Constrained conditionally activated binding proteins

Abstract

The present invention relates to a conditionally bispecific redirected activation construct or COBRA administered in the form of an active prodrug. Upon exposure to a tumor protease, the constructs are cleaved and activated such that they are able to bind both tumor target antigens (TTAs) and CD3, thereby recruiting CD3-expressing T cells to the tumor for treatment. In some embodiments, the tumor target antigen is B7H3.

Description

[0001] This application is a divisional application based on a patent application with an application date of March 5, 2020, an earliest priority date of March 5, 2019, an application number of 202080034144.2, and an invention name of "Constrained conditionally activated binding protein".

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims priority to U.S. Provisional Application No. 62 / 814,210, filed on March 5, 2019; U.S. Provisional Application No. 62 / 814,744, filed on March 6, 2019; and U.S. Provisional Application No. 62 / 826,523, filed on March 29, 2019, the disclosures of which are incorporated herein by reference in their entireties. Background of the Invention

[0004] In various clinical settings, it is often necessary to selectively destroy individual cells or specific cell types. For example, a major goal of cancer therapy is to specifically destroy tumor cells while leaving healthy cells and tissues as intact as possible. One such approach is to induce an immune response against the tumor so that immune effector cells such as natural killer (NK) cells or cytotoxic T lymphocytes (CTLs) attack and destroy tumor cells.

[0005] The use of intact monoclonal antibodies (mAbs), which provide excellent binding specificity and affinity for tumor-associated antigens, has been successfully applied in the field of cancer treatment and diagnosis. However, the large size of intact mAbs, their poor biodistribution, low efficacy, and persistent presence in the blood pool limit their clinical application. For example, intact antibodies can show specific accumulation within the tumor region. In biodistribution studies, when precisely examining tumors, it is noted that there is an uneven distribution of antibodies with initial accumulation in the peripheral region. Due to tumor necrosis, uneven antigen distribution, and increased interstitial tissue pressure, it is impossible to reach the central part of the tumor with intact antibody constructs. In contrast, smaller antibody fragments show rapid tumor localization, penetrate deeper into the tumor, and are also cleared from the bloodstream relatively quickly. However, many antibodies (including scFv and other constructs) show an "on target / off tumor" effect, in which the molecule is active on non-tumor cells, thereby causing side effects, some of which may be toxic. The present invention relates to novel constructs that are selectively activated in the presence of tumor proteases. Summary of the Invention

[0006] The present invention provides a number of different protein compositions for treating cancer. Thus, in one aspect, the present invention provides a "Form 2" protein comprising, from N-terminus to C-terminus: a first single domain antigen binding domain (sdABD) that binds to a human tumor target antigen (TTA) (sdABD-TTA); b) a domain linker; c) a constrained Fv domain comprising: i) a variable heavy domain comprising vhCDR1, vhCDR2, and vhCDR3; ii) a constrained non-cleavable linker (CNCL); and iii) a variable light domain comprising vlCDR1, vlCDR2, and vlCDR3; d) a second domain linker; e) a second sdABD -TTA; f) a cleavable linker (CL); g) a constrained pseudo-Fv domain, the constrained pseudo-Fv domain comprising: i) a pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a pseudo heavy variable domain; h) a third domain linker; and i) a third sdABD that binds to human serum albumin; wherein the variable heavy domain and the variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind CD3; the variable heavy domain and the pseudo variable light domain associate intramolecularly to form an inactive Fv; and the variable light domain and the pseudo variable heavy domain associate intramolecularly to form an inactive Fv. In some embodiments, the human tumor target antigen is B7H3.

[0007] In another aspect, the present invention provides a protein comprising, from N-terminus to C-terminus: a first single domain antigen binding domain (sdABD) (sdABD-TTA) that binds to a human tumor target antigen (TTA), the first single domain antigen binding domain comprising sdFR1-sdCDR1-sdFR2-sdCDR2-sdFR3-sdCDR3-sdFR4; b) a first domain linker; c) a constrained Fv domain comprising: i) comprising vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCDR3-vhFR4 variable heavy domain; ii) a constrained non-cleavable linker (CNCL); and iii) a variable light domain comprising vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4; d) a second domain linker; e) a second sdABD-TTA; f) a cleavable linker (CL); g) a constrained pseudo-Fv domain, The constrained pseudo-Fv domain comprises: i) a pseudo light variable domain comprising sdFR1-sdCDR1-sdFR2-sdCDR2-sdFR3-sdCDR3-sdFR4; ii) a non-cleavable linker (NCL); and iii) a pseudo heavy variable domain comprising vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4; h) a third domain linker; and i) binds to human blood. A third sdABD of albumin, the third sdABD comprising sdFR1-sdCDR1-sdFR2-sdCDR2-sdFR3-sdCDR3-sdFR4; wherein the variable heavy domain and the variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind CD3; the variable heavy domain and the pseudo variable light domain associate intramolecularly to form an inactive Fv; and the variable light domain and the pseudo variable heavy domain associate intramolecularly to form an inactive Fv. In some embodiments, the human tumor target antigen is B7H3.

[0008] In some embodiments of form 2 protein, the variable heavy domain is located at the N-terminus of the variable light domain, and the pseudo light variable domain is located at the N-terminus of the pseudo variable heavy domain. In some embodiments, the variable heavy domain is located at the N-terminus of the variable light domain, and the pseudo variable light domain is located at the C-terminus of the pseudo variable heavy domain. In some embodiments, the variable heavy domain is located at the C-terminus of the variable light domain, and the pseudo variable light domain is located at the N-terminus of the pseudo variable heavy domain. In some embodiments, the variable heavy domain is located at the C-terminus of the variable light domain, and the pseudo variable light domain is located at the N-terminus of the pseudo variable heavy domain. In some embodiments, the variable heavy domain is located at the C-terminus of the variable light domain, and the pseudo variable light domain is located at the C-terminus of the pseudo variable heavy domain.

[0009] In some embodiments of Form 2 protein, the first sdABDTTA and the second sdABDTTA are the same. In some embodiments, the first sdABDTTA and the second sdABDTTA are different. In these embodiments, the sdABD-TTA is selected from those depicted in Figure 5, including SEQ ID NO: 1, SEQ ID NO: 5, SEQ ID NO: 9, SEQ ID NO: 13, SEQ ID NO: 17, SEQ ID NO: 21, SEQ ID NO: 25, SEQ ID NO: 29, SEQ ID NO: 33, SEQ ID NO: 37, SEQ ID NO: 41, SEQ ID NO: 45, SEQ ID NO: 49, SEQ ID NO: 53, SEQ ID NO: 57, SEQ ID NO: 61, SEQ ID NO: 65, SEQ ID NO: 69, SEQ ID NO: 73, 77, SEQ ID NO: 81, SEQ ID NO: 85, SEQ ID NO: 89, SEQ ID NO: 93, SEQ ID NO: 97, SEQ ID NO: 101, SEQ ID NO: 105, SEQ ID NO: 109 and SEQ ID NO: 113.

[0010] In some embodiments of the Form 2 protein, the pseudo heavy variable domain of the constrained pseudo Fv domain is selected from SEQ ID NO: 146 (V Hi ), SEQ ID NO: 150 (V Hi2 ) and SEQ ID NO: 154 (VHiGL4), as shown in Figure 5. In some embodiments, the pseudo light variable domain of the constrained pseudo Fv domain is selected from SEQ ID NO: 130 (V Li )、SEQ ID NO:134(V Li2) and SEQ ID NO: 138 (V LiGL ) group, as shown in Figure 5.

[0011] In another aspect, the invention provides a "Format 1" protein comprising, from N-terminus to C-terminus: a) a first sdABD-TTA; b) a first domain linker; c) a constrained Fv domain comprising: i) a first variable heavy domain comprising vhCDR1, vhCDR2, and vhCDR3; ii) a constrained cleavable linker (CCL); and iii) a first variable light domain comprising vlCDR1, vlCDR2, and vlCDR3; d) a second domain linker; e) a second sdABD-TTA; f) a cleavable linker (CL); g) a constrained pseudo Fv domain, the constrained pseudo-Fv domain comprises: i) a first pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a first pseudo heavy variable domain; h) a third domain linker; and i) a third sdABD that binds to human serum albumin; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3; wherein the first variable heavy domain and the first pseudo variable light domain associate intramolecularly to form an inactive Fv; and wherein the first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv. In another aspect, the invention provides a "Format 4" protein comprising, from N-terminus to C-terminus: a) a first single domain antigen binding domain (sdABD) that binds to a human tumor target antigen (TTA) (sdABD-TTA); b) a first domain linker; c) a constrained Fv domain comprising: i) a first variable heavy domain comprising vhCDR1, vhCDR2, and vhCDR3; ii) a constrained non-cleavable linker (CNCL); and iii) a first variable light domain comprising vlCDR1, vlCDR2, and vlCDR3; d) a cleavable linker (CLCL); ); e) a second sdABD that binds to human serum albumin; f) a domain linker; g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising: i) a first pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a first pseudo heavy variable domain; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind CD3; wherein the first variable heavy domain and the first pseudo variable light domain associate intramolecularly to form an inactive Fv; and the first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv.

[0012] In another aspect of the Form 1, Form 2, and Form 4 proteins listed above, the first variable heavy domain is located N-terminal to the first variable light domain, and the pseudo light variable domain is located N-terminal to the pseudo variable heavy domain.

[0013] In another aspect of the Form 1, Form 2, and Form 4 proteins listed above, the first variable heavy domain is located N-terminal to the first variable light domain, and the pseudo variable heavy domain is located N-terminal to the pseudo variable light domain.

[0014] In another aspect of the Form 1, Form 2 and Form 4 proteins listed above, the first variable light domain is located N-terminal to the first variable heavy domain, and the pseudo light variable domain is located N-terminal to the pseudo variable heavy domain.

[0015] In another aspect of the Form 1, Form 2 and Form 4 proteins listed above, the first variable light domain is located N-terminal to the first variable heavy domain, and the pseudo variable heavy domain is located N-terminal to the pseudo variable light domain.

[0016] In another aspect, the invention provides Version 1 and Version 2 proteins, wherein the first and second TTAs are identical.

[0017] In another aspect, the invention provides Version 1 and Version 2 proteins, wherein the first and second TTAs are different.

[0018] In another aspect, the present invention provides Form 1, Form 2, and Form 4 proteins, wherein the first and second TTAs are selected from EGFR, EpCAM, FOLR1, Trop2, ca9, and B7H3. These sequences can be selected from the group consisting of: SEQ ID NO: 1, SEQ ID NO: 5, SEQ ID NO: 9, SEQ ID NO: 13, SEQ ID NO: 17, SEQ ID NO: 21, SEQ ID NO: 25, SEQ ID NO: 29, SEQ ID NO: 33, SEQ ID NO: 37, SEQ ID NO: 41, SEQ ID NO: 45, SEQ ID NO: 49, SEQ ID NO: 53, SEQ ID NO: 57, SEQ ID NO: 61, SEQ ID NO: 65, SEQ ID NO: 69, SEQ ID NO: 73, 77, SEQ ID NO: 81, SEQ ID NO: 85, SEQ ID NO: 89, SEQ ID NO: 93, SEQ ID NO: 97, SEQ ID NO: 101, SEQ ID NO: 105, SEQ ID NO: 109 and SEQ ID NO: 113.

[0019] In another aspect, the present invention provides Version 1, Version 2, and Version 4 proteins, wherein the half-life extension domain has SEQ ID NO: 117 (aHSA (10GE)) and SEQ ID NO: 121 (aHSA with His tag).

[0020] In another aspect, the invention provides Form 1, Form 2, and Form 4 proteins, wherein the cleavable linker is cleaved by a human protease selected from the group consisting of MMP2, MMP9, Meprin A, Meprin B, cathepsin S, cathepsin K, cathepsin L, granzyme B, uPA, Kallekriein 7, matriptase, and thrombin, or other human proteases such as those described in FIG. 6 .

[0021] In another aspect, the present invention provides a protein selected from the group consisting of Pro186, Pro225, Pro226, Pro233, Pro262, Pro311, Pro312, Pro313, Pro356, Pro359, Pro364, Pro388, Pro448, Pro449, Pro450, Pro451, Pro495, Pro246, Pro254, Pro255, Pro256, Pro420, Pro421, Pro432, Pro479, Pro480, Pro187, Pro221, Pro222, Pro223, Pro224, Pro393, Pro394, Pro395, Pro396, Pro429, Pro430, Pro431, Pro601, Pro602, Pro603, Pro604, and V605. 65, Pro667, Pro694, Pro695, Pro565, Pro566, Pro567, Pro727, Pro728, Pro729, Pro730, Pro731, Pro676, Pro677, Pro678, Pro679, Pro808, Pro819, Pro621, Pro622, Pro640, Pro641, Pro642, Pro643, Pro744, Pro746, Pro638, Pro639, Pro396, Pro476, Pro706, Pro709, Pro470, Pro471, Pro551, Pro552, Pro623, Pro624, Pro698, Pro655, Pro656, Pro657, Pro658, Pro516, Pro517, Pro518, and Pro519.

[0022] In another aspect, the present invention provides nucleic acids encoding Form 1, Form 2, or Form 4 proteins as described herein, as well as expression vectors and host cells comprising nucleic acids encoding the proteins.

[0023] In another aspect, the invention provides methods of making the proteins of the invention and methods of treating a patient in need thereof.

[0024] In another aspect, the present invention provides a composition comprising a "Form 3A" prodrug protein pair, the composition comprising: a) a first protein comprising, from N-terminus to C-terminus: i) a first sdABD-TTA; ii) a first domain linker; iii) a pseudo-Fv domain comprising, from N-terminus to C-terminus: 1) a variable heavy chain comprising vhCDR1, vhCDR2, and vhCDR3; 2) a cleavable linker; 3) a first pseudo variable light domain comprising iVLCDR1, iVLCDR2, and iVLCDR3; iv) a second domain linker; v) sdABD-HSA; a) a second protein comprising, from N-terminus to C-terminus: i) a third sdABD that binds to a human tumor target antigen; ii) a third domain linker; linker; iii) a pseudo-Fv domain comprising, from N-terminus to C-terminus: 1) a variable light chain comprising VLCDR1, VLCDR2, and VLCDR3; 2) a cleavable linker; and 3) a first pseudo variable heavy domain comprising iVHCDR1, iVHCDR2, and iVHCDR3; iv) a fourth domain linker; v) an sdABD-HSA; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3 when associated; wherein the first variable heavy domain and the first pseudo variable light domain molecules associate to form an inactive Fv; wherein the first variable light domain and the first pseudo variable heavy domain molecules associate to form an inactive Fv; and wherein the first and third sdABDs are selected from the group consisting of: SEQ ID NO: 1, SEQ ID NO: 5, SEQ ID NO: 9, SEQ ID NO: 13; SEQ ID NO: 17; SEQ ID NO: 21, SEQ ID NO: 25, SEQ ID NO: 29, SEQ ID NO: 33, SEQ ID NO: 37, SEQ ID NO: 41, SEQ ID NO: 45, SEQ ID NO: 49, SEQ ID NO: 53, SEQ ID NO: 57, SEQ ID NO:61, SEQ ID NO:65, SEQ ID NO:69, SEQ ID NO:73,77, SEQ ID NO:81, SEQ ID NO:85, SEQ ID NO:89, SEQ ID NO:93, SEQ ID NO:97, SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109 and SEQ ID NO:113.

[0025] In another aspect, the present invention provides a composition comprising a "Form 3B" prodrug protein pair, the composition comprising a) a first protein comprising, from N-terminus to C-terminus: i) a first sdABD-TTA; ii) a first domain linker; iii) a second sdABD-TTA; iv) a second domain linker; iii) a pseudo-Fv domain comprising, from N-terminus to C-terminus: 1) a variable heavy chain comprising vhCDR1, vhCDR2, and vhCDR3; 2) a cleavable linker; 3) a first pseudo variable light domain comprising iVLCDR1, iVLCDR2, and iVLCDR3; iv) a third domain linker; and v) sdABD-HSA; a) a first second protein comprising, from N-terminus to C-terminus: i) a third sdABD -TTA; ii) a fourth domain linker; iii) a fourth sdABD-TTA; iv) a fifth domain linker; iii) a pseudo-Fv domain, said pseudo-Fv domain comprising, from N-terminus to C-terminus: 1) a variable light chain comprising VLCDR1, VLCDR2 and VLCDR3; 2) a cleavable linker; and 3) a first pseudo variable heavy domain comprising iVHCDR1, iVHCDR2 and iVHCDR3; iv) a sixth domain linker; v) sdABD-HSA; wherein said first variable heavy domain and said first variable light domain are capable of binding to human CD3 when associated; wherein said first variable heavy domain and said first pseudo variable light domain molecules associate to form an inactive Fv; and wherein said first variable light domain and said first pseudo variable heavy domain molecules associate to form an inactive Fv.

[0026] In another aspect, Form 3A and Form 3B proteins have an sdABD-HSA having SEQ ID NO:117 or SEQ ID NO:121.

