GPC3]H1.4_L1.68, [GPC3]H1.4_L1.70, [GPC3]H1.4_L1.71, [GPC3]H1.4_L1.72 and [GPC3]H1.4_L1.73 or a variant thereof.

In particular embodiments, the VH/VL pairs are selected from the group consisting of [GPC3]H1.1_L1.16 and [GPC3]H1.1_L1.69 or a variant thereof.

H. Particular Embodiments of the Invention

The invention specifically provides 1+1 and 2+1 formats that bind CD3 and GPC3. Certain embodiments include XENP38086 (Xtend analog to XENP34920, meaning they have identical sequences except that the Xtend analog includes 428L/434S on each Fc domain), XENP38087 (Xtend analog to XENP36935, ditto), and XENP38232 (Xtend analog to XENP37625, ditto.)

1. 1+1 Format

In particular 1+1 format embodiments, the αGPC3 ABD is the Fab and has the VH/VL pair [GPC3]H1.1_L1.16 and the αCD3 ABD is a scFv selected from the group consisting of 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31.

In particular 1+1 format embodiments, the αGPC3 ABD is the Fab and has the VH/VL pair [GPC3]H1.1_L1.69, and the αCD3 ABD is a scFv selected from the group consisting of 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31.

Particular embodiments that find use in a number of applications include those listed in FIGS. 16 and 17, including, but not limited to, XENP21971, XENP34364, XENP34365, XENP34367, XENP34368, XENP35843, XENP36140, XENP36931, XENP36932 XENP36933, XENP36934, XENP36935, XENP36936, XENP36937, XENP36938, XENP36939, XENP36941 and XENP38087.

2. 2+1 Format

In particular 2+1 format embodiments, the αGPC3 ABD is the Fab and has the VH/VL pair [GPC3]H1.1_L1.16 and the αCD3 ABD is a scFv selected from the group consisting of 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31.

In particular 2+1 format embodiments, the αGPC3 ABD is the Fab and has the VH/VL pair [GPC3]H1.1_L1.69, and the αCD3 ABD is a scFv selected from the group consisting of 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, anti-CD3 H1.31_L1.47, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.30, anti-CD3 L1.47_H1.32, anti-CD3 L1.47_H1.89, anti-CD3 L1.47_H1.90, anti-CD3 L1.47_H1.33, and anti-CD3 L1.47_H1.31.

Particular embodiments that find use in a number of applications include those listed in FIGS. 18, 19, 20 and 21, including, but not limited to, XENP33744, XENP33745, XENP27259, XENP33746, XENP34919, XENP34920, XENP34921, XENP34922, XENP34923, XENP35840, XENP35840, XENP37246, XENP37247, XENP38086, XENP33747, XENP35841, XENP37624, XENP37625, XENP37626, XENP38232, XENP37430, XENP37433 and XENP33748.

II. Nucleic Acids of the Invention

The disclosure further provides nucleic acid compositions encoding the anti-GPC3 antibodies provided herein, including, but not limited to, anti-GPC3×anti-CD3 bispecific antibodies and GPC3 monospecific antibodies.

As will be appreciated by those in the art, the nucleic acid compositions will depend on the format and scaffold of the heterodimeric protein. Thus, for example, when the format requires three amino acid sequences, such as for the 1+1 Fab-scFv-Fc format (e.g. a first amino acid monomer comprising an Fc domain and a scFv, a second amino acid monomer comprising a heavy chain and a light chain), three nucleic acid sequences can be incorporated into one or more expression vectors for expression. Similarly, some formats (e.g. dual scFv formats such as disclosed in FIG. 44) only two nucleic acids are needed; again, they can be put into one or two expression vectors.

As is known in the art, the nucleic acids encoding the components of the antibodies described herein can be incorporated into expression vectors as is known in the art, and depending on the host cells used to produce the heterodimeric antibodies described herein. Generally the nucleic acids are operably linked to any number of regulatory elements (promoters, origin of replication, selectable markers, ribosomal binding sites, inducers, etc.). The expression vectors can be extra-chromosomal or integrating vectors.