[0027] In another aspect, the Form 3A and Form 3B proteins have an sdABD-TTA that binds to a TTA selected from EGFR, EpCAM, Trop2, CA9, FOLR1, and B7H3. The sdABD-TTA may be selected from the group consisting of SEQ ID NO: 1, SEQ ID NO: 5, SEQ ID NO: 9, SEQ ID NO: 13, SEQ ID NO: 17, SEQ ID NO: 21, SEQ ID NO: 25, SEQ ID NO: 29, SEQ ID NO: 33, SEQ ID NO: 37, SEQ ID NO: 41, SEQ ID NO: 45, SEQ ID NO: 49, SEQ ID NO: 53, SEQ ID NO: 57, SEQ ID NO: 61, SEQ ID NO: 65, SEQ ID NO: 69, SEQ ID NO: 73, 77, SEQ ID NO: 81, SEQ ID NO: 85, SEQ ID NO: 89, SEQ ID NO: 93, SEQ ID NO: 97, SEQ ID NO: 101, SEQ ID NO: 105, SEQ ID NO: 109 and SEQ ID NO: 113.

[0028] In another aspect, the present invention provides an sdABD that binds to human Trop2, the sdABD having a sequence selected from the group consisting of SEQ ID NO:77, SEQ ID NO:81, SEQ ID NO:85, SEQ ID NO:89, and SEQ ID NO:93.

[0029] In another aspect, the present invention provides an sdABD that binds to human B7H3, the sdABD having a sequence selected from the group consisting of SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53, and SEQ ID NO:57.

[0030] In another aspect, the present invention provides an sdABD that binds to human CA9, the sdABD having a sequence selected from the group consisting of SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109, and SEQ ID NO:113.

[0031] In another aspect, the present invention provides an sdABD that binds to human EpCAM, the sdABD having a sequence selected from the group consisting of SEQ ID NO:69 and SEQ ID NO:73.

[0032] In another aspect, the present invention provides nucleic acid compositions comprising a first nucleic acid encoding a first protein member of a prodrug pair and a second nucleic acid encoding a second protein member of the pair; and expression vectors and host cells containing the nucleic acids.

[0033] The present invention includes the following embodiments:

[0034] 1. A fusion protein comprising, from N-terminus to C-terminus:

[0035] a) first sdABD-TTA;

[0036] b) a first domain linker;

[0037] c) a constrained Fv domain, said constrained Fv domain comprising:

[0038] i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3;

[0039] ii) a constrained non-cleavable linker (CNCL); and

[0040] iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3;

[0041] d) a second domain linker;

[0042] e) second sdABD-TTA;

[0043] f) a cleavable linker (CL);

[0044] g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising:

[0045] i) a first pseudo light variable domain;

[0046] ii) a non-cleavable linker (NCL); and

[0047] iii) a first pseudo heavy variable domain;

[0048] h) a third domain linker; and

[0049] i) a third sdABD that binds to human serum albumin;

[0050] wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3;

[0051] The first variable heavy domain and the first pseudo variable light domain intermolecularly associate to form an inactive Fv;

[0052] The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv; and

[0053] wherein at least one of the sdABD-TTAs is a sdABD-B7H3 having a sequence selected from the group consisting of SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53, and SEQ ID NO:57.

[0054] 2. The fusion protein of embodiment 1, which is Pro664 and has SEQ ID NO: 282.

[0055] 3. A fusion protein comprising, from N-terminus to C-terminus:

[0056] a) a first single domain antigen binding domain (sdABD) that binds to a human tumor target antigen (TTA) (sdABD-TTA);

[0057] b) a first domain linker;

[0058] c) a constrained Fv domain, said constrained Fv domain comprising:

[0059] i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3;

[0060] ii) a constrained non-cleavable linker (CNCL); and

[0061] iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3;

[0062] d) a second domain linker;

[0063] e) second sdABD-TTA;

[0064] f) a cleavable linker (CL);

[0065] g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising:

[0066] i) a first pseudo light variable domain;

[0067] ii) a non-cleavable linker (NCL); and

[0068] iii) a first pseudo heavy variable domain;

[0069] h) a third domain linker; and

[0070] i) a third sdABD that binds to human serum albumin;

[0071] wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3;

[0072] The first variable heavy domain and the first pseudo variable light domain intermolecularly associate to form an inactive Fv;

[0073] The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv; and

[0074] wherein at least one of the sdABD-TTAs is a sdABD-EpCAM having a sequence selected from the group consisting of SEQ ID NO: 69 and SEQ ID NO: 73.

[0075] 4. A fusion protein comprising, from N-terminus to C-terminus:

[0076] a) first sdABD-TTA;

[0077] b) a first domain linker;

[0078] c) a constrained Fv domain, said constrained Fv domain comprising:

[0079] i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3;

[0080] ii) a constrained non-cleavable linker (CNCL); and

[0081] iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3;

[0082] d) a second domain linker;

[0083] e) second sdABD-TTA;

[0084] f) a cleavable linker (CL);

[0085] g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising:

[0086] i) a first pseudo light variable domain;

[0087] ii) a non-cleavable linker (NCL); and

[0088] iii) a first pseudo heavy variable domain;

[0089] h) a third domain linker; and

[0090] i) a third sdABD that binds to human serum albumin;

[0091] wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3;

[0092] The first variable heavy domain and the first pseudo variable light domain intermolecularly associate to form an inactive Fv;

[0093] The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv; and

[0094] wherein at least one of the sdABD-TTAs is sdABD-Trop2 having a sequence selected from the group consisting of SEQ ID NO:77, SEQ ID NO:81, SEQ ID NO:85, SEQ ID NO:89, SEQ ID NO:93, and SEQ ID NO:97.

[0095] 5. A fusion protein comprising, from N-terminus to C-terminus:

[0096] a) first sdABD-TTA;

[0097] b) a first domain linker;

[0098] c) a constrained Fv domain, said constrained Fv domain comprising:

[0099] i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3;

[0100] ii) a constrained non-cleavable linker (CNCL); and

[0101] iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3;

[0102] d) a second domain linker;

[0103] e) second sdABD-TTA;

[0104] f) a cleavable linker (CL);

[0105] g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising:

[0106] i) a first pseudo light variable domain;

[0107] ii) a non-cleavable linker (NCL); and

[0108] iii) a first pseudo heavy variable domain;

[0109] h) a third domain linker; and

[0110] i) a third sdABD that binds to human serum albumin;

[0111] wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3;

[0112] The first variable heavy domain and the first pseudo variable light domain associate intramolecularly to form an inactive Fv; and

[0113] The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv.

[0114] wherein at least one of the sdABD-TTAs is sdABD-CA9 having a sequence selected from the group consisting of SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109, and SEQ ID NO:113.

[0115] 6. A fusion protein according to any one of embodiments 1 and 3 to 5, wherein the first variable heavy domain is located at the N-terminus of the first variable light domain, and the pseudo light variable domain is located at the N-terminus of the pseudo variable heavy domain.

[0116] 7. A fusion protein according to any one of embodiments 1 and 3 to 5, wherein the first variable heavy domain is located at the N-terminus of the first variable light domain, and the pseudo variable heavy domain is located at the N-terminus of the pseudo variable light domain.

[0117] 8. A fusion protein according to any one of embodiments 1 and 3 to 5, wherein the first variable light domain is located at the N-terminus of the first variable heavy domain, and the pseudo light variable domain is located at the N-terminus of the pseudo variable heavy domain.

[0118] 9. A fusion protein according to any one of embodiments 1 and 3 to 5, wherein the first variable light domain is located at the N-terminus of the first variable heavy domain, and the pseudo variable heavy domain is located at the N-terminus of the pseudo variable light domain.

[0119] 10. The fusion protein of any one of embodiments 1 and 3 to 9, wherein the first and second TTAs are identical.

[0120] 11. The fusion protein of any one of embodiments 1 and 3 to 9, wherein the first and second TTAs are different.

[0121] 12. The fusion protein of any one of embodiments 1 and 3 to 11, wherein the half-life extension domain has SEQ ID NO: 117.

[0122] 13. The fusion protein of any one of embodiments 1 and 3 to 12, wherein the fusion protein has a sequence selected from the group consisting of: Pro601 Pro602, V3 and V4, Pro665, Pro666, Pro667, Pro694, Pro695, Pro565, Pro566, Pro567, Pro727-731, Pro676-679, Pro808, Pro819, Pro621, Pro622, Pro640-643, Pro744, Pro746, Pro638, Pro639, Pro396, Pro476, Pro706, Pro709, Pro470, Pro471, Pro551, Pro552, Pro623, Pro624, Pro698, Pro655, Pro656, Pro657, Pro658, Pro516, Pro517, Pro518, and Pro519.

[0123] 14. A nucleic acid encoding the fusion protein according to any one of embodiments 1 to 13.

[0124] 15. An expression vector comprising the nucleic acid of embodiment 14.

[0125] 16. A host cell comprising the expression vector of embodiment 15.

[0126] 17. A method for preparing a fusion protein, the method comprising culturing the host cell of embodiment 16 under conditions wherein the protein is expressed and recovering the fusion protein.

[0127] 18. A method of treating cancer, comprising administering to a patient the protein of any one of embodiments 1 to 13.

[0128] 19. A single domain antigen binding domain that binds to human Trop2 having a sequence selected from the group consisting of SEQ ID NO: 77, SEQ ID NO: 81, SEQ ID NO: 85, SEQ ID NO: 89 and SEQ ID NO: 93.

[0129] 20. A single domain antigen binding domain that binds to human B7H3 having a sequence selected from the group consisting of SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53 and SEQ ID NO:57.

[0130] 21. A single domain antigen binding domain that binds to human CA9, having a sequence selected from the group consisting of SEQ ID NO: 101, SEQ ID NO: 105, SEQ ID NO: 109 and SEQ ID NO: 113.

[0131] 22. A single domain antigen binding domain that binds to human EpCAM, having a sequence selected from SEQ ID NO: 69 and SEQ ID NO: 73. BRIEF DESCRIPTION OF THE DRAWINGS

[0132] Figure 1 The "Form 1" type of protease activation of the present invention is depicted, referred to herein as a "constrained, non-cleavable construct" or "cc construct." In this embodiment, a representative construct is Pro140: two TTAs present in the presence of an ABD (e.g. Figure 1 (As depicted, these are all identical, although as described herein, they can be different.) Upon cleavage, the prodrug construct is split into three component parts, one containing the α-TTA domain of the active VH linked to αCD3 via a domain linker, a second containing the α-TTA domain of the active VL linked to αCD3 via a domain linker, and the "remainder" comprising the half-life extension domain linked to the inactive VH and VL. The two active variable domains are then free to associate to form a functional anti-CD3 binding domain. It should be noted that in the "Format 1" embodiment, the resulting active component is trivalent: there is monovalent binding to CD3 and bivalent binding to TTA, thereby generating a bispecific binding protein, although in some cases such trivalency can be trispecific, with monovalent binding to CD3, monovalent binding to the first TTA, and monovalent binding to the second TTA. Figure 1 Also shown is an anti-human serum albumin (HSA) domain as a half-life extending domain, in many embodiments, as an sdABD as defined herein, although as described herein, this is optional and / or may be replaced by other half-life extending domains; alternatively, the half-life extending domain may also be at the N-terminus or internal to the construct. Figure 1Also included are the VH and VL of the Fv and the iVH and iVL of the pseudo-Fv in a specific order, for example, from N-terminus to C-terminus, VH-linker-VL (and iVL-linker-iVH), although as will be appreciated by those skilled in the art, these can be reversed (VL-linker-VH and iVH-linker-iVL). Alternatively, one of the Fvs can be in one orientation and the other in another, although expression of the protein in the orientation shown here is surprisingly higher than in other orientations.

[0133] Figure 2 The "Format 2" type of protease activation of the present invention is depicted, referred to herein as a "constrained, non-cleavable construct" or "CNCL construct," and sometimes also referred to herein as a "dimerization construct" as discussed herein. These constructs do not isomerize as discussed herein. Upon cleavage, the two prodrug constructs split into four component parts: two half-life extension domains (in this case, sdABD to HSA) linked to two pseudodomains (which may or may not be able to self-associate, depending on the length of the linker and the inactivating mutation), and two active moieties that self-assemble into a dimeric active moiety comprising four anti-TTA domains (which may all be identical or two identical and the other two different). It should be noted that in the "Format 2" embodiment, the resulting active component is hexavalent: there is bivalent binding to CD3 and tetravalent binding to TTA, thereby generating a bispecific binding protein, although in some cases this hexavalent can be trispecific, with bivalent binding to CD3, bivalent binding to the first TTA, and bivalent binding to the second TTA. Figure 2 Also shown is an anti-human serum albumin (HSA) domain as a half-life extending domain, in many embodiments, as an sdABD as defined herein, although as described herein, this is optional and / or may be replaced by other half-life extending domains; alternatively, the half-life extending domain may also be at the N-terminus or internal to the construct. Figure 2 Also included are the VH and VL of the Fv and the iVH and iVL of the pseudo-Fv in a specific order, for example, from N-terminus to C-terminus, VH-linker-VL (and iVL-linker-iVH), although as will be appreciated by those skilled in the art, these can be reversed (VL-linker-VH and iVH-linker-iVL). Alternatively, one of the Fvs can be in one orientation and the other in another, although expression of the protein in the orientation shown here is surprisingly higher than in other orientations.

[0134] Figure 3A-Figure 3B Depicts a "Format 3" type of construct as outlined herein, sometimes also referred to as a "half construct" or "half-COBRA TM", as these are two different polypeptide chains that together constitute the MCE therapeutic, as discussed further herein. In this embodiment, the constructs are delivered in pairs with pre-cleavage intramolecular self-assembly, resulting in an inactive anti-CD3 Fv domain. After cleavage, the inert variable domain is released, and the two active variable domains then assemble intermolecularly to form the active anti-CD3 binding domain. The two sdABD-TTAs bind to corresponding receptors on the surface of tumor cells, and the cleavage is performed by a protease. This allows intermolecular assembly because the molecules are physically held in place, thereby facilitating the assembly of the active anti-CD3 domains. As above for versions 1 and 2, in this embodiment, the N-terminal to C-terminal order of the variable domains can also be reversed or mixed. In addition, the sdABD (HSA) can be located at the N-terminus or C-terminus of each half-construct. Pro16 has the sdABD (HSA) at the C-terminus, and Pro17 has it at the N-terminus (see Pro19, SEQ ID NO: XX, which has the sdABD (HSA) at the C-terminus). Figure 3A Form 3 constructs are shown with each half-construct having a single sdABD-TTA domain, and Figure 3B Format 3 constructs are shown with two sdABD-TTA per half construct in a "dual targeting" or "heterogeneous targeting" format. Note that Figure 3B FOLR1 and EGFR were used as the two TTAs, but other combinations as outlined herein may also be used.

[0135] Figure 4 A "Format 4" type construct is depicted, similar to the "Format 2" construct, but with only a single sdABD-TTA. The figure shows an sdABD-TTA directed against EGFR, but as will be appreciated by those skilled in the art, other TTAs may also be used. Upon cleavage, the prodrug construct splits into its two components: a half-life extension domain (in this case, an sdABD directed against HSA) linked to a pseudo-Fv, and an active moiety, which self-assembles into a dimeric active moiety containing two anti-TTA domains in the presence of a second active moiety from a different cleavage molecule. It should be noted that in the "Format 4" embodiment, the resulting active moiety is tetravalent: there is bivalent binding to CD3 and bivalent binding to TTA, thereby generating a bispecific binding protein. Figure 4 Also shown is an anti-human serum albumin (HSA) domain as a half-life extension domain, in many embodiments, as sdABD (1 / 2) as defined herein, although as described herein, this is optional and / or can be replaced by other half-life extension domains; in addition, the half-life extension domain can also be at the N-terminus or internal to the construct. Figure 4Also included are the VH and VL of the Fv and the iVH and iVL of the pseudo-Fv in a specific order, for example, from N-terminus to C-terminus, VH-linker-VL (and iVL-linker-iVH), although as will be appreciated by those skilled in the art, these can be reversed (VL-linker-VH and iVH-linker-iVL). Alternatively, one of the Fvs can be in one orientation and the other in another, although expression of the protein in the orientation shown here is surprisingly higher than in other orientations.

[0136] Figure 5A – Figure 5J Depict multiple sequences of the present invention. For antigen binding domains, the CDRs are underlined. As more fully outlined herein, in the present invention these domains can be assembled in a wide range of configurations, including "Format 1," "Format 2," "Format 3," and "Format 4" orientations.

[0137] Figure 6A – Figure 6D Many applicable protease cleavage sites are described. As will be appreciated by those skilled in the art, these cleavage sites can be used as cleavable joints. In some embodiments, for example, when a more flexible cleavable joint is desired, other amino acids (normally glycine and serine) may be present in the N-terminal or C-terminal or both of these cleavage sites.

[0138] Figures 7A-7D Depicted with "Form 3" or "Semi-COBRA TM This demonstrates that the Version 3 constructs bind cooperatively to CD3 and generate CD3 binding sites upon cleavage by a protease (in this case, EK protease, but any of the protease cleavage sites outlined herein and depicted in Figures 5 and 6 may be used), as shown by sandwich FACS analysis.

[0139] Figures 8A-8D Shown, protease cleavage by complementary half-COBRA TM T cell killing of synergistically activated EGFR+ target cells. Figure 8A and Figure 8B It was shown that constructs isolated but cleaved with different concentrations of protease did not affect target cell viability. Figure 8C It was shown that in the presence of the proteases in combination, target cell viability was significantly reduced. Figure 8D Shows the general mechanism.

[0140] Figure 9 Some non-target controls used to determine the efficacy of the test format 1 constructs are shown.

[0141] Figures 10A-10FIt was shown that the generation of an active CD3 binding domain is dependent on target binding by two "arms" (eg, sdABD-TTA domains), one of which is present on each of the two constructs. ELISA assays were performed as described in the Examples.

[0142] Figure 11 Suitable semi-COBRA is shown TM "Mep" represents a meprin protease cleavage site, "His-6" is a tag as discussed more fully herein, ST14 is a matriptase protease cleavage site, and "Thb" is a thrombin protease cleavage site.

[0143] Figures 12A-12C Shown with Figure 11 TDCC data related to the constructs. Figure 12A It is shown that the addition of pre-cleaved half-COBRA pairs results in efficacy against OvCAR8 cells. Figure 12B It was shown that the addition of pre-cleaved half-COBRA resulted in efficacy against HCT116 cells, and Figure 12C It was shown that the addition of pre-cleaved half-COBRA pairs resulted in efficacy against LoVo cells, all cancer cell lines.

[0144] Figure 13A – Figure 13B It is shown that the MMP9 linker is stable in vivo. Pro40 (MMP9 cleavable) and Pro74 (non-cleavable) were administered as a single intravenous bolus dose to NSG mice at a dose of 0.5 mg / kg via the tail vein. A dose solution of each compound was prepared in a vehicle of 25 mM citric acid, 75 mM L-arginine, 75 mM NaCl and 4% sucrose pH 7.0. Two blood samples were collected from each animal at preselected times, one at the beginning of the study by orbital bleeding or submandibular bleeding, and the other at the endpoint by cardiac puncture. The time points for blood collection were 0.083, 1, 6, 24, 72 and 168 hours. Plasma was prepared from each individual blood sample using K2 EDTA tubes. Concentrations were determined using an MSD assay using a monoclonal antibody specific for HSA sdABD and detected using the EGFR extracellular domain.

[0145] Figure 14 Depicted in Figure 15 The experiments depicted in this paper used form 3A hemi-COBRA. TMSchematic diagram of the constructs. Pro51 is a positive control because it is "always on," so it forms an active anti-CD3 Fv. Pro98 is a negative control because its sdABD targets hen egg lysozyme, which is not expressed by tumors. Pro77 and Pro53 are a prodrug Format 3A pair, using an sdABD targeting EGFR and an MMP9 cleavage site. Pro74 and Pro72 are a negative control Format 3A pair because they lack a cleavage site.

[0146] Figure 15 It was shown that using the protocol in the Examples, the Format 1 construct regressed tumors in vivo using two different tumor cell lines implanted in mice. The anti-tumor activity of the half-COBRA construct (Pro77 and Pro53) was dependent on the inclusion of an anti-EGFR sdABD and an MMP9 cleavable linker as well as an active anti-CD3 Fv.

[0147] Figure 16 A schematic diagram of the next generation format is shown, i.e., a full-length construct with two pseudo-Fv domains with a cleavable site between them, as generally described in US2018 / 0134789, which is hereby incorporated by reference. However, as shown in the figure below, this first generation full-length construct did not show very good conditionality because it could isomerize to form active and inactive constructs.

[0148] Figure 17 It was shown that the version 3A construct pair actually displayed better conditionality than the Pro100 first generation full length construct.

[0149] Figure 18 Depicts additional first generation full-length constructs that are Figure 19 Tested in .

[0150] Figure 19 It was shown that the first generation constructs displayed high activity even in the uncleaved form, ie poor conditionality.

[0151] Figure 20 The first generation full length construct is shown to display two monomeric peaks on analytical SEC.

[0152] Figure 21 Schematic diagram showing the reason for the non-lytic activity, which is that the full-length first-generation construct isomerizes to form two conformations, one of which is inactive because no active anti-CD3 Fv is formed ("bivalent scFv"), and the other of which is active in the absence of proteases, i.e., a "single-chain diabody" type of configuration. See PEDS 23(8):667-677 (2010).

[0153] Figure 22Shown are the results of a TDCC assay run at 37°C for 2 days using a first generation single chain construct. The results indicate that the uncleaved construct exhibited robust killing. These results led to the generation of a version 1 construct.

[0154] Figures 23A-23G Format 1 constructs used in the present invention are shown. As will be appreciated by those skilled in the art and as described herein, these are depicted as having sdABD-EGFR targeting moieties, but sdABDs directed to other TTAs may be used.

[0155] Figure 24 The Version 1 construct (in this case Pro140) is shown to form a single isomer that is stable at 37°C.

[0156] Figure 25 Depicts very low binding of the Format 1 construct to human CD3 in uncleaved form as measured by Octet assay. The top line is Pro120, the middle line is Pro51 (positive control), and the bottom line is Pro140 at 4°C or 37°C for 3 days.

[0157] Figure 26 Similarly depicted, the Version 1 construct has very low TDCC activity in the uncleaved form.

[0158] Figure 27 Depicted is the specific version 1 construct Pro140 used in in vivo testing, using sdABD-EGFR and the MMP9 cleavage site as targeting moieties.

[0159] Figure 28A – Figure 28B Tumor regression using Format 1 construct is shown.

[0160] Figure 29 Due to the cleavage site in the constrained Fv, several different fragments can be generated: partially cleaved fragments and fully cleaved fragments. Surprisingly, the partially cleaved form is more active than the fully cleaved form, resulting in the generation of form 2.

[0161] Figure 30 A number of Form 2 schematics are shown, all using the sdABD-EGFR targeting domain, but sdABDs for other TTAs can be used as described herein and listed in the sequences. Pro51 and Pro201 are positive controls (in the active "half" and active dimer configurations, respectively), and Pro214 is a full-length negative control because there is no cleavage site.

[0162] Figure 31TDCC activity of Pro187, a Format 2 construct, using a meprin cleavage site is shown. The activity of Pro187 in the TDCC assay was 1200-fold higher when pre-cleaved than when uncleaved. Pre-cleaved Pro187 exhibited an activity intermediate between the positive controls Pro51 and Pro201. Uncleaved Pro187 exhibited activity similar to Pro214, which does not contain a protease-cleavable linker.

[0163] Figure 32 TDCC activity of Pro186, a construct of Version 2, using n MMP9 cleavage sites, is shown. Pro186 exhibited 18-fold greater activity in the TDCC assay when pre-cleaved compared to when uncleaved. Pre-cleaved Pro186 exhibited activity intermediate between the positive controls Pro51 and Pro201. Uncleaved Pro186 exhibited higher activity than Pro214, which does not contain a protease-cleavable linker, likely due to MMP activity produced by cells during the 48-hour assay.

[0164] Figure 33 The Pro186 construct is depicted to bind to cells with varying levels of EGFR receptors, but not to CHO cells that do not express EGFR on the cell surface. Pro186 saturates cells expressing varying levels of EGFR at similar COBRA concentrations.

[0165] Figure 34 Shown in Figure 35 Schematic representation of the version 2 constructs used in in vivo studies, all of which employed the sdABD-EGFR targeting domain.

[0166] Figure 35 It was shown that the Form 2 construct Pro186 was highly effective at both dose levels and outperformed the Form 1 construct Pro140 at the lower dose.

[0167] Figure 36 Many version 2 constructs based on Pro186 but with different protease cleavage sites are depicted. Although all of these constructs use sdABD-EGFR for both targeting domains, other sdABDs targeting different TTAs can also be used, and they can be the same or different. That is, either homogeneous targeting (sdABDs targeting the same TTA) or heterogeneous targeting (one sdABD targeting a first TTA and another sdABD targeting a different TTA) can be performed.

[0168] Figure 37Schematic diagrams depicting different Format 2 constructs with varying linker lengths between the Fv domains. These are shown using the MMP9 cleavage site, but other cleavage sites can also be used as described herein. Similarly, while all of these constructs use sdABD-EGFR for both targeting domains, other sdABDs for different TTAs can also be used, and they can be the same or different.

[0169] Figure 38 It is shown that the linker length of the pseudo-Fv can be varied, for example, a construct with a short linker between active Fvs ("short active") and a longer linker between pseudo-Fvs ("long inactive") exhibited similar activity to "short active" and "short inactive." Thus, the conditionality of the COBRA construct is independent of both the active and inactive scFv linkers being constrained; as long as one of them is constrained, the single-chain diabody fold appears to be more favorable than the bivalent scFv fold.

[0170] Figure 39 The length of the active Fv linker can be varied; for example, a version 2 construct with a "long active" and a "short inactive" structure behaves similarly to a "short active" and a "short inactive" structure. Thus, the conditionality of the COBRA construct is independent of both the active and inactive scFv linkers being constrained; as long as one of them is constrained, the single-chain diabody fold appears to be more favorable than the bivalent scFv fold.

[0171] Figure 40A – Figure 40C Schematic diagrams of many different constructs are shown. Pro188 is the construct of form 1, similar to Pro140 but with long joints (16 aggressiveness) in pseudo-Fv. Pro189 and Pro190 (form 2 constructs) are similar to Pro186 and Pro187, except having long joints (16 aggressiveness) in pseudo-Fv domains. Pro191 and Pro192 (also form 2 constructs) are similar to Pro189 and Pro190, except that they have extra cleavage sites in sdABD (1 / 2) upstream. Pro193 (form 4) has a single EGFR targeting domain, is rearranged to iVH and iVL in reverse order and in extra cleavage sites in sdABD (1 / 2) upstream. Pro195 is a form 2 construct similar to Pro186, which has a targeting domain that is attached to same TTA (i.e. EGFR) but is attached to different epi-positions. Pro196, Pro197 and Pro198 are form 2 constructs with rearranged variable domains.

[0172] Figure 41Depicts the fact that different sdABD clones against human FOLR1 show different killing.A Pro22-type construct (with a FLAG sequence instead of Pro51 of NCL) binding to human FOLR1 was compared with a Pro22-EGFR construct against a panel of cell lines.

[0173] Figure 42 Schematic diagram depicting four sdABD-FOLR1 constructs, including an active domain dimer utilizing Pro201 with sdABD-EGFR2 as a positive control (two molecules associate intermolecularly to form two active Fvs against CD3), as well as two format 2 test articles utilizing Pro311 with 77.2 sdABD and Pro312 with h59.3 sdABD, and two negative controls utilizing Pro299 with h77.2 sdABD and Pro303 with h59.3 sdABD.

[0174] Figure 43 Schematic diagram depicting the Format 2 construct used for FOLR / MMP9 in vivo design.

[0175] Figure 44 The efficacy of the Pro312 construct in vivo was shown and demonstrated that the MMP9 cleavable linker is necessary for anti-tumor activity.

[0176] Figure 45 Schematic diagram depicting some formats using sdABD against human B7H3 (sdABD-B7H3), including Pro244, a positive control (using sdABD-B7H3 (hF7), and two Format 2 testers Pro225 (Format 2 construct) and Pro295 (negative control lacking the cleavage site).

[0177] Figure 46 It is shown that Pro225 has greater conditionality compared to the control Pro295.

[0178] Figure 47 It is shown that Pro373 of the Version 2 construct using the meprin linker displays a great deal of conditionality compared to Pro295.

[0179] Figure 48 A number of sdABD-B7H3 (using hF12 sequence) constructs are depicted showing a Pro51 positive control using sdABD-EGFR, a Pro244 positive control using sdABD-hF12 B7H3, a test construct Pro226, and a negative control Pro296 containing no cleavage site.

[0180] Figure 49Showing good conditionality of the Pro226 construct in the TDCC assay.

[0181] Figure 50 Humanization of the sdABD against human EpCAM is shown.

[0182] Figure 51 Schematic diagram showing various formats: Pro244 is a standard T cell engager positive control, and Pro205 is an active domain dimer positive control, Pro199 is a format 2 construct, and Pro175 is a negative control.

[0183] Figure 52 TDCC activity of the sdABD-EpCAM construct is shown, indicating good conditionality.

[0184] Figure 53A – Figure 53B The TDCC activity of the sdABD-EpCAM Pro199 construct was shown to be well-conditional in both HT29 and LoVo cell models.

[0185] Figure 54A – Figure 54B The TDCC activity of the sdABD-EpCAM Pro200 construct was shown to be well-conditional in both HT29 and LoVo cell models.

[0186] Figure 55 Schematic diagram of Pro255 is shown, which uses two different sdABD-TTAs, one for EGFR (sdABD-EGFR) and the other for EpCAM (sdABD-EpCAM), compared to Pro199 which uses a dual EpCAM sdABD. These are sometimes referred to herein as "heterogeneous targeting" constructs, in this case, Format 2 constructs.

[0187] Figure 56 It was shown that the Pro255 dual targeting molecule with MMP9 cleavage site showed good conditionality.

[0188] Figures 57A-57D Results from experiments on three different cell types are shown. First, Raji transfectants with similar expression levels of EpCAM, EGFR, and EpCAM+EGFR were generated (data not shown). Pro255 targeting both EpCAM and EGFR was then tested in the TDCC assay using each cell type. Figure 57A The parental Raji line, which does not express either receptor, is shown. Figure 57B Conditionality shown on EpCAM lines. Figure 57C Conditionality is shown on the EGRF line. Figure 57D Conditionality shown on EpCAM / EGFR line.

[0189] Figure 58 Schematic diagram depicting Version 4 construct Pro258.

[0190] Figure 59A – Figure 59B Pro258 is shown to be conditional in both FBS and human serum. The conditionality of the MMP9 linker was underestimated due to MMP9 activity in culture. Interestingly, Pro51 TDCC activity was inhibited by HSA binding, while Pro258 TDCC activity was similar to Pro51 in the presence of HSA. Finally, Pro258 conditionality was enhanced 6-fold in the presence of HSA.

[0191] Figures 60A-60C . Figure 60A Cleavage of MMP9 substrates by other MMPs is shown. Figure 60B and 60C Cleavage of a FRET probe containing an MMP9 linker sequence is shown.

[0192] Figure 61A-Figure 61B Some exemplary constructs and their formats are shown.

[0193] Figures 62A-62V Many sequences of the present invention are shown, but many others are found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (where cleavable linkers are italicized and double underlined), and domain separations are indicated with " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and as purification tags.

[0194] Figures 63A to 63EE Depicted are the amino acid sequences of exemplary Format 2 constructs comprising multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A The amino acid sequences of Pro601 and Pro602 are depicted. Figure 63B Depict the amino acid sequences of V3 and V4. Pro601 comprises two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 comprises two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 comprises two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 comprises two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb).

[0195] Figures 64A-64C The COBRA design and predicted folding mechanism are shown. Figure 64A Describing Pro186 COBRA( Figure 62B Schematic diagram of SEQ ID NO: 145). Figure 64B The predicted COBRA fold is shown. Figure 64C Shown is the analytical size exclusion chromatogram of Pro186.

[0196] Figures 65A-65C Depicted are exemplary embodiments of the constructs described herein, including Pro186 (pre-cleaved), Pro186 cleavage products, and PRO186 active dimers.

[0197] Figure 66 Provides an illustration of the conversion of COBRA to the active dimer following protease cleavage.

[0198] Figure 67A-Figure 67B Characterization of COBRA binding is provided. Figure 67A Binding activity is shown for human, cynomolgus monkey and mouse preparations. Figure 67B Shown are PRO186 binding to human CD3ε; active PRO186 binding to human CD3ε and active PRO186 binding to human EGFR.

[0199] Figures 68A-68B Cleavage of the PRO186 linker by MMP2 and MMP9 is shown. Figure 68A Western blot depicting active binding product molecules after cleavage. Figure 68B The accumulation of active binding product molecules is shown relative to the cleavage time.

[0200] Figure 69 In vitro activity of the conditional PRO186 construct is shown. Figure 69 - Left panel shows the results of T cell killing assay. Figure 69 - The right panel shows the level of IFN-γ release in relation to the concentration of the test article.

[0201] Figure 70 Shown are EGFR expression relative to activity in three tumor cell lines - LoVo (colorectal cancer (CRC) cell line), HT-29 (colorectal cancer (CRC) cell line), and SCC25 (head and neck cancer cell line).

[0202] Figure 71A and Figure 71B EGFR, MMP2, and MMP9 expression on tumor cells and tumor xenografts is shown. Figure 71A Shown are the cell surface densities of EGFR on three cancer cell lines - LoVo, HT-29 and SCC25. Figure 71BImmunohistochemical staining of tumor xenografts for EGFR, MMP2, and MMP9 is shown.

[0203] Figure 72 Provides a schematic diagram of the experimental procedures for the adoptive human T cell transfer model in tumor-bearing mice.

[0204] Figure 73 PRO186 is shown to cause regression of established solid tumors in mice. Figure 73 - Left panel shows regression of LoVo derived tumors. Figure 73 - Middle panel shows regression of HT-29 derived tumors. Figure 73 - Right panel shows regression of SCC25-derived tumors.

[0205] Figure 74A-74B It was shown that cleaved PRO186 was cleared faster than intact (uncleaved) PRO186. Figure 74A Shown are the pharmacokinetics of the test articles in the plasma of non-tumor-bearing mice. Figure 74B Shown are the tumor volumes of LoVo-derived tumors in mice administered the test articles.

[0206] Figure 75 Depicts the results of a T cell-dependent cytotoxicity (TDCC) assay using Pro233 cleaved by MMP9 and Pro233. Pro233 is dependent on the humanized anti-EGFR binding domain. The results show that cleaved Pro233 exhibits potency on HT29 cells expressing EGFR compared to the uncleaved form.