The nucleic acids and/or expression vectors of the antibodies described herein are then transformed into any number of different types of host cells as is well known in the art, including mammalian, bacterial, yeast, insect and/or fungal cells, with mammalian cells (e.g. CHO cells), finding use in many embodiments.

In some embodiments, nucleic acids encoding each monomer and the optional nucleic acid encoding a light chain, as applicable depending on the format, are each contained within a single expression vector, generally under different or the same promoter controls. In embodiments of particular use in the antibodies described herein, each of these two or three nucleic acids are contained on a different expression vector. As shown herein and in 62/025,931, hereby incorporated by reference, different vector ratios can be used to drive heterodimer formation. That is, surprisingly, while the proteins comprise first monomer:second monomer:light chains (in the case of many of the embodiments herein that have three polypeptides comprising the heterodimeric antibody) in a 1:1:2 ratio, these are not the ratios that give the best results.

The heterodimeric antibodies described herein are made by culturing host cells comprising the expression vector(s) as is well known in the art. Once produced, traditional antibody purification steps are done, including an ion exchange chromatography step. As discussed herein, having the pIs of the two monomers differ by at least 0.5 can allow separation by ion exchange chromatography or isoelectric focusing, or other methods sensitive to isoelectric point. That is, the inclusion of pI substitutions that alter the isoelectric point (pI) of each monomer so that such that each monomer has a different pI and the heterodimer also has a distinct pI, thus facilitating isoelectric purification of the “1+1 Fab-scFv-Fc” and “2+1” heterodimers (e.g., anionic exchange columns, cationic exchange columns). These substitutions also aid in the determination and monitoring of any contaminating dual scFv-Fc and mAb homodimers post-purification (e.g., IEF gels, cIEF, and analytical IEX columns).

III. Biological and Biochemical Functionality of the Heterodimeric Bispecific Antibodies

Generally the bispecific GPC3×CD3 antibodies described herein are administered to patients with cancer, and efficacy is assessed, in a number of ways as described herein. Thus, while standard assays of efficacy can be run, such as cancer load, size of tumor, evaluation of presence or extent of metastasis, etc., immuno-oncology treatments can be assessed on the basis of immune status evaluations as well. This can be done in a number of ways, including both in vitro and in vivo assays.

IV. Treatments

Once made, the compositions of the antibodies described herein find use in a number of applications including cancer such as liver cancer, such that the heterodimeric compositions of the antibodies described herein find use in the treatment of such GPC3 positive cancers.

V. Antibody Compositions for In Vivo Administration

Formulations of the antibodies used in accordance with the antibodies described herein are prepared for storage by mixing an antibody having the desired degree of purity with optional pharmaceutically acceptable carriers, excipients or stabilizers (Remington's Pharmaceutical Sciences 16th edition, Osol, A. Ed. [1980]), in the form of lyophilized formulations or aqueous solutions.

VI. Administrative Modalities

The antibodies and chemotherapeutic agents described herein are administered to a subject, in accord with known methods, such as intravenous administration as a bolus or by continuous infusion over a period of time.

VII. Treatment Modalities

In the methods described herein, therapy is used to provide a positive therapeutic response with respect to a disease or condition. By “positive therapeutic response” is intended an improvement in the disease or condition, and/or an improvement in the 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 reduction in the number of neoplastic cells; (2) an increase in neoplastic cell death; (3) inhibition of neoplastic cell survival; (5) inhibition (i.e., slowing to some extent, preferably halting) of tumor growth; (6) an increased patient survival rate; and (7) some relief from one or more symptoms associated with the disease or condition.

Positive therapeutic responses in any given disease or condition can be determined by standardized response criteria specific to that disease or condition. Tumor response can be assessed for changes in tumor morphology (i.e., overall tumor burden, tumor size, and the like) using screening techniques such as magnetic resonance imaging (MRI) scan, x-radiographic imaging, computed tomographic (CT) scan, bone scan imaging, endoscopy, and tumor biopsy sampling including bone marrow aspiration (BMA) and counting of tumor cells in the circulation.