[0207] Figure 76 Depicts the results of a TDCC assay using Pro565, Pro565 cleaved by MMP9, and Pro568 (uncleavable control). Pro565 is dependent on the hVIB664 anti-EpCAM binding domain. The results show that cleaved Pro565 is more effective than the uncleaved version and the uncleavable control on EpCAM-expressing HT29 tumor cells.

[0208] Figure 77 Depicts the results of a TDCC assay using Pro225, Pro225 cleaved by MMP9, and Pro295 (a non-cleavable control). Pro225 is dependent on the hF7 anti-B7H3 binding domain. The results show that cleaved Pro566 is more effective than the non-cleaved version and the non-cleavable control on HT29 tumor cells expressing B7H3 (note that HT29 expresses EGFR, B7H3, and EpCAM).

[0209] Figure 78Depicts the results of a TDCC assay using Pro566, Pro566 cleaved by MMP9, and Pro569 (uncleavable control). Pro566 is dependent on the hVIB665 anti-EpCAM binding domain. The results show that cleaved Pro566 is more effective than the uncleaved version and the uncleavable control on EpCAM-expressing HT29 tumor cells.

[0210] Figure 79 Depicts the results of a TDCC assay using a dual targeting construct to EGFR and EpCAM (sometimes referred to herein as "iso-COBRA") using the EGFR2 binding domain and the hVIB664 EpCAM binding domain. The results show that MMP9 cleaved Pro623 is more effective than uncleaved Pro623 or the uncleavable control Pro625.

[0211] Figure 80 Depicts the results of a TDCC assay using a dual targeting construct to EGFR and EpCAM using the EGFR2 binding domain and the hVIB665 EpCAM binding domain. The results show that MMP9 cleavage of Pro624 is more effective than uncleaved Pro624 or the uncleavable control Pro626.

[0212] Figure 81 Depicts the results of a TDCC assay using a dual targeting construct for EGFR and EpCAM (in the opposite orientation to Pro624) using the hEGFR2 binding domain and the hVIB665 EpCAM binding domain. The results show that MMP9 cleavage of Pro698 is more effective than uncleaved Pro698 or the uncleavable control Pro699.

[0213] Figure 82 Depicted are the results of a TDCC assay using a dual targeting construct for B7H3 and EpCAM using the hF7 B7H3 binding domain and the hVIB664 EpCAM binding domain in Pro655. The results show that MMP9 cleavage of Pro665 is more effective than uncleaved Pro665 or the uncleavable control Pro659.

[0214] Figure 83 Depicted are the results of a TDCC assay using a dual targeting construct for B7H3 and EpCAM using the hF7 B7H3 binding domain and the hVIB664 EpCAM binding domain in Pro657 (in the opposite orientation to Pro655). The results show that MMP9 cleavage of Pro657 is more effective than uncleaved Pro657 or the uncleavable control Pro661.

[0215] Figure 84Depicted are the results of a TDCC assay using a dual targeting construct for B7H3 and EpCAM using the hF7 B7H3 binding domain and the hVIB665 EpCAM binding domain in Pro656. The results show that MMP9 cleavage of Pro656 is more effective than uncleaved Pro656 or the uncleavable control Pro660.

[0216] Figure 85 Depicted are the results of a TDCC assay using a dual targeting construct for B7H3 and EpCAM using the hF7 B7H3 binding domain and the hVIB665 EpCAM binding domain in Pro658 (in the opposite orientation to Pro656). The results show that MMP9 cleavage of Pro658 is more effective than uncleaved Pro658 or the uncleavable control Pro662.

[0217] Figure 86 The experiments described show that the dual targeting constructs Pro656 and Pro658, which bind to both B7H3 and EpCAM (the two domains are oriented differently), kill all three cell types (those expressing B7H3 and not EpCAM, those expressing EpCAM and not B7H3, and those expressing both). In contrast, Pro225 (which has two anti-B7H3 domains) kills only two cell types (those expressing B7H3 and both B7H3 and EpCAM). Similarly, Pro566 (which has two anti-EpCAM domains) kills only two cell types, those expressing EpCAM and both B7H3 and EpCAM. The Raji F cell line transiently expressing the appropriate proteins was used.

[0218] Figure 87 Depicts the results of a TDCC assay using the B7H3 targeting construct Pro226. The results show that Pro226 and Pro226 cleaved by MMP9 showed higher activity compared to the uncleavable control Pro296. Pre-cleaved Pro226 showed an EC50 of approximately 8 pM.

[0219] Figure 88 Depicted are the results of a TDCC assay using Pro226 expressed in the presence of tunicamycin, which increases the potency of Pro226. Tunicamycin prevents glycosylation, and the EC50 of the cleavage product increases to 1 pM, indicating that glycosylation is decreasing the EC50 (increasing potency).

[0220] Figure 89Depicts the results of a TDCC assay using Pro664, which contains an anti-B7H3 hF12 domain with an amino acid variant (S59Y) to remove a glycosylation site, compared to Pro226 (same construct but without the amino acid variant). Pro664 exhibited similar activity to Pro226.

[0221] Figure 90 Depicts the results of a TDCC assay using Pro665, which contains an anti-B7H3 hF12 domain with an amino acid variant (N57Q) to remove a glycosylation site, compared to Pro226 (same construct but without the amino acid variant). Pro665 showed higher potency than Pro226.

[0222] Figure 91 Depicts the results of a TDCC assay using Pro667, which contains an anti-B7H3 hF12 domain with an amino acid variant (N57E) to remove a glycosylation site, compared to Pro226 (same construct but without the amino acid variant). Pro667 exhibited lower potency than Pro226.

[0223] Figure 92 Depicts the results of a TDCC assay using Pro694, which contains an anti-B7H3 hF12 domain with an amino acid variant (S59A) to remove a glycosylation site, compared to Pro226 (same construct but without the amino acid variant). Pro694 showed higher potency than Pro226.

[0224] Figure 93 Depicts the results of a TDCC assay using Pro695, which contains an anti-B7H3 hF12 domain with an amino acid variant (N57D) to remove a glycosylation site, compared to Pro226 (same construct but without the amino acid variant). Pro695 showed similar potency to Pro226.

[0225] Figure 94 Shown is a TDCC assay comparing two different inactivations in the inactive domain. Pro186 has a signature inactivation in both Vhi and Vli, and Pro476 has an i2 inactivation, and they show similar potency.

[0226] Figure 95 Shown is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro393 containing the S9 cleavage site; the results indicate that both constructs behave similarly with different protease cleavage sites.

[0227] Figure 96Depicted is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro394 containing the ST14 MV cleavage site; results show that both constructs behave similarly with different protease cleavage sites.

[0228] Figure 97 Depicted is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro395 containing the CathS cleavage site; the results show that both constructs behave similarly with different protease cleavage sites.

[0229] Figure 98 Depicted is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro396 containing the MMP9v cleavage site; the results show that the two constructs behave similarly with different protease linker sequences. Note that the MMP9 site is cleaved by both MMP9 and CathS, while MMP9v is MMP9 specific.

[0230] Figure 99 Depicted is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro429 containing the MepGzb cleavage site; the results show that both constructs behave similarly with different protease cleavage sites.

[0231] Figure 100 Depicted is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro430 containing the MMP9-2 cleavage site; despite the differences in the MMP9 cleavage sites, they behave similarly.

[0232] Figure 101 Depicted is a TDCC assay comparing Pro186 containing the MMP9 cleavage site and Pro431 containing the ST14 MS cleavage site; the results show that both constructs behave similarly with different protease cleavage sites.

[0233] Figure 102 Depicts a TDCC assay using Pro676 (a Trop2-containing construct) showing that Pro677 and its active dimer (Pro684AD) exhibited higher activity against BXPC3 tumor cells compared to Pro680 (a non-cleavable control). The active dimer was made by expressing the self-dimerizing active domain.

[0234] Figure 103 Depicted is a TDCC assay using Pro677, a Trop2-containing construct, showing that Pro677 and its active dimer (Pro685AD) displayed higher activity against BXPC3 tumor cells compared to Pro681, a non-cleavable control.

[0235] Figure 104 Depicts a TDCC assay using Pro677 (a Trop2-containing construct) showing that Pro677 and its active dimer (Pro685AD) exhibited higher activity against HT29 tumor cells compared to Pro681 (a non-cleavable control). Pro677 exhibited less activation by HT29 cells compared to BXPC3 cells.

[0236] Figure 105 Depicted is a TDCC assay using Pro678, a Trop2-containing construct, showing that Pro678 and its active dimer (Pro686AD) displayed higher activity against BXPC3 tumor cells compared to Pro682, a non-cleavable control.

[0237] Figure 106 Depicted is a TDCC assay using Pro679, a Trop2-containing construct, showing that Pro679 and its active dimer (Pro687 active domain, AD) displayed higher activity against BXPC3 tumor cells compared to Pro683 (uncleavable control).

[0238] Figure 107 Depicted is a TDCC assay using Pro808, a Trop2-containing construct, showing that Pro808, its cleaved form, and the active dimer (Pro810AD) displayed higher activity against BXPC3 tumor cells compared to Pro809, a non-cleavable control.

[0239] Figure 108 Depicts a TDCC assay using Pro808 (a Trop2-containing construct) in HT29 tumor cells. Pro810AD (active domain) showed higher activity in the TDCC assay than Pro809 (uncleavable control). Full-length Pro808 did not show much activation in the assay using HT29 cells.

[0240] Figure 109 Depicted is a TDCC assay using Pro819, a Trop2-containing construct, showing that Pro819 and its active dimer (Pro821AD) displayed higher activity against BXPC3 tumor cells compared to Pro820, a non-cleavable control.

[0241] Figure 110Depicts a TDCC assay using Pro819 (a construct containing Trop2) showing that the active dimer (Pro821AD) exhibits higher activity compared to the uncleavable control. Full-length Pro819 did not show much activation of HT29 cells when compared to the uncleavable control Pro820. Figure 107 and 108 as well as Figure 109 and 110 The data suggest that different target cell types produce different levels of MMP activity when combined with data showing that addition of an MMP9 inhibitor reduces activity (see Figure 133 ).

[0242] Figure 111 Depicted is a TDCC assay using Pro311, a FOLR1 construct, showing that Pro311 cleaved by MMP9 and Pro311 exhibited higher activity against H292 tumor cells compared to a Pro299 uncleavable control.

[0243] Figure 112 Depicted is a TDCC assay using Pro312, a FOLR1 construct, showing that Pro312 cleaved by MMP9 and Pro312 exhibited higher activity against H292 tumor cells compared to a Pro303 uncleavable control.

[0244] Figure 113 Depicts the results of a TDCC assay using the dual targeting construct Pro420 with FOLR1 and EGFR targeting. The results show that Pro420 and Pro420 cleaved by MMP9 showed higher activity compared to the uncleavable control Pro299.

[0245] Figure 114 Depicts the results of a TDCC assay using the dual targeting construct Pro421, which has both FOLR1 and EGFR targeting (in the opposite orientation to Pro420). The results show that Pro421 and Pro421 cleaved by MMP9 exhibited higher activity compared to the uncleavable control Pro299.

[0246] Figure 115 Depicts the results of a TDCC assay using the dual targeting construct Pro551 with EGFR and FOLR1 targeting. The results show that Pro551 and Pro551 cleaved by MMP9 showed higher activity compared to the uncleavable control Pro550.

[0247] Figure 116Depicts the results of a TDCC assay using the dual targeting construct Pro552, which has both FOLR1 and EGFR targeting (in the opposite orientation to Pro551). The results show that Pro522 and Pro522 cleaved by MMP9 displayed higher activity compared to the uncleavable control Pro303.

[0248] Figure 117 Depicted are the results of a TDCC assay using the dual-targeting construct Pro254, in which each targeting domain binds to a different epitope of EGFR and is effective against EGFR-expressing HT29 cells.

[0249] Figure 118 Depicts the results of a TDCC assay using two different dual-targeting constructs that also use two different targeting domains but both bind to B7H3. Pro479 and Pro480 are identical except for the orientation of the two binding domains, and both show activity after cleavage by MMP9 on HT29 cells expressing B7H3.

[0250] Figure 119 Depicted are the results of a TDCC assay using Pro233 and Pro233 cleaved on EGFR-expressing HT-29 cells. Pro233 showed good conditionality and good potency when cleaved by MMP9.

[0251] Figure 120 Depicts the results of a tumor regression study using Pro225, a construct with B7H3 targeting and an MMP9 linker. The study was conducted using a human PBMC engraftment model in which NSG-β2M- / - mice (Jackson) were intravenously implanted with human PBMCs; 3 days after implantation, the mice were subcutaneously implanted with tumor cell lines. Once tumor growth was established, mice were randomized based on tumor volume and given the test article intravenously as indicated. Tumor volume was assessed by caliper measurement. The results demonstrate that Pro225 resulted in regression of established solid tumors compared to the non-cleavable control, Pro295.

[0252] Figure 121 Depicts the results of a tumor regression study using Pro226, a construct with B7H3 targeting and an MMP9 linker. Figure 120 The results were performed as described in , and the results showed that Pro226 caused regression of established solid tumors compared to the uncleavable control Pro295.

[0253] Figure 122Depicts the results of tumor regression studies using Pro565 and Pro566, both of which have EpCAM targeting and an MMP9 linker. The results show that both Pro565 and Pro566 exhibited anti-tumor responses compared to the uncleavable control Pro568.

[0254] Figure 123 Depicts the results of a tumor regression study using Pro393, which utilizes EGFR targeting and an S9 linker, using Figure 122 The results showed that Pro393 caused regression of established solid tumors compared to the uncleavable control Pro214.

[0255] Figure 124 Depicts the results of a tumor regression study using Pro394, which utilizes EGFR targeting and the ST14 MV linker, using Figure 122 The results showed that Pro394 exhibited a small antitumor response compared to the uncleavable control Pro214.

[0256] Figure 125 Depicts the results of a tumor regression study using Pro395, which utilizes EGFR targeting and a CathS linker, using Figure 122 The results showed that Pro395 caused regression of established solid tumors compared to the uncleavable control Pro214.

[0257] Figure 126 Depicts the results of a tumor regression study using Pro396, which utilizes EGFR targeting and an MMP9v linker, using Figure 122 The results showed that Pro396 exhibited an antitumor response compared to the uncleavable control Pro214.

[0258] Figure 127 Depicts the results of a tumor regression study using Pro430, which utilizes EGFR targeting and an MMP9-2 linker, using Figure 122 The results showed that Pro430 caused regression of established solid tumors compared to the uncleavable control Pro214.

[0259] Figure 128 Depicts the results of a tumor regression study using Pro431, which utilizes EGFR targeting and the ST14 MS linker, using Figure 122 The results showed that Pro431 exhibited an antitumor response compared to the uncleavable control Pro214.

[0260] Figure 129 Depicts the results of a tumor regression study using Pro476, which utilizes EGFR targeting, an MMP9-2 linker, and inactive domains Vli2 and VHi2. Figure 122 The results showed that Pro476 caused regression of established solid tumors compared to the uncleavable control Pro214.

[0261] Figure 130 Depicts the results of a tumor regression study using Pro517, which utilizes EGFR targeting and an MMP9-2 linker, using Figure 122 The results showed that Pro517 caused regression of established solid tumors compared to the uncleavable control Pro513.

[0262] Figure 131 Depicts the results of a tumor regression study using Pro664, which utilizes B7H3 targeting and an MMP9 linker, using Figure 120 The results showed that Pro664 caused regression of established solid tumors compared to the uncleavable control Pro766.

[0263] Figure 132 Depicts the results of a tumor regression study using Pro676, which utilizes Trop2 targeting and an MMP9 linker, using Figure 122 The results showed that Pro676 caused regression of established solid tumors compared to the uncleavable control Pro681.

[0264] Figure 133 Depicts the results of studies using Pro225 (anti-B7H3 construct with an MMP9 linker) and increasing amounts of the MMP-specific inhibitor Batimastat, showing that the potency of uncleaved Pro225 is reduced by Batimastat, demonstrating that COBRA is cleaved and activated by cells in the assay. DETAILED DESCRIPTION

[0265] introduction

[0266] The present invention relates to methods for reducing the toxicity and side effects of bispecific antibodies (including antibody-like functional proteins) that bind to important physiological targets such as CD3 and tumor antigens. Many antigen-binding proteins (such as antibodies) can have significant side effects due to targeting normal tissues, and therefore the binding ability of therapeutic molecules needs to be activated only in the vicinity of diseased tissue to avoid normal tissue interactions. Therefore, the present invention relates to multivalent conditionally effective ("MCE") proteins having many functional protein domains. Generally speaking, one of these domains is an antigen binding domain (ABD) that will bind to a target tumor antigen (TTA), and another is an ABD that will bind to a T cell antigen such as CD3 under certain conditions. In addition, the MCE protein also includes one or more protease cleavage sites. That is, the therapeutic molecule is prepared in a "prodrug"-like form, wherein the CD3 binding domain is inactive before exposure to the tumor environment. The tumor environment contains proteases, so that when exposed to the proteases, the prodrug is cleaved and becomes active.