In addition to these positive therapeutic responses, the subject undergoing therapy may experience the beneficial effect of an improvement in the symptoms associated with the disease.

Treatment according to the disclosure includes a “therapeutically effective amount” of the medicaments used. A “therapeutically effective amount” refers to an amount effective, at dosages and for periods of time necessary, to achieve a desired therapeutic result.

A therapeutically effective amount may vary according to factors such as the disease state, age, sex, and weight of the individual, and the ability of the medicaments to elicit a desired response in the individual. A therapeutically effective amount is also one in which any toxic or detrimental effects of the antibody or antibody portion are outweighed by the therapeutically beneficial effects.

A “therapeutically effective amount” for tumor therapy may also be measured by its ability to stabilize the progression of disease. The ability of a compound to inhibit cancer may be evaluated in an animal model system predictive of efficacy in human tumors.

Alternatively, this property of a composition may be evaluated by examining the ability of the compound to inhibit cell growth or to induce apoptosis by in vitro assays known to the skilled practitioner. A therapeutically effective amount of a therapeutic compound may decrease tumor size, or otherwise ameliorate symptoms in a subject. One of ordinary skill in the art would be able to determine such amounts based on such factors as the subject's size, the severity of the subject's symptoms, and the particular composition or route of administration selected.

Dosage regimens are adjusted to provide the optimum desired response (e.g., a therapeutic response). For example, a single bolus may be administered, several divided doses may be administered over time or the dose may be proportionally reduced or increased as indicated by the exigencies of the therapeutic situation. Parenteral compositions may be formulated in dosage unit form for ease of administration and uniformity of dosage. Dosage unit form as used herein refers to physically discrete units suited as unitary dosages for the subjects to be treated; each unit contains a predetermined quantity of active compound calculated to produce the desired therapeutic effect in association with the required pharmaceutical carrier.

The specification for the dosage unit forms of the disclosure are dictated by and directly dependent on (a) the unique characteristics of the active compound and the particular therapeutic effect to be achieved, and (b) the limitations inherent in the art of compounding such an active compound for the treatment of sensitivity in individuals.

The efficient dosages and the dosage regimens for the bispecific antibodies described herein depend on the disease or condition to be treated and may be determined by the persons skilled in the art.

An exemplary, non-limiting range for a therapeutically effective amount of an bispecific antibody used in the antibodies described herein is about 0.1-100 mg/kg.

All cited references are herein expressly incorporated by reference in their entirety.

Whereas particular embodiments of the disclosure have been described above for purposes of illustration, it will be appreciated by those skilled in the art that numerous variations of the details may be made without departing from the invention as described in the appended claims.

EXAMPLES
A. Example 1: Engineering αGPC3×αCD3 Bispecific Antibodies

A number of formats for αGPC3×αCD3 bispecific antibodies (bsAbs) were conceived, illustrative formats for which are outlined below and in FIG. 15.

One such format is the 1+1 Fab-scFv-Fc format which comprises a single-chain Fv (“scFv”) covalently attached to a first heterodimeric Fc domain, a heavy chain variable region (VH) covalently attached to a complementary second heterodimeric Fc domain, and a light chain (LC) transfected separately so that a Fab domain is formed with the variable heavy domain.

Another format is the 2+1 Fab2-scFv-Fc format which comprises a VH domain covalently attached to a CH1 domain covalently attached to an scFv covalently attached to a first heterodimeric Fc domain (VH-CH1-scFv-Fc), a VH domain covalently attached to a complementary second heterodimeric Fc domain, and a LC transfected separately so that Fab domains are formed with the VH domains.

DNA encoding chains of the αGPC3×αCD3 bsAbs were generated by standard gene synthesis followed by isothermal cloning (Gibson assembly) or subcloning into a pTT5 expression vector containing fusion partners (e.g. domain linkers as depicted in FIG. 6 and/or backbones as depicted in FIGS. 7-8). DNA was transfected into HEK293E cells for expression. Sequences for illustrative αGPC3×αCD3 bsAbs in the 1+1 Fab-scFv-Fc format and in the 2+1 Fab2-scFv-Fc format are depicted respectively in FIGS. 16-21.