[0267] This is typically achieved by using a protein containing a "pseudo" variable heavy domain and a "pseudo" variable light domain directed against a T cell antigen such as CD3, which constrains the CD3 Fv of the MCE to an inactive form as discussed herein. Since TTA targets the MCE near the tumor, it exposes the MCE to proteases. After cleavage, the active variable heavy domain and the active light domain are now able to pair to form one or more active ABDs directed against CD3 and thus recruit T cells to the tumor for treatment.

[0268] In general, the CD3 binding domain ("Fv") is in a constrained form in which the linker between the active variable heavy domain and the active variable light domain that traditionally form the Fv is too short to allow the two active variable domains to bind to each other; this is referred to as a "constrained linker"; these can be constrained and cleavable (CCL, as used in Form 1) or constrained and non-cleavable (CNCL, as used in Form 2). Instead, in a prodrug (e.g., uncleaved) form, the prodrug polypeptide also comprises a "pseudo-Fv domain." The pseudo-Fv domain comprises a variable heavy chain and light chain domain with a standard framework region but "inert" or "inactive" CDRs. The pseudo-Fv domain also has a constrained linker between the inactive variable heavy chain domain and the inactive variable light chain domain. Because both Fv and pseudo-Fv domains cannot self-assemble due to steric constraints, there is an intramolecular assembly that pairs aVL with iVH and aVH with iVL due to the affinity of their respective framework regions. However, due to the "inert" CDRs of the pseudo-domain, the resulting ABD will not bind CD3, thereby preventing toxicity outside of diseased tissues (such as tumors). However, in the presence of proteases in or near the tumor, the prodrug construct is cleaved, releasing the pseudo-Fv domain from the surface and thus allowing the "real" variable heavy domain and variable light domain to associate intermolecularly (e.g., the two cleaved constructs bind together), thereby triggering active CD3 binding and resulting tumor efficacy. These constructs are generally referred to herein as conditional bispecific redirected activation constructs or "COBRAs." TM The stability of the intramolecular assembly is demonstrated by conditional experiments herein, whereby in the absence of a protease, the uncleaved construct is inactive (e.g., does not form an active CD3 binding domain).

[0269] Interestingly, although these constructs are referred to herein as "constrained" for ease of description, additional work has shown that even if one of the Fv domains is unconstrained, for example, one of the domains can have a longer, flexible linker to facilitate intramolecular assembly. Figures 37-39 As shown in , if only one Fv domain is constrained (domains with active VL and VH or pseudo-Fv domains), intramolecular assembly still occurs (e.g., the uncleaved construct is inactive in the absence of protease cleavage). However, in the current system, the protein has better expression when both linkers are constrained. However, as will be appreciated by those skilled in the art, any of the form 1, form 2, or form 4 constructs herein can have one of these Fv domains with an "unconstrained" or "flexible" linker. For ease of reference, a construct with two Fv domains in a constrained form is shown.

[0270] The constructs and formats of the present invention are variations on the invention described in WO2017 / 156178, which is hereby expressly incorporated by reference in its entirety. Figure 17-21 As shown in , due to the presence of two sets of VH and VL domains in a single polypeptide, the previous construct has the ability to isomerize, thereby forming a bivalent scFv and a single-chain diabody. Even after purification of each isoform, the bivalent construct can still reach equilibrium with the diabody isoform. Due to the ability of the single-chain diabody to bind to CD3 in the absence of protease cleavage, the utility of the construct is reduced.

[0271] To address this issue, the present invention provides four separate types of constructs to accomplish this conditional activation. Prodrug activation can occur in one of four general ways, as generally illustrated in the accompanying figures. Figure 1 , a "Format 1" mechanism is shown. In this embodiment, the prodrug construct has two cleavage sites: one between the VH and VL domains of the constrained Fv, thereby releasing the two variable domains to associate; and a second site that releases the pseudo-Fv domain from the prodrug construct, leaving two molecules that associate due to the innate self-assembly of the variable heavy and variable light domains, each of which also has an antigen binding domain to the tumor antigen, thereby allowing T cell recruitment to the tumor site.

[0272] In an alternative embodiment, the prodrug construct is shown in Figure 2 In the "Format 2" mechanism. In this embodiment, the domain linker between the active variable heavy chain and the active light chain is a constrained but non-cleavable linker ("CNCL"). In the prodrug form, the inactive VH and VL of the constrained pseudo-Fv domain associate with the VH and VL of the constrained Fv domain so that there is no CD3 binding. However, once cleavage occurs in the tumor environment, two different activated proteins (each comprising an active variable heavy chain and a light chain domain) associate to form two anti-CD3 binding domains. This Format 2 has two target tumor antigen binding domains ("TTA-ABD"), as described in more detail below, which may be the same (e.g., "same-COBRA") or different (e.g., "hetero-COBRA"). If different, they may each be directed against a different tumor antigen, or they may be directed against the same tumor antigen but different epitopes, as described more fully below.

[0273] In addition to the "single-chain protein" COBRA format discussed above, in which all components are contained on a single amino acid sequence, there are also constructs that rely on two proteins acting in pairs, "half-COBRA," as shown in Figure 3. In this embodiment, each protein has an active variable domain and an inert variable domain separated by a protease cleavage site. Each molecule contains a TTA binding domain, such that when the molecule is bound to TTA and exposed to tumor proteases, the inert domain is cleaved and the two active variable domains self-assemble to form an anti-CD3 binding domain.

[0274] In addition, the present invention also provides "Form 4" constructs, such as Figure 4 These are similar to the “Format 2” design, except that a single ABD directed against TTA is used, such that upon cleavage, the two prodrug molecules now form a tetravalent bispecific construct containing two active anti-CD3 domains, as described further below.

[0275] Thus, the forms and constructs of the invention are useful in treating disease.

[0276] definition

[0277] In order that this application may be more fully understood, several definitions are set forth below. Such definitions are intended to encompass grammatical equivalents.

[0278] As used herein, "amino acid" and "amino acid identity" mean one of the 20 naturally occurring amino acids or any non-natural analog that may be present at a specific, defined position. In many embodiments, "amino acid" refers to one of the 20 naturally occurring amino acids. "Protein" herein means at least two covalently attached amino acids, which includes proteins, polypeptides, oligopeptides, and peptides.

[0279] As used herein, "amino acid modification" means an amino acid substitution, insertion and / or deletion in a polypeptide sequence or an alteration of a portion chemically linked to a protein. For example, the modification can be an altered carbohydrate or PEG structure attached to a protein. For clarity, unless otherwise noted, amino acid modifications are always directed to amino acids encoded by DNA, such as the 20 amino acids for which there are codons in DNA and RNA. Preferred amino acid modifications herein are substitutions.

[0280] As used herein, "amino acid substitution" or "substitution" means that the amino acid at a specific position in the parent polypeptide sequence is replaced by a different amino acid. In particular, in some embodiments, the replacement is an amino acid that is not naturally occurring at the specific position, or is not naturally occurring in the organism or in any organism. For the sake of clarity, a protein that has been engineered to change the nucleic acid coding sequence but does not change the starting amino acid (e.g., exchanging CGG (encoding arginine) for CGA (still encoding arginine) to increase expression levels in the host organism) is not an "amino acid substitution"; that is, although a new gene encoding the same protein is generated, if the protein has the same amino acid at the specific position where it starts, the protein is not an amino acid substitution.

[0281] As used herein, "amino acid insertion" or "insertion" means the addition of an amino acid sequence at a specific position in a parent polypeptide sequence.

[0282] As used herein, "amino acid deletion" or "deletion" means the removal of an amino acid sequence at a specific position in a parent polypeptide sequence.

[0283] As outlined herein, the polypeptides of the present invention specifically bind to CD3 and a target tumor antigen (TTA), such as a target cell receptor. "Specific binding" or "specifically binds to" or "is specific for a particular antigen or epitope" refers to binding that is distinct from non-specific interactions. Specific binding can be measured, for example, by measuring the binding of a molecule compared to the binding of a control molecule, which is typically a molecule of similar structure that has no binding activity. For example, specific binding can be measured by competition with a control molecule that is similar to the target.

[0284] Specific binding for a particular antigen or epitope can be determined, for example, by a KD for the antigen or antigenic epitope of at least about 10 -4 M, at least about 10 -5 M, at least about 10 -6 M, at least about 10 -7 M, at least about 10 -8 M, at least about 10 -9 M, or at least about 10 -10 M, at least about 10 -11 M, at least about 10 -12 Antibodies with KDs of 50, 100, 500, 1000, 5,000, 10,000, or more are typically displayed that specifically bind an antigen with a KD of 20, 50, 100, 500, 1000, 5,000, 10,000, or more relative to the antigen or epitope for a control molecule.

[0285] In addition, specific binding to a particular antigen or antigenic epitope can be exhibited, for example, by an antibody having a KA or Ka for the antigen or antigenic epitope that is at least 20, 50, 100, 500, 1000, 5,000, 10,000 or more times greater than that of a control, where KA or Ka refers to the on-rate of the specific antibody-antigen interaction. Binding affinity is typically measured using Biacore assays or Octet assays known in the art.

[0286] As used herein, "parent polypeptide" or "precursor polypeptide" (including Fc parent or precursor) means a polypeptide that is subsequently modified to produce a variant. The parent polypeptide can be a naturally occurring polypeptide, or a variant or engineered version of a naturally occurring polypeptide. The parent polypeptide can refer to the polypeptide itself, a composition comprising the parent polypeptide, or an amino acid sequence encoding the polypeptide. Thus, as used herein, "parent Fc polypeptide" means an unmodified Fc polypeptide that is modified to produce a variant, and "parent antibody" as used herein means an unmodified antibody that is modified to produce a variant antibody.

[0287] As used herein, "position" refers to a position in the sequence of a protein. Positions can be numbered sequentially or according to an established format, such as the EU index for antibody numbering.

[0288] As used herein, "target antigen" refers to a molecule that is specifically bound by the variable region of a given antibody. The target antigen can be a protein, carbohydrate, lipid, or other chemical compound. A range of suitable exemplary target antigens are described herein.

[0289] As used herein, "target cell" means a cell that expresses a target antigen. Generally speaking, for the purposes of the present invention, a target cell is a tumor cell expressing TTA or a T cell expressing the CD3 antigen.

[0290] As used herein, the term "Fv" or "Fv fragment domain" or "Fv region" refers to a polypeptide comprising the VL and VH domains of the antigen binding domain typically from an antibody. If the Fv domain contains active VH and VL domains, the Fv domain typically forms an "antigen binding domain" or "ABD" as discussed herein (although in some cases, an Fv containing a constrained linker is used so that an active ABD is not formed before cleavage). As discussed below, in the present invention, the Fv domain can be organized in a variety of ways and can be "active" or "inactive," such as in the form of scFv, constrained Fv, pseudo-Fv, and the like. It should be understood that in the present invention, in some cases, the Fv domain consists of the VH and VL domains on a single polypeptide chain, such as Figure 1 and Figure 2In some embodiments, the Fv domain is composed of a VH and a VL domain, one of which is inert so that an intermolecular ABD is formed only after cleavage. As discussed below, in the present invention, the Fv domain can be organized in a variety of ways and can be "active" or "inactive", such as in scFv form, constrained Fv form, pseudo-Fv form, etc. In addition, as discussed herein, an Fv domain containing VH and VL can be / form an ABD, and other ABDs that do not comprise VH and VL domains can be formed using sdABDs.

[0291] By "variable region" herein is meant an immunoglobulin region comprising one or more Ig domains substantially encoded by any of the Vκ, Vλ, and / or VH genes that constitute the κ, λ, and heavy and light chain immunoglobulin loci, respectively. In some cases, a single variable domain may be used, such as sdFv (also referred to herein as sdABD).

[0292] In embodiments utilizing both a variable heavy (VH) domain and a variable light (VL) domain, each VH and VL is composed of three hypervariable regions ("complementarity determining regions," "CDRs") and four "framework regions," or "FRs," arranged from amino-terminus to carboxyl-terminus in the following order: FR1-CDR1-FR2-CDR2-FR3-CDR3-FR4. Thus, the VH domain has the structure vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCDR3-vhFR4, and the VL domain has the structure vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4. As described more fully herein, the vhFR and vlFR regions self-assemble to form an Fv domain. Generally speaking, in the prodrug forms of the invention, there is a "constrained Fv domain" in which the VH and VL domains are unable to self-associate, and a "pseudo-Fv domain" in which the CDRs do not form an antigen binding domain when self-associated.

[0293] The hypervariable region confers antigen-binding specificity and generally encompasses amino acid residues 24-34 (LCDR1; "L" for light chain), 50-56 (LCDR2), and 89-97 (LCDR3) from the light chain variable region and approximately 31-35B (HCDR1; "H" for heavy chain), 50-65 (HCDR2), and 95-102 (HCDR3) from the heavy chain variable region; Kabat et al., SEQUENCES OF PROTEINS OF IMMUNOLOGICAL INTEREST, Public Health Service, 5th ed., National Institutes of Health. Health, Bethesda Md. (1991) and / or those residues forming the hypervariable loop in the light chain variable region (e.g., residues 26-32 (LCDR1), 50-52 (LCDR2), and 91-96 (LCDR3)) and those residues forming the hypervariable loop in the heavy chain variable region, 26-32 (HCDR1), 53-55 (HCDR2), and 96-101 (HCDR3); Chothia and Lesk (1987) J. Mol. Biol. 196:901-917. Specific CDRs of the invention are described below.

[0294] As will be appreciated by those skilled in the art, the exact numbering and position of CDRs may be different in different numbering systems. However, it will be understood that the disclosure of variable heavy and / or variable light sequences includes the disclosure of the associated (intrinsic) CDRs. Thus, the disclosure of each variable heavy region is the disclosure of the vhCDRs (e.g., vhCDR1, vhCDR2, and vhCDR3), and the disclosure of each variable light region is the disclosure of the vlCDRs (e.g., vlCDR1, vlCDR2, and vlCDR3).

[0295] A useful comparison of CDR numbering is as follows, see Lafranc et al., Dev. Comp. Immunol. 27(1):55-77 (2003):

[0296] Table 1

[0297] Kabat+Chothia IMGT Kabat AbM Chothia Contact vhCDR1 26-35 27-38 31-35 26-35 26-32 30-35 vhCDR2 50-65 56-65 50-65 50-58 52-56 47-58 vhCDR3 95-102 105-117 95-102 95-102 95-102 93-101 vlCDR1 24-34 27-38 24-34 24-34 24-34 30-36 vlCDR2 50-56 56-65 50-56 50-56 50-56 46-55 vlCDR3 89-97 105-117 89-97 89-97 89-97 89-96

[0298] Throughout this specification, the Kabat numbering system and the EU numbering system for Fc regions (e.g., Kabat et al., supra (1991)) are generally used when referring to residues in the variable domain (approximately, residues 1-107 in the light chain variable region and residues 1-113 in the heavy chain variable region).

[0299] The present invention provides a large number of different CDR sets. In this case, a "full CDR set" in the context of an anti-CD3 component comprises three variable light CDRs and three variable heavy CDRs, e.g., vlCDR1, vlCDR2, vlCDR3, vhCDR1, vhCDR2, and vhCDR3. As will be appreciated by those skilled in the art, each set of CDRs, i.e., the VH and VL CDRs, can bind to an antigen both individually and as a group. For example, in a constrained Fv domain, the vhCDRs can, for example, bind to CD3, and the vlCDRs can bind to CD3, but in a constrained form, they cannot bind to CD3.

[0300] In the context of a single-domain ABD ("sdABD"), as typically used herein for binding to a target tumor antigen (TTA), the set of CDRs is only three CDRs; these are sometimes also referred to in the art as "VHH" domains.

[0301] These CDRs can be part of a larger variable light domain or a variable heavy domain. In addition, as more fully outlined herein, in the case of scFv sequences, the variable heavy domain and the variable light domain can be on separate polypeptide chains or on a single polypeptide chain, depending on the form and configuration of the moieties herein.

[0302] CDRs facilitate the formation of antigen binding sites, or more particularly epitope binding sites. An "epitope" refers to a determinant that interacts with a specific antigen binding site in the variable region, called a paratope. An antigenic epitope is a group of molecules such as amino acids or sugar side chains, and typically has specific structural characteristics, as well as specific charge characteristics. A single antigen may have more than one epitope.

[0303] An epitope may include amino acid residues that are directly involved in binding (also called the immunodominant component of the epitope) as well as other amino acid residues that are not directly involved in binding, such as amino acid residues that are effectively blocked by the specific antigen-binding peptide; in other words, the amino acid residues are within the footprint of the specific antigen-binding peptide.

[0304] Epitopes can be conformational or linear. Conformational epitopes are produced by spatially juxtaposed amino acids in different segments of a linear polypeptide chain. Linear epitopes are produced by adjacent amino acid residues in a polypeptide chain. Conformational and nonconformational epitopes may be distinguished by the loss of binding to the former, while not to the latter, in the presence of denaturing solvents.

[0305] An epitope typically includes at least 3, and more typically at least 5 or 8-10 amino acids in a unique spatial conformation. Antibodies that recognize the same epitope can be validated in a simple immunoassay that shows the ability of one antibody to block another antibody from binding to the target antigen, such as "binning." As outlined below, the present invention encompasses not only the antigen-binding domains and antibodies listed herein, but also those that compete for binding to the epitopes bound by the listed antigen-binding domains.

[0306] The variable heavy and variable light domains of the present invention may be "active" or "inactive."

[0307] As used herein, "inactive VH" ("iVH") and "inactive VL" ("iVL") refer to components of pseudo-Fv domains that, when paired with their cognate VL or VH partners, respectively, form a resulting VH / VL pair that does not specifically bind to the antigen to which an "active" VH or "active" VL would bind if it were bound to a similar VL or VH that is not "inactive." Exemplary "inactive VH" and "inactive VL" domains are formed by mutation of wild-type VH or VL sequences, as more fully summarized below. Exemplary mutations are within CDR1, CDR2, or CDR3 of a VH or VL. Exemplary mutations include placing a domain linker within CDR2, thereby forming an "inactive VH" or "inactive VL" domain. In contrast, an "active VH" or "active VL" is one that is capable of specifically binding to its target antigen when paired with its "active" cognate partner, i.e., VL or VH, respectively. Thus, it will be understood that a pseudo-Fv can be a VH / iVL pair, an iVH / VL pair, or an iVH / iVL pair.

[0308] In contrast, as used herein, the term "active" refers to a CD-3 binding domain that is capable of specifically binding to CD-3. This term is used in two contexts: (a) when referring to a single member of an Fv binding pair (i.e., VH or VL), which has a sequence that is capable of pairing with its cognate partner and specifically binding to CD-3; and (b) a pair of homologs (i.e., VH and VL) of sequences that are capable of specifically binding to CD-3. An exemplary "active" VH, VL, or VH / VL pair is the wild-type or parental sequence.

[0309] "CD-x" refers to a group of differentiation (CD) proteins. In exemplary embodiments, CD-x is selected from those CD proteins that have an effect on the recruitment or activation of T cells in a subject to which a polypeptide construct of the invention is administered. In an exemplary embodiment, CD-x is CD3, the sequence of which is shown in Figure 5.

[0310] The term "binding domain" in the context of the present invention refers to a domain that (specifically) binds to / interacts with / recognizes a given target epitope or a given target site on a target molecule (antigen) (e.g., EGFR and CD-3, respectively). The structure and function of the target antigen binding domain (recognizing EGFR), and preferably also the structure and / or function of the CD-3 binding domain (recognizing CD3), are based on the structure and / or function of an antibody (e.g., a full-length or entire immunoglobulin molecule, including an sdABD). According to the present invention, the target antigen binding domain is typically characterized by the presence of three CDRs that bind to the target tumor antigen (commonly referred to in the art as a variable heavy domain, even though the corresponding light chain CDRs are not present). Alternatively, an ABD for TTA may comprise three light chain CDRs (i.e., CDR1, CDR2, and CDR3 of the VL region) and / or three heavy chain CDRs (i.e., CDR1, CDR2, and CDR3 of the VH region). The CD-3 binding domain preferably also comprises at least the minimum structural requirements of an antibody that permits target binding. More preferably, the CD-3 binding domain comprises at least three light chain CDRs (i.e., CDR1, CDR2, and CDR3 of the VL region) and / or three heavy chain CDRs (i.e., CDR1, CDR2, and CDR3 of the VH region). It is envisioned that in exemplary embodiments, the target antigen and / or CD-3 binding domain is generated or obtainable by phage display or library screening methods.

[0311] As used herein, "domain" refers to a protein sequence with a function, as summarized herein. The domains of the present invention include tumor target antigen binding domains (TTA domains), variable heavy domains, variable light domains, scFv domains, linker domains, and half-life extension domains.

[0312] "Domain linker" herein refers to an amino acid sequence connecting two domains as outlined herein. Domain linkers can be cleavable linkers, constrained cleavable linkers, non-cleavable linkers, constrained non-cleavable linkers, scFv linkers, etc.

[0313] As used herein, a "cleavable linker" ("CL") refers to an amino acid sequence that can be cleaved by proteases, preferably human proteases, in diseased tissues as outlined herein. Cleavable linkers are generally at least 3 amino acids in length, with 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 or more amino acids being useful in the present invention, depending on the desired flexibility. Many cleavable linker sequences are found in Figures 6 and 5.

[0314] A "non-cleavable linker" ("NCL") herein refers to an amino acid sequence that cannot be cleaved by human proteases under normal physiological conditions.

[0315] A "constrained cleavable linker" ("CCL") herein refers to a short polypeptide containing a protease cleavage site (as defined herein) that connects two domains as outlined herein in a manner such that the two domains cannot significantly interact with each other until they reside on different polypeptide chains, e.g., after cleavage. When the CCL connects the VH and VL domains as defined herein, due to spatial constraints in an intramolecular manner, the VH and VL cannot self-assemble to form a functional Fv before cleavage (although they can assemble into pseudo-Fv domains in an intermolecular manner). After cleavage by the relevant protease, the VH and VL can assemble in an intermolecular manner to form an active antigen binding domain. In general, CCLs are less than 10 amino acids in length, of which 9, 8, 7, 6, 5, and 4 amino acids can be used in the present invention. In general, the length of the protease cleavage site is typically at least 4+ amino acids to confer sufficient specificity, as shown in Figure 6.

[0316] A "constrained non-cleavable linker" ("CNCL") herein refers to a short polypeptide that connects two domains as outlined herein in a manner that the two domains cannot significantly interact with each other and are not significantly cleaved by human proteases under physiological conditions.

[0317] By "constrained Fv domain" herein is meant an Fv domain comprising an active variable heavy domain and an active variable light domain, said Fv domain being covalently linked to a constrained linker as outlined herein in such a manner that the active heavy and light variable domains cannot interact intramolecularly to form an active Fv that will bind to an antigen (such as CD3). Thus, constrained Fv domains are Fv domains that are similar to scFvs but cannot bind to antigens due to the presence of a constrained linker (although they can assemble intermolecularly with inert variable domains to form pseudo-Fv domains).

[0318] As used herein, a "pseudo-Fv domain" refers to a domain comprising a pseudo or inactive variable heavy domain or a pseudo or inactive variable light domain connected using a domain linker (which can be cleavable, constrained, non-cleavable, non-constrained, etc.). The iVH and iVL domains of the pseudo-Fv domain do not bind to human antigens when associated with each other (iVH / iVL) or with an active VH or VL; thus, iVH / iVL, iVH / VL, and iVL / VH Fv domains do not significantly bind to human proteins, rendering these domains inert in humans.

[0319] By "single-chain Fv" or "scFv" herein is meant a variable heavy (VH) domain covalently attached to a variable light (VL) domain, typically using an scFv linker as discussed herein, to form an scFv or scFv domain. The scFv domains can be located in either orientation from N-terminus to C-terminus (VH-linker-VL or VL-linker-VH).

[0320] "Single domain Fv", "sdFv" or "sdABD" herein refers to an antigen binding domain having only three CDRs, typically based on camelid antibody technology. See: Protein Engineering 9(7):1129-35 (1994); Rev Mol Biotech 74:277-302 (2001); Ann Rev Biochem 82:775-97 (2013). As outlined herein, two general types of sdABD are used herein: sdABDs that bind to TTA and are annotated as such (sdABD-TTA for the general term, sdABD-EGFR for one that binds to EGFR, and sdABD-FOLR1 for one that binds to FOLR1) and sdABDs that bind to HSA ("sdABD-HSA" or "sdABD (1 / 2)").

[0321] A "protease cleavage site" refers to an amino acid sequence that is recognized and cleaved by a protease. Suitable protease cleavage sites are summarized below and shown in Figures 5 and 6.

[0322] As used herein, a "protease cleavage domain" refers to a domain that is incorporated into a "protease cleavage site" and the space between individual protease cleavage sites and between the protease cleavage sites and other functional components of the construct of the invention (e.g., V H 、V L The peptide sequence of any linker between the protease cleavage domain, the iVL target antigen binding domain, the half-life extension domain, etc. As described herein, the protease cleavage domain may also contain additional amino acids when necessary, for example, to impart flexibility.

[0323] The term "COBRA TM " and "conditional bispecific redirected activation" refer to bispecific conditionally active proteins having a number of functional protein domains. In some embodiments, one of the functional domains is an antigen binding domain (ABD) that binds to a target tumor antigen (TTA). In certain embodiments, another domain is an ABD that binds to a T cell antigen under certain conditions. T cell antigens include but are not limited to CD3. The term "half-COBRA TM" refers to the variable heavy chain of a half-COBRA that can bind to another half-COBRA due to innate self-assembly when concentrated on the surface of the target expressing cell. TM (Complementary COBRA TM ) is a conditionally effective protein that can bind to a T cell antigen when the variable light chain of the DETAILED DESCRIPTION

[0325] I. Fusion Proteins of the Present Invention

[0326] The fusion proteins of the present invention have many different components (generally referred to herein as domains) connected together in a variety of ways. Some of the domains are binding domains that each bind to a target antigen (e.g., TTA or CD3). Because they bind to more than one antigen, they are referred to herein as "multispecific"; for example, the prodrug constructs of the present invention can bind to TTA and CD3 and are therefore "bispecific." Proteins can also have higher specificity; for example, if the first αTTA binds to EGFR, the second αTTA binds to EpCAM, and there is an anti-CD3 binding domain, this will be a "trispecific" molecule. Similarly, adding an anti-HSA binding domain to this construct will be "tetraspecific," such as Figure 3B As shown in .

[0327] As will be appreciated by those skilled in the art, the proteins of the invention may have different valencies and be multispecific. That is, the proteins of the invention may bind to targets having more than one binding site; for example, Pro140 is bivalent for EGFR.

[0328] The proteins of the invention may include a CD3 antigen binding domain, a tumor target antigen binding domain, a half-life extending domain, a linker, etc., arranged in a variety of ways as outlined herein.

[0329] A. CD3 antigen binding domain

[0330] The specificity of T cell responses is mediated by the recognition of antigens by the T cell receptor complex (presented in the context of the major histocompatibility complex, MHC). As part of the T cell receptor complex, CD3 is a protein complex that contains the CD3γ (γ) chain, CD3δ (δ) chain, two CD3e (ε) chains, and two CD3ζ (ζ) chains present on the cell surface. The CD3 molecule associates with the α (α) and β (β) chains of the T cell receptor (TCR) to form the TCR complex. Clustering of CD3 on T cells (e.g., by binding to the Fv domain of CD3) leads to T cell activation, similar to engagement of the T cell receptor, but independent of its clone-typical specificity.

[0331] However, as is known in the art, CD3 activation can cause a number of toxic side effects, and therefore, the present invention is directed to providing active CD3 binding of the polypeptides of the present invention only in the presence of tumor cells in which specific proteases are found, which then cleave the prodrug polypeptides of the present invention to provide an active CD3 binding domain. Thus, in the present invention, the binding of the anti-CD-3 Fv domain to CD-3 is regulated by a protease cleavage domain that restricts the binding of the CD-3 Fv domain to CD-3 only in the microenvironment of diseased cells or tissues where protease levels are elevated (e.g., in the tumor microenvironment described herein).

[0332] Thus, the present invention provides two sets of VH and VL domains, one active set (VH and VL) and one inactive set (iVH and iVL), all four of which are present in a prodrug construct. The construct is formatted so that the VH and VL sets cannot self-associate, but instead associate with an inactive partner (e.g., iVH and VL and iVL and VH as shown herein).

[0333] 1. Active anti-CD3 variable heavy domain and variable light domain

[0334] There are many suitable active CDR sets and / or VH and VL domains known in the art that are suitable for use in the present invention. For example, the CDRs and / or VH and VL domains are derived from known anti-CD-3 antibodies, such as, for example, Muromonab-CD-3 (OKT3), Ocinetuzumab (TRX4), Teplizumab (MGA031), Visilizumab (Nuvion), SP34 or I2C, TR-66 or X35-3, VIT3, BMA030 (BW264 / 56), CLB-T3 / 3, CRISPR-Cas9, or the like. 7, YTH12.5, F111-409, CLB-T3.4.2, TR-66, WT32, SPv-T3b, 11D8, XIII-141, XIII-46, XIII-87, 12F6, T3 / RW2-8C8, T3 / RW2-4B6, OKT3D, M-T301, SMC2, F101.01, UCHT-1, and WT-31.

[0335] In one embodiment, the VH and VL sequences that form active Fv domains that bind to human CD3 are shown in Figure 5. As shown herein, these active VH ("aVH") and active VL ("aVL") domains can be used in different configurations and formats 1, 2, 3, and 4.

[0336] 2. Inactive anti-CD3 variable heavy domain and variable light domain

[0337] The inactive iVH and iVL domains contain "conventional" framework regions (FRs) that allow association such that the inactive variable domain will associate with the active variable domain, rendering the pair inactive, e.g., unable to bind CD3.

[0338] As will be appreciated by those skilled in the art, there are many "inactive" variable domains that can be used in the present invention. Essentially, any variable domain with a human framework region that allows self-assembly with another variable domain can be used, regardless of the amino acids in the CDR positions in the variable region. For clarity, an inactive domain is said to contain CDRs, even though technically an inactive variable domain does not confer binding ability.

[0339] As will be appreciated in the art, the generation of inactive VH or VL domains is generally simple and can be accomplished in a variety of ways. In some embodiments, the generation of inactive variable domains is typically accomplished by changing one or more CDRs of the active Fv, including changing one or more of the three CDRs of the active variable domain. This can be accomplished by performing one or more amino acid substitutions at functionally important residues in one or more CDRs, replacing some or all of the CDR residues with random sequences, replacing one or more CDRs with tags or marker sequences, and / or exchanging CDRs and / or variable regions with CDRs and / or variable regions from unrelated antibodies (e.g., CDRs and / or variable regions of proteins directed against different organisms).

[0340] In some cases, only one of the CDRs in the variable region may be altered to render it inactive, although other embodiments include alterations in 1, 2, 3, 4, 5, or 6 CDRs.

[0341] In some cases, the inactive domains can be engineered to promote selective binding in a prodrug form to promote formation of intramolecular iVH-VL and VH-iVL domains prior to cleavage (e.g., over intermolecular pair formation). See, e.g., Igawa et al. Protein Eng. Des. Selection 23(8):667-677 (2010), hereby expressly incorporated by reference in its entirety, and particularly with respect to amino acid substitutions of interface residues.

[0342] In certain embodiments, the CD-3 binding domains of the polypeptide constructs described herein not only exhibit effective CD-3 binding affinity to human CD-3, but also exhibit excellent cross-reactivity with the corresponding cynomolgus monkey CD-3 protein. In some cases, the CD-3 binding domains of the polypeptide constructs cross-react with CD-3 from cynomolgus monkeys. In some cases, the human:cynomolgus monkey KD ratio for CD-3 is between 5 and 0.2.

[0343] In some embodiments, the CD-3 binding domain of the antigen binding protein can be any domain that binds to CD-3, including but not limited to domains from monoclonal antibodies, polyclonal antibodies, recombinant antibodies, human antibodies, and humanized antibodies. In some cases, it is beneficial for the CD-3 binding domain to be derived from the same species in which the antigen binding protein will ultimately be used. For example, for use in humans, it may be beneficial for the CD-3 binding domain of the antigen binding protein to comprise human or humanized residues from the antigen binding domain of an antibody or antibody fragment.

[0344] Thus, in one aspect, the antigen binding domain comprises a humanized or human binding domain. In one embodiment, the humanized or human anti-CD-3 binding domain comprises one or more (e.g., all three) light chain complementary determining region 1 (LC CDR1), light chain complementary determining region 2 (LC CDR2), and light chain complementary determining region 3 (LC CDR3) of a humanized or human anti-CD-3 binding domain described herein, and / or one or more (e.g., all three) heavy chain complementary determining region 1 (HC CDR1), heavy chain complementary determining region 2 (HC CDR2), and heavy chain complementary determining region 3 (HC CDR3) of a humanized or human anti-CD-3 binding domain described herein, e.g., a humanized or human anti-CD-3 binding domain comprising one or more (e.g., all three) LC CDRs and one or more (e.g., all three) HC CDRs.

[0345] In some embodiments, the humanized or human anti-CD3 binding domain comprises a humanized or human light chain variable region that is specific for CD3, wherein the light chain variable region that is specific for CD3 comprises human or non-human light chain CDRs in a human light chain framework region. In some cases, the light chain framework region is a lambda (λ) light chain framework. In other cases, the light chain framework region is a kappa (κ) light chain framework.

[0346] In some embodiments, one or more CD-3 binding domains are humanized or fully human. In some embodiments, one or more activated CD-3 binding domains bind to CD-3 on cells expressing CD-3 with a KD of 1000 nM or less. In some embodiments, one or more activated CD-3 binding domains bind to CD-3 on cells expressing CD-3 with a KD of 100 nM or less. In some embodiments, one or more activated CD-3 binding domains bind to CD-3 on cells expressing CD-3 with a KD of 10 nM or less. In some embodiments, one or more CD-3 binding domains cross-react with cynomolgus monkey CD-3. In some embodiments, one or more CD-3 binding domains comprise an amino acid sequence provided herein.

[0347] In some embodiments, the humanized or human anti-CD-3 binding domain comprises a humanized or human heavy chain variable region that is specific for CD-3, wherein the heavy chain variable region that is specific for CD-3 comprises human or non-human heavy chain CDRs within a human heavy chain framework region.