B. Example 2: Engineering GPC3 Binding Domains
2A: Humanization of a Murine GPC3 Binding Domain

A murine clone that binds that C terminal peptide of GPC3 (amino acids 524-563) of the GPC3 protein, located close to the cell membrane was humanized using string content optimization (see, e.g., U.S. Pat. No. 7,657,380, issued Feb. 2, 2010). Sequences for the humanized variant are herein referred to as GPC3-A and are depicted in FIG. 12. Variants reduced degradation (e.g. aspartic acid isomerization and deamidation) liability, modulated GPC3 binding affinity, and/or selectivity for high GPC3 expression cell lines. Sequences for illustrative such variants are depicted in FIG. 13.

2B: Tuning the GPC3 Binding Affinity

After humanization, GPC3 binding arms were engineered with single point mutations in the variable light domain with the aim to create a ladder of GPC3 binding affinity, illustrative sequences for which are depicted in FIG. 13. 73 variable light domain variants were engineered and paired with GPC3-A H1.1 (see Example 3C for description of GPC3-A H1.1 variant). The GPC3-A variants were engineered in the 1+1 Fab-scFv-Fc bsAb format and binding for GPC3 antigen was measured using Octet, a BioLayer Interferometry (BLI)-based method. Experimental steps for Octet generally include the following: Immobilization (capture of ligand to a biosensor); Association (dipping of ligand-coated biosensors into wells containing the analyte); and Dissociation (returning of biosensors to well containing buffer). For ease of clinical development (e.g. by investigating the therapeutics in model animals), it is useful for the binding domains to be cross-reactive for cynomolgus antigen; therefore, binding to both human and cynomolgus GPC3 antigen were investigated. His-tagged human and cynomolgus GPC3 were captured on HIS1K sensors then dipped into multiple concentrations of the bispecific antibodies. The resulting dissociation constant (KD) are depicted in FIG. 23. A broad range of KD values were obtained from high affinity variant H1.1_L1.6 (4 nM), to medium affinity variants H1.1_L1.16 and H1.1_L1.29 (100 nM and 70 nM respectively), to low affinity variant H1.1_L1.31 (400 nM).

Next, binding of the variants to cell surface GPC3 was investigated. Illustrative affinity variants H1.1_L1.29, H1.1_L1.16, and H1.1_L1.31 were engineered in the 1+1 Fab-scFv-Fc format. BsAbs XENP36941 (L1.29, 70 nM GPC3 affinity), XENP35843 (L1.16, 100 nM GPC3 affinity), and XENP36140 (L1.31, 400 nM GPC3 affinity) were incubated with GPC3^highHepG2 cells at indicated concentrations for 1 hr at 4° C. Cells were then washed and stained with a secondary antibody (typically anti-human Fc AlexaFlour647) for 1 hr at 4° C. After 2 more washes, cells were analyzed by flow cytometry. The resulting binding curves portrayed in FIG. 24 show that cell binding correlates with the GPC3 affinity, with XENP36941 (70 nM variable light domain) showing the tightest binding, and the XENP36140 (400 nM variable light domain) showing the weakest.

2C: Engineering to Remove Degradation Liable Residues

The sequences for humanized GPC3-A antibodies were investigated for degradation liable residues. The GPC3-A H1 heavy chain CDR2 included D52/P52a (Kabat numbering) as an aspartic acid isomerization motif. Further, the L1 variable light domain contained an N28/G29 (Kabat numbering) as a deamidation motif. Accordingly, a library was made with mutations at these residues to investigate whether the liability could be removed without impacting on GPC3 binding, binding data for which are depicted in FIG. 25. D52E mutation in the variable heavy variant H1.1 removed the aspartic acid isomerization liability (relative to H1) while maintaining similar GPC3 binding (D52N resulted in reduced binding, while D52Q and D52S were acceptable). G29T in the variable light variant L1.6 removed the deamidation liability (relative to L1) while maintaining similar GPC3 binding (N28Q, N28Y, N28S, and N28H drastically reduced binding, while G29A and G29K were acceptable). It should be noted that different mutations may be required to remove the deamidation liability from the affinity variants, binding data for illustrative variants as depicted in FIG. 26. G29A mutation was utilized in the L1.69 variant to remove the deamidation liability from the L1.29 variant (N28Y mutation resulted in decreased binding, while G29K was also acceptable); and G29K mutation was utilized in the L1.73 variant to remove the deamidation liability from the L1.31 variant (N28Y and G29A mutations resulted in low or aberrant response). Sequences for the variants described here are depicted in FIG. 13.

C. Example 3: Tuning and Optimizing αGPC3×αCD3 bsAbs

αGPC3×αCD3 bsAbs were engineered with various affinity-tuned GPC3 and CD3 binding domains and with different GPC3 binding valency and produced as generally described in Example 1 to optimize redirected T cell cytotoxicity (RTCC), selectivity, and potential therapeutic index.

RTCC assays were used to investigate the potential of αGPC3×αCD3 bispecific antibodies (bsAbs) to redirect CD3+ effector T cells to destroy GPC3-expressing cell lines. RTCC assays were generally performed using HepG2 cells (a liver hepatocellular carcinoma line) as GPC3^hightarget cells, Huh7 cells (also a liver hepatocellular carcinoma line) as a GPC3^medtarget cells, and/or HEK293 cells as GPC3^lowtarget cells (a surrogate for cells outside of the tumor environment). Two methods of redirected T cell cytotoxicity (RTCC) assays were used: flow cytometry based, and lactate dehydrogenase (LDH) based. For flow cytometry based RTCC assays, target cells were incubated with human PBMCs and test articles at indicated effector to target cell ratios at 37° C. After incubation, cells were stained with Aqua Zombie stain for 15 minutes at room temperature. Cells were then washed and stained with antibodies for cell surface markers and analyzed by flow cytometry. Induction of RTCC was determined using Zombie Aqua staining on CSFE+ target cells; and activation and degranulation of T cells were determined by CD107a, CD25, and CD69 expression on lymphocytes. For LDH-based RTCC assays, cells and indicated concentrations of test articles would be incubated in a total volume of 200 ul in a flat bottom 96 well plate for 48 or 72 hours at 37° C. Then the cells were lysed and mixed with substrate using the Promega CytoTox-one kit according to its protocol. A Wallac machine was used to read the plate. A V-PLEX proinflammatory panel 1 human kit was used to measure cytokines. It should also be noted that some of the data sets are from the same experiment, as several engineering approaches were simultaneously explored.

Throughout this section, a comparator bsAb XENP31308, which is based on the αGPC3×αCD3 bsAb as disclosed in WO 2016/047722 and sequences for which are depicted in FIG. 22, was also utilized to benchmark the novel bsAbs of the invention. XENP31308 was characterized in RTCC assays as generally described above. Data depicted in FIG. 27 show that at 10:1 effector:target ratio, XENP31308 induced potent RTCC and cytokine secretion on both high and low GPC3 expressing cell lines. In another experiment utilizing 1:1 effector:target ratio data for which are depicted in FIG. 28, XENP31308 induces little RTCC in the presence of low GPC3 expressing cell line HEK293 but still potently induces cytokine secretion. As the GPC3^lowcell line is a surrogate for healthy tissues, maintaining very little to no killing and cytokine release in the presence of a low GPC3 expressing cell line is important as it may translate to less toxicity and cytokine storm in a clinical setting. Strong killing of a GPC3^lowcell line would indicate likelihood of on-target off-tissue killing in vivo, which could cause undesirable toxicity.