[0348] In one embodiment, the anti-CD-3 binding domain is an Fv comprising a light chain and a heavy chain having an amino acid sequence provided herein. In one embodiment, the anti-CD-3 binding domain comprises: a light chain variable region comprising an amino acid sequence having a light chain variable region provided herein, at least one, two or three modifications (e.g., substitutions), but no more than 30, 20 or 10 modifications (e.g., substitutions), or a sequence with 95%-99% identity to an amino acid sequence provided herein; and / or a heavy chain variable region comprising an amino acid sequence having a heavy chain variable region provided herein, at least one, two or three modifications (e.g., substitutions), but no more than 30, 20 or 10 modifications (e.g., substitutions), or a sequence with 95%-99% identity to an amino acid sequence provided herein. In one embodiment, the humanized or human anti-CD-3 binding domain is an scFv, and the light chain variable region comprising the amino acid sequence described herein is attached to the heavy chain variable region comprising the amino acid sequence described herein via an scFv linker. The light chain variable region and heavy chain variable region of the scFv can be, for example, in any of the following orientations: light chain variable region-scFv linker-heavy chain variable region or heavy chain variable region-scFv linker-light chain variable region.

[0349] In some embodiments, the CD-3 binding domain of the antigen binding protein has an affinity for CD-3 on cells expressing CD3 with a KD of 1000 nM or less, 100 nM or less, 50 nM or less, 20 nM or less, 10 nM or less, 5 nM or less, 1 nM or less, or 0.5 nM or less. In some embodiments, the CD-3 binding domain of the antigen binding protein has an affinity for CD-3 epsilon with a KD of 1000 nM or less, 100 nM or less, 50 nM or less, 20 nM or less, 10 nM or less, 5 nM or less, 1 nM or less, or 0.5 nM or less. In other embodiments, the CD-3 binding domain of the antigen binding protein has a low affinity for CD-3, i.e., about 100 nM or greater.

[0350] The affinity for binding to CD-3 can be determined, for example, by the ability of the antigen-binding protein itself or its CD-3 binding domain to bind to CD-3 coated on an assay plate; displayed on the surface of a microbial cell; in solution, etc., as known in the art, typically using Biacore or Octet assays. The binding activity of the antigen-binding protein itself or its CD-3 binding domain of the present disclosure to CD-3 can be determined by immobilizing a ligand (e.g., CD-3) or the antigen-binding protein itself or its CD-3 binding domain to beads, substrates, cells, etc. The reagent can be added in an appropriate buffer and the binding partner can be incubated at a given temperature for a period of time. After washing to remove unbound material, the bound protein can be released with, for example, SDS, a buffer having a high pH, etc., and analyzed, for example, by surface plasmon resonance (SPR).

[0351] In many embodiments, preferred active and inactive binding domains are those shown in Figure 5. Figure 5 depicts one active VH and VL and three inactive VHi and three inactive VLi that have been inactivated in different ways.

[0352] As shown in Figure 5, a particularly useful pair of active anti-CD3 VL and VH domains has a VL having: vlCDR1 having SEQ ID NO: 127, vlCDR2 having SEQ ID NO: 128, and vlCDR3 having SEQ ID NO: 129; and a VH having: vhCDR1 having SEQ ID NO: 143, vhCDR2 having SEQ ID NO: 144, and vhCDR3 having SEQ ID NO: 145.

[0353] As shown in FIG5 , a particularly useful pair of active anti-CD3 VL and VH domains has a VL having SEQ ID NO: 126 and a VH having SEQ ID NO: 142.

[0354] B. Antigen-binding domains targeting tumor target antigens

[0355] In addition to the described CD3 and half-life extension domains, the polypeptide constructs described herein also include a target domain that binds to one or more target antigens or one or more regions on a single target antigen. It is contemplated herein that the polypeptide constructs of the present invention are cleaved, for example, at the protease cleavage domain in a disease-specific microenvironment or in the blood of a subject, and each target antigen binding domain will bind to the target antigen on the target cell, thereby activating the CD3 binding domain to bind to the T cell. In general, the TTA binding domain can bind to its target before protease cleavage, so they can "wait" on the target cell to be activated as a T cell adaptor. At least one target antigen is involved in a disease, disorder or condition and / or is associated with a disease, disorder or condition. Exemplary target antigens include those associated with proliferative diseases, neoplastic diseases, inflammatory diseases, immune disorders, autoimmune diseases, infectious diseases, viral diseases, allergic reactions, parasitic reactions, graft-versus-host disease, or host-versus-graft disease. In some embodiments, the target antigen is a tumor antigen expressed on a tumor cell. Alternatively, in some embodiments, the target antigen is associated with a pathogen (such as a virus or bacteria). At least one target antigen may also be directed against healthy tissue.

[0356] In some embodiments, the target antigen is a cell surface molecule, such as a protein, lipid, or polysaccharide. In some embodiments, the target antigen is on a tumor cell, a virally infected cell, a bacterially infected cell, a damaged red blood cell, an arterial plaque cell, or a fibrotic tissue cell.

[0357] Preferred embodiments of the present invention utilize sdABDs as targeting domains. These are preferred over scFv ABDs because the addition of additional VH and VL domains to the constructs of the present invention may complicate the formation of pseudo-Fv domains.

[0358] In some embodiments, the prodrug constructs of the present invention utilize a single TTA binding domain, such as generally in Figure 3A Depicted as a paired sdABD-TTA in Figure 4 Depicted as "Form 4" configuration Figure 4 The use of a single anti-EGFR ABD is shown, although other TTA binding domains can be used.

[0359] In some embodiments, particularly in Form 1 and Form 2 constructs, the prodrug constructs of the present invention utilize two TTA ABDs, again preferably in the form of sdABD-TTA. When dual targeting domains are used, they can bind to the same epitope of the same TTA. For example, as discussed herein, many constructs herein utilize two identical targeting domains. In some embodiments, two targeting domains can be used that bind to different epitopes of the same TTA, for example, as shown in FIG5 , two EGFR sdABDs bind to different epitopes on human EGFR. In some embodiments, the two targeting domains bind to different TTAs, see, for example, FIG5 .

[0360] The polypeptide constructs contemplated herein comprise at least one antigen binding domain, wherein the antigen binding domain binds to at least one target antigen. In some embodiments, the target antigen binding domain specifically binds to a cell surface molecule. In some embodiments, the target antigen binding domain specifically binds to a tumor antigen. In some embodiments, the target antigen binding domain specifically and independently binds to a tumor target antigen ("TTA") selected from at least one of EpCAM, EGFR, HER-2, HER-3, cMet, LyPD3, B7H3, CEA, Trop2, and FOLR1.

[0361] (a) EGFR sdABD

[0362] As shown in FIG5 , there are a number of particularly useful sdABDs that bind to human EGFR, referred to herein as “sdABD-EGFR” or “EGFR ABD”.

[0363] In one useful embodiment, the sdABD-EGFR1 has: an sdCDR1 having SEQ ID NO: 10, an sdCDR2 having SEQ ID NO: 11, and an sdCDR3 having SEQ ID NO: 12. In some cases, the sdABD-EGFR has SEQ ID NO:9.

[0364] In one useful embodiment, the sdABD-EGFR2a has: an sdCDR1 having SEQ ID NO: 14, an sdCDR2 having SEQ ID NO: 15, and an sdCDR3 having SEQ ID NO: 16. In some cases, the sdABD-EGFR has SEQ ID NO:13.

[0365] In one useful embodiment, the sdABD-EGFR2d has: an sdCDR1 having SEQ ID NO: 18, an sdCDR2 having SEQ ID NO: 19, and an sdCDR3 having SEQ ID NO: 20. In some cases, the sdABD-EGFR has SEQ ID NO: 17.

[0366] (b)EpCAM sdABD

[0367] As shown in FIG5 , there are a number of particularly useful sdABDs that bind to human EpCAM, referred to herein as “sdABD-EpCAM” or “EpCAMABD”.

[0368] In one useful embodiment, the sdABD-EpCAM h13 has: sdCDR1 having SEQ ID NO: 62, sdCDR2 having SEQ ID NO: 63, sdCDR3 having SEQ ID NO: 64. In some cases, the sdABD-EpCAM has SEQ ID NO: 61.

[0369] In one useful embodiment, sdABD-EpCAM h23 has: sdCDR1 having SEQ ID NO: 66, sdCDR2 having SEQ ID NO: 67, sdCDR3 having SEQ ID NO: 68. In some cases, the sdABD-EpCAM has SEQ ID NO: 65.

[0370] In one useful embodiment, the sdABD-EpCAM hVIB665 has: an sdCDR1 having SEQ ID NO: 70, an sdCDR2 having SEQ ID NO: 71, an sdCDR3 having SEQ ID NO: 72. In some cases, the sdABD-EpCAM has SEQ ID NO: 69. It should be noted that, in contrast to the h13 and h23 EpCAM sdABDs, hVIB665 (also referred to as "acEpCAM hVIB665") binds to both cleaved and uncleaved forms of EpCAM (which are known to undergo cleavage in vivo).

[0371] In one useful embodiment, the sdABD-EpCAM hVIB666 has: an sdCDR1 having SEQ ID NO: 74, an sdCDR2 having SEQ ID NO: 75, an sdCDR3 having SEQ ID NO: 76. In some cases, the sdABD-EpCAM has SEQ ID NO: 73. It should be noted that, in contrast to the h13 and h23 EpCAM sdABDs, hVIB666 (also referred to as "acEpCAM hVIB666") binds to both cleaved and uncleaved forms of EpCAM (which are known to undergo cleavage in vivo).

[0372] (c)B7H3 sdABD

[0373] As shown in FIG5 , there are a number of particularly useful sdABDs that bind to human B7H3, referred to herein as “sdABD-B7H3” or “B7H3-ABD”.

[0374] In one useful embodiment, the sdABD-B7H3 hF7 has: an sdCDR1 having SEQ ID NO: 34, an sdCDR2 having SEQ ID NO: 35, and an sdCDR3 having SEQ ID NO: 36. In some cases, the sdABD-B7H3 has SEQ ID NO: 33.

[0375] In one useful embodiment, the sdABD-B7H3 hF12 has: an sdCDR1 having SEQ ID NO: 38, an sdCDR2 having SEQ ID NO: 39, and an sdCDR3 having SEQ ID NO: 40. In some cases, the sdABD-B7H3 has SEQ ID NO: 37.

[0376] In one useful embodiment, the sdABD-B7H3 hF12(N57Q) has: an sdCDR1 having SEQ ID NO: 42, an sdCDR2 having SEQ ID NO: 43, and an sdCDR3 having SEQ ID NO: 44. In some cases, the sdABD-B7H3 has SEQ ID NO: 41. Compared to hF7 and hF12 B7H3 sdABD, the amino acid substitution N57Q removes a glycosylation site.

[0377] In one useful embodiment, sdABD-B7H3 HF12(N57E) has: an sdCDR1 having SEQ ID NO: 46, an sdCDR2 having SEQ ID NO: 47, and an sdCDR3 having SEQ ID NO: 48. In some cases, the sdABD-B7H3 has SEQ ID NO: 45. Compared to hF7 and hF12 B7H3 sdABDs, the amino acid substitution N57E removes a glycosylation site.

[0378] In one useful embodiment, the sdABD-B7H3 hF12(N57D) has: an sdCDR1 having SEQ ID NO: 50, an sdCDR2 having SEQ ID NO: 51, and an sdCDR3 having SEQ ID NO: 52. In some cases, the sdABD-B7H3 has SEQ ID NO: 49. Compared to hF7 and hF12 B7H3 sdABDs, the amino acid substitution N57D removes a glycosylation site.

[0379] In one useful embodiment, the sdABD-B7H3 hF12(S59A) has: an sdCDR1 having SEQ ID NO: 54, an sdCDR2 having SEQ ID NO: 55, and an sdCDR3 having SEQ ID NO: 56. In some cases, the sdABD-B7H3 has SEQ ID NO: 53. Compared to hF7 and hF12 B7H3 sdABDs, the amino acid substitution S59A removes a glycosylation site.

[0380] In one useful embodiment, the sdABD-B7H3 hF12(S59Y) has: an sdCDR1 having SEQ ID NO: 58, an sdCDR2 having SEQ ID NO: 59, and an sdCDR3 having SEQ ID NO: 60. In some cases, the sdABD-B7H3 has SEQ ID NO: 57. Compared to hF7 and hF12 B7H3 sdABDs, the amino acid substitution NS59Y removes a glycosylation site.

[0381] (d)FOLR1 sdABD

[0382] As shown in FIG5 , there are a number of particularly useful sdABDs that bind to human FOLR1 , referred to herein as “sdABD-FOLR1 ” or “FOLR1 -ABD”.

[0383] In one useful embodiment, sdABD-FOLR1 h77-2 has: sdCDR1 having SEQ ID NO: 22, sdCDR2 having SEQ ID NO: 23, sdCDR3 having SEQ ID NO: 24. In some cases, the sdABD-FOLR1 has SEQ ID NO: 21.

[0384] In one useful embodiment, sdABD-FOLR1 h59.3 has: sdCDR1 having SEQ ID NO: 26, sdCDR2 having SEQ ID NO: 27, sdCDR3 having SEQ ID NO: 28. In some cases, the sdABD-FOLR1 has SEQ ID NO: 25.

[0385] In one useful embodiment, sdABD-FOLR1 h22-4 has: sdCDR1 having SEQ ID NO: 30, sdCDR2 having SEQ ID NO: 31, sdCDR3 having SEQ ID NO: 32. In some cases, the sdABD-FOLR1 has SEQ ID NO: 29.

[0386] (e)Trop2 sdABD

[0387] As shown in FIG5 , there are a number of particularly useful sdABDs that bind to human Trop2, referred to herein as “sdABD-Trop2” or “Trop2-ABD”.

[0388] In one useful embodiment, the sdABD-Trop2 hVIB557 has: an sdCDR1 having SEQ ID NO: 78, an sdCDR2 having SEQ ID NO: 79, an sdCDR3 having SEQ ID NO: 80. In some cases, the sdABD-Trop2 has SEQ ID NO: 77.

[0389] In one useful embodiment, the sdABD-Trop2 hVIB565 has: an sdCDR1 having SEQ ID NO: 82, an sdCDR2 having SEQ ID NO: 83, and an sdCDR3 having SEQ ID NO: 84. In some cases, the sdABD-Trop2 has SEQ ID NO: 81.

[0390] In one useful embodiment, the sdABD-Trop2 hVIB575 has: an sdCDR1 having SEQ ID NO: 86, an sdCDR2 having SEQ ID NO: 87, an sdCDR3 having SEQ ID NO: 88. In some cases, the sdABD-Trop2 has SEQ ID NO: 85.

[0391] In one useful embodiment, the sdABD-Trop2 hVIB578 has: an sdCDR1 having SEQ ID NO: 90, an sdCDR2 having SEQ ID NO: 01, and an sdCDR3 having SEQ ID NO: 92. In some cases, the sdABD-Trop2 has SEQ ID NO: 89.

[0392] In one useful embodiment, the sdABD-Trop2 hVIB609 has: an sdCDR1 having SEQ ID NO: 94, an sdCDR2 having SEQ ID NO: 95, and an sdCDR3 having SEQ ID NO: 96. In some cases, the sdABD-Trop2 has SEQ ID NO: 93.

[0393] In one useful embodiment, the sdABD-Trop2 hVIB619 has: an sdCDR1 having SEQ ID NO: 98, an sdCDR2 having SEQ ID NO: 99, an sdCDR3 having SEQ ID NO: 100. In some cases, the sdABD-Trop2 has SEQ ID NO: 97.

[0394] (f)CA9 sdABD

[0395] As shown in FIG5 , there are a number of particularly useful sdABDs that bind to human CA9, referred to herein as “sdABD-CA9” or “CA9-ABD”.

[0396] In one useful embodiment, sdABD-CA9 hVIB456 has: sdCDR1 having SEQ ID NO: 102, sdCDR2 having SEQ ID NO: 103, sdCDR3 having SEQ ID NO: 104. In some cases, the sdABD-Trop2 has SEQ ID NO: 101.

[0397] In one useful embodiment, sdABD-CA9 hVIB476 has: sdCDR1 having SEQ ID NO: 106, sdCDR2 having SEQ ID NO: 107, sdCDR3 having SEQ ID NO: 108. In some cases, the sdABD-Trop2 has SEQ ID NO: 105.

[0398] In one useful embodiment, sdABD-CA9 hVIB407 has: sdCDR1 having SEQ ID NO: 110, sdCDR2 having SEQ ID NO: 111, sdCDR3 having SEQ ID NO: 112. In some cases, the sdABD-Trop2 has SEQ ID NO:109.

[0399] In one useful embodiment, sdABD-CA9 hVIB445 has: sdCDR1 having SEQ ID NO: 114, sdCDR2 having SEQ ID NO: 115, sdCDR3 having SEQ ID NO: 116. In some cases, the sdABD-Trop2 has SEQ ID NO:113.

[0400] In some embodiments, the protein before the protease cleavage domain cleavage is less than about 100kDa. In some embodiments, the protein after the protease cleavage domain cleavage is about 25 to about 75kDa. In some embodiments, the protein before the protease cleavage has a size higher than the kidney threshold for first-pass clearance. In some embodiments, the protein before the protease cleavage has an elimination half-life of at least about 50 hours. In some embodiments, the protein before the protease cleavage has an elimination half-life of at least about 100 hours. In some embodiments, the protein has increased tissue penetration compared to IgG for the same target antigen. In some embodiments, the protein has increased tissue distribution compared to IgG for the same target antigen.