3A: Investigating the Effect of GPC3 Binding Domain Binding Different GPC3 Epitopes

GPC3 is membrane bound and consists of an alpha subunit and a beta subunit. The alpha subunit (residues 25-358) may be cleaved and released as soluble GPC3, the while the beta subunit remains attached to the cellular membrane. As described in Example 2A, GPC3 binding domain GPC3-A binds C-terminal residues 524-563 of GPC3 which is part of the beta subunit. Additional GPC3 binding domains have been described to bind the N-terminal residues 359-524 of the beta subunit or the alpha subunit. To investigate the effect of binding to different GPC3 epitopes, αGPC3×αCD3 bsAbs in the 1+1 format with CD3 High or 2+1 format with CD3 High-Int #1 scFv were engineered with different GPC3 binding domains binding different GPC3 epitopes. An RTCC assay was performed using the bsAbs and HepG2 cells (10:1 effector:target ratio). The data in FIG. 29 (depicting activity of the bsAbs as indicated by induction of IFNγ release) show that each of the additional epitopes were less potent at inducing T cell activity than GPC3-A. While some of the difference may be due to differences in binding affinity (between clones) or avidity (between formats), it is likely that the GPC3 epitope contributes to the reduced potency.

3B: Investigating the Effect of GPC3 Binding Avidity

It was hypothesized that bsAbs in the 2+1 Fab₂-scFv-Fc format having bivalent binding for GPC3 may be useful in achieving higher potency with a lower affinity binding domain due to the improved avidity from two binding domains. The avidity benefit of this format may confer selectivity to cell lines with higher target expression, reducing the potential for on-target off-tissue (e.g. healthy tissues expressing low levels of GPC3) effects in vivo. Additionally, high levels of soluble GPC3 may be present in vivo, potentially acting as an antigen sink for GPC3-targeting drugs. Lower GPC3 binding affinity coupled with bivalent GPC3 binding may help overcome the sink by preferably binding cells with high antigen density over soluble antigen. Accordingly, the cell binding of αGPC3×αCD3 bispecific antibodies in the 1+1 Fab-scFv-Fc format, the 2+1 Fab2-scFv-Fc format, and the traditional monospecific bivalent format were investigated. The GPC3-A variable light domains L1.16 (having a GPC3 binding affinity of 100 nM) and L1.31 (having a GPC3 binding affinity of 400 nM) were engineered into the 1+1 Fab-scFv-Fc format, the 2+1 Fab2-scFv-Fc (VL/VH), and bivalent monospecific IgG1 format. Test articles were incubated with GPC3^highHepG2 cells at indicated concentrations, washed, stained with a secondary antibody, washed again, and analyzed by flow cytometry as described above. As seen in FIG. 30, the avidity conferred by the 2+1 Fab₂-scFv-Fc (VL/VH) format improves binding compared to the 1+1 Fab-scFv-Fc format, reaching a level similar to that of the standard bivalent IgG. Additionally, the lower affinity L1.31 clone more drastically shows this improvement over the 1+1 Fab-scFv-Fc indicating that the lower binding affinity creates more of a dependence on avidity and binding valency.

3C: Investigating the Effect of Variant CD3 Binding Domains on RTCC

To investigate the effect of CD3 binding affinity, αGPC3×αCD3 bsAbs were produced in the 1+1 Fab-scFv-Fc and the 2+1 Fab₂-scFv-Fc formats with CD3 High scFv, CD3 High-Int #1 scFv, CD3 High-Int #2 scFv, and CD3 Intermediate scFvs and GPC3-A variants having 100 nM or 400 nM binding affinity for GPC3 and investigated in a flow-based RTCC performed as described above using a 10:1 E:T ratio on GPC3^highHepG2 target cells with a 48 hour incubation time. The data plotted in FIGS. 31A and 32A (respectively for bsAbs in the 1+1 format and in the 2+1 format) showed that affinity-engineered αGPC3×αCD3 bsAbs demonstrated a range of potencies to GPC3^highHepG2 cells, from high (XENP33744 having a 5 nM CD3 arm and 2+1 format, with an EC50 value of 62.63 pg/ml) to low (XENP27259 having a 30 nM CD3 arm in a 1+1 format, with an EC50 value of 19920 pg/ml). The bispecific antibodies also induced release of IFNγ (as depicted in FIGS. 31B and 32B) and T cell activation (as indicated by expression of activation markers such as CD69 and 107a; data not shown) in a manner correlated to their potency. Notably, de-tuning the CD3 binding affinity from CD3 High to CD3 High-Int #1 in the 1+1 Fab-scFv-Fc format provided a much greater potency reduction in induction of cytokine release in comparison to in the 2+1 Fab₂-scFv-Fc format. However, in the 2+1 Fab₂-scFv-Fc format, de-tuning the CD3 binding affinity with CD3 High-Int #2 and CD3 Intermediate still resulted in potency reduction in induction of cytokine release.

3D: Investigating the Effect of Affinity Detuned GPC3 Binding Domains on RTCC

In another experiment, the effect of engineering affinity reductions in the GPC3 binding domains of a set of αGPC3×αCD3 bsAbs as described in Example 2 on the ability to redirect CD3+ effector T cells was investigated. An LDH-based RTCC assay was performed as described above, using GPC3^medHuh7 as target cells, mixed with indicated concentrations of test articles and PBMC effector cells at a 10:1 E:T ratio, for a 48 hour incubation time. FIG. 33 shows that XENP37625 (having a variable light domain L1.69, 70 nM GPC3 affinity), XENP37624 (having a variable light domain L1.16, 100 nm GPC3 affinity), and XENP37626 (having a variable light domain L1.73, 400 nM GPC3 affinity) each demonstrated reduced potency of killing and IFNγ secretion roughly in accordance with their respective affinities.

3E: Investigating the Ability of αGPC3×αCD3 bsAbs to Induce RTCC on Cell Lines with Different GPC3 Expression Levels

On-target, off-tumor toxicity and cytokine release syndrome can result in severe adverse effects in patients; therefore, it is important to tune the bsAbs of the invention to avoid killing and induction of cytokine release in the presence of healthy tissues which express low levels of GPC3.

Accordingly, based on the various observations described above, additional αGPC3×αCD3 bsAbs were engineered, produced, and investigated to identify bsAbs with maximal therapeutic potential with minimal potential for toxicity. Towards this, the effects of modulating GPC3 binding affinity, GPC3 binding valency, and CD3 binding affinity on selectivity of the bispecific antibodies for cell lines with high and low expression were investigated.

HepG2 (GPC3^high) and HEK293 (GPC3^low) cells were each incubated with human PBMCs (1:1 effector to target cell ratio) and indicated concentrations of the test articles for 72 hours at 37° C. Data depicting induction of RTCC and cytokine secretion are depicted in FIGS. 34 and 35. Consistent with the above, the bispecific antibodies demonstrated a range of potencies in inducing RTCC and cytokine release in the presence of GPC3^highHepG2 cell. Notably, in the presence of GPC3^lowHEK293 cells, each of the bispecific antibodies demonstrated little to no induction of RTCC and cytokine release. 3F: Further characterization of αGPC3×αCD3 bsAbs using 2D RTCC on Incucyte

The novel αGPC3×αCD3 bsAbs and the comparator molecule XENP31308 were investigated utilizing a different system (2D RTCC on Incucyte). Indicated concentrations of indicated test articles were incubated with HepG2 or Huh7 cells and T cells at 10:1 or 1:1 effector:target ratio. Data depicting RTCC are depicted in FIGS. 36-38. At a 10:1 effector:target (HepG2) ratio, bsAbs XENP34920, XENP37430, XENP37625, and XENP37624 demonstrated similar potency (less than EC50 of 100 ng/ml) as comparator XENP31308; other bsAbs XENP35843, XENP36935, XENP37433, and XENP37626 were much less potent but were able to achieve efficacious killing at higher concentrations, and XENP36939 did not demonstrate any killing. Notably on lower density Huh7 and at lower effector:target ratio, several of the bsAbs of the invention (e.g. XENP34920, XENP36935, and XENP37625) demonstrate much greater difference in EC50 in comparison to the XENP31308 comparator indicating potential for enhanced therapeutic index. Additionally, when GPC3high HepG2 and GPC3med Huh7 target cells were treated with test articles at a concentration of 10 pg/ml at a 10:1 E:T ratio and observed over different time points ranging from 24 to 144 hours, similar results were seen. As depicted in FIGS. 39-40, the affinity detuned αGPC3×αCD3 bsAbs including XENP34920, XENP37430, XENP36935, XENP37624, and XENP37625 were effective at inhibiting tumor growth over time in both HepG2 and Huh7 target cells, and their efficacy was very similar to that of XENP31308.

3G: Further Characterization of αGPC3×αCD3 bsAbs Using a 3D Spheroid RTCC Model

An additional RTCC system was used to further investigate the affinity detuned αGPC3×αCD3 bsAbs. In this system, target tumor cells grow in a 3D spheroid format, which is physiologically more similar to an in vivo tumor compared to cells growing in a monolayer on a flat surface. In this experiment, 1,000 HepG2 cells (which have a doubling time of 24 hours) were seeded and given 72-96 hours to grow before adding 40,000 PBMCs (resulting in an E:T ratio of approximately 3:1) along with the indicated bsAb at a concentration of 10 ug/ml. The intensity of each signal (tumor or immune cell) was integrated over each well across different time points, and this data is depicted in FIGS. 41-42. FIG. 43 additionally shows photographic images taken of the tumor cells and PBMCs. With PBS treatment only, the tumor spheroids and the PBMCs clustered around them remain unchanged over time, whereas when treated with the αGPC3×αCD3 bsAbs PBMCs proliferate dramatically (as depicted quantitatively in FIG. 42), and the tumor cells are destroyed (as depicted quantitatively in FIG. 41). All αGPC3×αCD3 bsAbs except XENP36939 show 100% efficacy after 144 hours.

The 3D spheroid model also produced an unexpected result in test articles with the lower affinity High-Int #2 CD3 binding domain. Generally, for 2+1 bsAbs in a 2D model, the CD3 affinity difference between the High-Int #1 binding domain and the High-Int #2 binding domain results in roughly a 10-fold potency difference, such as what can be seen in FIG. 38. However, in this 3D model, XENP34920 and XENP37624, both having with the same 100 nM GPC3 binding domain but XENP34920 having the High-Int #1 CD3 binding domain and XENP37624 having the High-Int #2 binding domain, showed potency curves that were very similar. Depicted in FIG. 41, the unexpected result of this 3D model, which should more closely mimic an in vivo model, provides useful insight on the potency of a weaker CD3 binder in this context.

D. Example 4: Identifying αGPC3×αCD3 bsAbs with Optimal Selectivity and Therapeutic Index

Based on the above in vitro experiments, several bsAbs were selected for further analysis in vivo. These antibodies were further engineered with Xtend Fc (M428L/N434S) to enhance serum half-life, illustrative sequences which are depicted in FIGS. 17, 19, and 20 as XENP38086 (Xtend analog to XENP34920), XENP38087 (Xtend analog to XENP36935), and XENP38232 (Xtend analog to XENP37625).

1. 4A: In Vivo Investigation for Toxicity in Cynomolgus

Studies in cynomolgus are planned to investigate whether the improvement in vitro (i.e. selectivity for GPC^highover GPC^lowtarget cells, and reduced cytokine release in the presence of GPC^lowcells) translates to improved safety in an in vivo setting. In a Phase 1 dose escalation study, animals (n=1) are intravenously dosed with 1×, 3×, 10×, 30×, and 60× dose of XENP38086, XENP38087, or XENP38232. Blood is drawn to determine IL-6 concentrations as an indicator of cytokine release syndrome. Animals may be sacrificed to investigate additional signs of toxicity.

	Number	Date	Country
	63159058	Mar 2021	US
	63173127	Apr 2021	US

	Number	Date	Country
Parent	17692045	Mar 2022	US
Child	18358884		US

HETERODIMERIC ANTIBODIES THAT BIND CD3 AND GPC3

Information

Publication Number

Date Filed

Date Published

Inventors

Original Assignees

CPC

International Classifications

Abstract

Description

Claims

PRIORITY CLAIM

Provisional Applications (2)

Continuations (1)