[0401] C. Half-life extension domain

[0402] The MCE protein of the present invention (again also referred to herein as "COBRA TM The "protein or construct" optionally comprises a half-life extending domain. Such domains are contemplated to include, but are not limited to, HSA binding domains, Fc domains, small molecules, and other half-life extending domains known in the art.

[0403] Human serum albumin (HSA) (molecular weight approximately 67 kDa) is the most abundant protein in plasma, present at approximately 50 mg / ml (600 uM) and having a half-life of approximately 20 days in humans. HSA is used to maintain plasma pH, promote colloidal blood pressure, act as a carrier for many metabolites and fatty acids, and serves as the major drug transporter in plasma.

[0404] Non-covalent association with albumin prolongs the elimination half-life of short-lived proteins. For example, recombinant fusions of albumin-binding domains with Fab fragments reduced in vivo clearance by 25- and 58-fold and prolonged half-life by 26- and 37-fold when administered intravenously to mice and rabbits, respectively, compared to administration of the Fab fragment alone. In another example, when insulin was acylated with fatty acids to promote association with albumin, a prolonged effect was observed when injected subcutaneously in rabbits or pigs. Together, these studies demonstrate a link between albumin binding and prolonged effect.

[0405] In one aspect, the antigen binding proteins described herein comprise a half-life extension domain, such as a domain that specifically binds to HSA. In other embodiments, the HSA binding domain is a peptide. In other embodiments, the HSA binding domain is a small molecule. It is contemplated that in some embodiments, the HSA binding domain of the antigen binding protein is relatively small and does not exceed 25 kD, does not exceed 20 kD, does not exceed 15 kD, or does not exceed 10 kD. In some cases, if the HSA binding domain is a peptide or small molecule, it is 5 kD or less.

[0406] In many embodiments, the half-life extension domain is a single domain antigen binding domain from a single domain antibody that binds to HSA. This domain is generally referred to herein as an "sdABD" for human HSA (sdABD-HSA), or alternatively, "sdABD(1 / 2)" to distinguish these binding domains from sdABDs for TTA. A particularly useful sdABD(1 / 2) is shown in FIG5 .

[0407] The half-life extension domain of the antigen-binding protein provides the pharmacodynamics and pharmacokinetics of the change of the antigen-binding protein itself. As mentioned above, the half-life extension domain extends the elimination half-life. The half-life extension domain also changes the pharmacodynamic properties, including changing the tissue distribution, penetration and diffusion of the antigen-binding protein. In some embodiments, compared with the protein without the half-life extension binding domain, the half-life extension domain provides improved tissue (including tumor) targeting, tissue penetration, tissue distribution, diffusion within the tissue and enhanced efficacy. In one embodiment, the therapeutic method effectively and efficiently utilizes the antigen-binding proteins of reduced amount, thereby producing the side effect of reduction, such as the non-tumor cell cytotoxicity of reduction.

[0408] In addition, characteristics of a half-life extension domain, such as an HSA binding domain, include the binding affinity of the HSA binding domain for HSA. The affinity of the HSA binding domain can be selected to target a specific elimination half-life in a particular polypeptide construct. Thus, in some embodiments, the HSA binding domain has a high binding affinity. In other embodiments, the HSA binding domain has an intermediate binding affinity. In other embodiments, the HSA binding domain has a low or marginal binding affinity. Exemplary binding affinities include KD concentrations of 10 nM or less (high), between 10 nM and 100 nM (intermediate), and greater than 100 nM (low). As described above, binding affinity for HSA is determined by known methods such as surface plasmon resonance (SPR).

[0409] D. Protease cleavage site

[0410] As outlined herein, the protein compositions, and in particular the prodrug constructs of the invention comprise one or more protease cleavage sites, typically located in a cleavable linker.

[0411] As described herein, the prodrug constructs of the present invention comprise at least one protease cleavage site comprising an amino acid sequence that is cleaved by at least one protease. In some cases, the MCE proteins described herein comprise 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20 or more protease cleavage sites that are cleaved by at least one protease. As discussed more fully herein, when more than one protease cleavage site is used in a prodrug construction, the protease cleavage sites can be the same (e.g., multiple sites cleaved by a single protease) or different (two or more cleavage sites cleaved by at least two different proteases). As will be understood by those skilled in the art, constructs containing three or more protease cleavage sites can utilize one, two, three, etc.; for example, some constructs can utilize three sites for two different proteases, etc.

[0412] The amino acid sequence of the protease cleavage site will depend on the protease being targeted.As is known in the art, there are many human proteases that are found in vivo and that can be associated with disease states.

[0413] Known proteases are secreted by some diseased cells and tissues, such as tumors or cancer cells, to produce a microenvironment rich in proteases (microenvironment that is rich in proteases) or a microenvironment rich in proteases (protease-rich microenvironment). In some cases, the blood of the experimenter is rich in proteases. In some cases, the cells around the tumor secrete proteases into the tumor microenvironment. The cells around the tumor that secrete proteases include but are not limited to tumor stroma cells, myofibroblasts, hemocytes, mast cells, B cells, NK cells, regulatory T cells, macrophages, cytotoxic T lymphocytes, dendritic cells, mesenchymal stem cells, polymorphonuclear cells and other cells. In some cases, proteases are present in the blood of the experimenter, such as the protease of the amino acid sequence found in the targeted microbial peptide. This feature allows targeted therapeutic agents such as antigen-binding proteins to have other specificity, because except in the microenvironment rich in proteases in the targeted cells or tissues, T cells will not be combined by antigen-binding proteins.

[0414] Proteases are proteins that cleave proteins in a sequence-specific manner in some cases. Proteases include, but are not limited to, serine proteases, cysteine proteases, aspartic proteases, threonine proteases, glutamic proteases, metalloproteases, asparagine peptide lysing enzymes, serum proteases, cathepsins (e.g., cathepsin B, cathepsin C, cathepsin D, cathepsin E, cathepsin K, cathepsin L, cathepsin S), kallikrein, hK1, hK10, hK15, KLK7, granzyme B, plasmin, collagenase, type IV collagenase, stromelysin, factor XA, chymotrypsin-like protease, trypsin-like protease, elastase-like protease, subtilisin-like protease, actinidin, bromelain, calpain, caspases (e.g., caspase-3), Mir1-C P, papain, HIV-1 protease, HSV protease, CMV protease, chymosin, renin, pepsin, malic acid, leguminase, plasmepsin, nepenthesin, metalloexopeptidases, metalloendopeptidases, matrix metalloproteinases (MMPs), MMP1, MMP2, MMP3, MMP8, MMP9, MMP13, MMP11, MMP14, Meprin, urokinase plasminogen activator (uPA), enterokinase, prostate-specific antigen (PSA, hK3), interleukin-1β converting enzyme, thrombin, FAP (FAP-α), dipeptidyl peptidase, and dipeptidyl peptidase IV (DPPIV / CD26).

[0415] Some suitable proteases and protease cleavage sequences are shown in Figures 5 and 6.

[0416] E.Connector

[0417] As discussed herein, the different domains of the present invention are typically linked together using amino acid linkers, which can also confer functionality, including flexibility or rigidity (e.g., steric constraint) and the ability to be cleaved using in situ proteases. These linkers can be classified in a variety of ways.

[0418] The present invention provides "domain linkers" that are used to link two or more domains (e.g., VH and VL, target tumor antigen binding domain (TTABD, sometimes also referred to herein as "αTTA") (for "anti-TTA") to VH or VL, linking a half-life extension domain to another component, etc. Domain linkers can be non-cleavable (NCL), cleavable ("CL"), constrained and cleavable (CCL), and constrained and non-cleavable (CNCL), for example.

[0419] 1. Non-cleavable linker

[0420] In some embodiments, the domain linker is non-cleavable. Generally speaking, these can be one of two types: non-cleavable and flexible, thereby allowing the components "upstream" and "downstream" of the linker in the construct to self-assemble in some manner intramolecularly; or non-cleavable and constrained, in which the two components separated by the linker cannot self-assemble intramolecularly. However, it should be noted that in the latter case, although the two component domains separated by the non-cleavable constrained linker do not self-assemble intramolecularly, the other intramolecular components will self-assemble to form a pseudo-Fv domain.

[0421] (i) Non-cleavable but flexible linker

[0422] In this embodiment, linkers are used to connect domains, typically through longer flexible domains that are not cleaved by in situ proteases in the patient's body to preserve the functionality of the domains. Examples of internal, non-cleavable linkers suitable for connecting domains in the polypeptides of the present invention include, but are not limited to, (GS)n, (GGS)n, (GGGS)n, (GGSG)n, (GGSGG)n, or (GGGGS)n, where n is 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10. In some embodiments, the linker can be about 15 amino acids in length.

[0423] (ii) Non-cleavable and constrained linker

[0424] In some cases, the linker does not contain a cleavage site and is too short to allow intramolecular self-assembly of the protein domains separated by the linker and is a "constrained non-cleavable linker" or "CNCL". For example, in Pro186, the active VH and active VL are separated by 8 amino acids ("8-mer"), which does not allow VH and VL to self-assemble into active antigen-binding domains. In some embodiments, the linker is still flexible; for example, (GGGS)n, where n=2. In other embodiments, although generally less preferred, more rigid linkers such as those comprising proline or bulky amino acids can be used.

[0425] 2. Cleavable linker

[0426] All prodrug constructs herein comprise at least one cleavable linker. Thus, in one embodiment, the domain linker is cleavable (CL), sometimes referred to herein as a "protease cleavage domain" ("PCD"). In this embodiment, the CL contains a protease cleavage site, as outlined herein and shown in Figures 5 and 6. In some cases, the CL contains only a protease cleavage site. Optionally, depending on the length of the cleavage recognition site, there may be additional connecting amino acids at one or both of the N-terminus or C-terminus of the CL. For example, there may be 1, 2, 3, 4, or 5 amino acids at one or both of the N-terminus and C-terminus of the cleavage site. Thus, the cleavable linker may also be constrained (e.g., 8-mer) or flexible.

[0427] Of particular interest in the present invention are MMP9-cleavable linkers and Meprin-cleavable linkers, particularly MMP9-constrained cleavable linkers and Meprin-constrained cleavable linkers.

[0428] II. Domains of the Invention

[0429] The present invention provides a variety of different forms of the prodrug polypeptides of the present invention. The present invention provides constrained Fv domains and constrained pseudo-Fv domains. In addition, the present invention provides multivalent conditionally effective ("MCE") proteins containing two Fv domains, but as non-isomerizable constructs. As outlined herein, although each construct contains at least one protease cleavage domain, these can be non-isomerizable cleavable forms or non-isomerizable non-cleavable forms.

[0430] Importantly, although both domains (Fv domain and pseudo-Fv domain) are referred to herein as "constrained," this means that the Figure 36 、 Figure 37 and Figure 38As shown in , only one of these needs to be constrained, although generally, the protein has better expression when both linkers are constrained.

[0431] Those skilled in the art will appreciate that for formats 1, 2, and 4, there are four possibilities for the N-terminal to C-terminal order of the constrained and pseudo-Fv domains of the present invention (linkers not shown): aVH-aVL and iVL-iVH, aVH-aVL and iVH-iVL, aVL-aVH and iVL-iVH, aVL-aVH and iVL-iVH, aVL-aVH and iVH-iVL. All four have been tested, and all four are active, but the first order aVH-aVL and iVL-iVH showed better expression than the other three. Therefore, although the description herein is generally shown in this aVH-aVL and iVL-iVH format, all disclosures herein also include other orders of these domains.

[0432] Note that, in general, the N-terminal to C-terminal order of the full-length constructs of the invention is based on the aVH-aVL and iVL-iVH orientations.

[0433] In addition, it is known in the art that C-terminal sequences derived from certain ABDs may be immunogenic in humans. Therefore, in general, a histidine tag (His6 or His10) may be used, particularly when the C-terminus of the construct terminates in an sdABD (e.g., the sdABD-HSA domain of many constructs). For purification reasons, many or most sequences herein were generated using a His6 C-terminal tag, but these sequences can also be used to reduce immunogenicity in humans, as shown in Holland et al., DOI 10.1007 / s10875-013-9915-0 and WO 2013 / 024059.

[0434] A. Constrained Fv Domain

[0435] The present invention provides constrained Fv domains comprising an active VH and an active VL domain covalently attached using a constrained linker (which, as outlined herein, can be cleavable (Format 1) or non-cleavable (Formats 2 and 4)). The constrained linker prevents intramolecular association between aVH and aVL in the absence of cleavage. Thus, a constrained Fv domain generally comprises a set of six CDRs contained within the variable domain, wherein vhCDR1, vhCDR2, and vhCDR3 of VH bind to human CD-3, and vlCDR1, vCDR2, and vlCDR3 of VL bind to human CD-3, but in a prodrug form (e.g., uncleaved), VH and VL cannot spatially associate to form an active binding domain, but rather prefer to pair intramolecularly with a pseudo-Fv.

[0436] As described herein, a constrained Fv domain may comprise an active VH and an active VL (aVH and aVL) or an inactive VH and VL (iVH and iVL, in which case they are constrained pseudo-Fv domains), or a combination thereof.

[0437] As will be understood by those skilled in the art, in a constrained Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0438] As outlined herein, for Format 1 constructs, the constrained Fv domain may comprise a VH and VL connected using a cleavable linker, in cases such as those shown in Figures 5 and 6. In this embodiment, the constrained Fv domain has the structure (N-terminus to C-terminus) vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCDR3-vhFR4-CCL-vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4. In general, a constrained Fv domain comprises active VH and VL domains (e.g., capable of binding CD3 when associated), and thus has the structure (N-terminus to C-terminus) vhFR1-avhCDR1-vhFR2-avhCDR2-vhFR3-avhCDR3-vhFR4-CCL-vlFR1-avlCDR1-vlFR2-avlCDR2-vlFR3-avlCDR3-vlFR4.

[0439] As outlined herein, for Format 2 constructs, the constrained Fv domain can comprise a VH and a VL connected using a non-cleavable linker. In this embodiment, the constrained Fv domain has the structure (N-terminus to C-terminus) vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCDR3-vhFR4-CNCL-vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4. In general, a constrained Fv domain comprises active VH and VL domains (e.g., capable of binding CD3 when associated), and thus has the structure (N-terminus to C-terminus) vhFR1-avhCDR1-vhFR2-avhCDR2-vhFR3-avhCDR3-vhFR4-CNCL-vlFR1-avlCDR1-vlFR2-avlCDR2-vlFR3-avlCDR3-vlFR4.

[0440] Particularly useful in the present invention is a constrained, non-cleavable Fv domain having: an aVH having SEQ ID NO: 142, an aVL having SEQ ID NO: 126, and a domain linker having SEQ ID NO: 233.

[0441] B. Constrained pseudo-Fv domain

[0442] The present invention provides constrained pseudo-Fv domains comprising inactive or pseudo iVH and iVL domains covalently attached using constrained linkers (as outlined herein, the constrained linkers may be cleavable or non-cleavable). The constrained linker prevents intramolecular association between iVH and iVL in the absence of cleavage. Thus, constrained pseudo-Fv domains typically comprise iVH and iVL, with a framework region (when in a non-constrained form) that allows iVH and iVL to associate, but the resulting pseudo-Fv domains do not bind to human proteins. The iVH domain can be assembled with the aVL domain, and the iVL domain can be assembled with the aVH domain, but the resulting structure does not bind to CD3.

[0443] The constrained pseudo-Fv domains comprise inactive VH and VL (iVH and iVL).

[0444] As will be understood by those skilled in the art, in a constrained pseudo-Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0445] As outlined herein, a constrained pseudo-Fv domain can comprise an iVH and an iVL connected using a non-cleavable linker (as shown in formats 1, 2, and 4) or with a cleavable linker (as shown in format 3).

[0446] In general, a constrained Fv domain comprises inert VH and VL domains (e.g., capable of binding CD3 when associated), and thus has the structure (N-terminus to C-terminus) vhFR1-ivlCDR1-vhFR2-ivlCDR2-vhFR3-ivlCDR3-vhFR4-CNCL-vlFR1-ivhCDR1-vlFR2-ivhCDR2-vlFR3-ivhCDR3-vlFR4.

[0447] Particularly useful in the present invention are constrained, non-cleavable pseudo-Fv domains having: an iVH having SEQ ID NO: 146, SEQ ID NO: 150, or SEQ ID NO: 154, an iVL having SEQ ID NO: 130, SEQ ID NO: 134, or SEQ ID NO: 138, and a domain linker having SEQ ID NO: 233.

[0448] III. Forms of the Invention

[0449] As discussed herein, the prodrug constructs of the present invention can take a variety of different forms, including cleavable forms having dual TTA binding domains, non-cleavable forms having dual TTA binding domains (either of which can have the same TTA binding domain or different binding domains), and non-cleavable forms having a single targeting domain.

[0450] A. Cleavable form with dual targeting

[0451] The present invention Figure 1 A non-isomerizable cleavable form of the "Format 1" type is provided. In this embodiment, the constrained Fv domain comprises VH and VL domains connected using a constrained cleavable linker, and the constrained pseudo-Fv domain uses a constrained non-cleavable linker. For ease of discussion, both will be referred to herein as "constrained," but as discussed above and in Figure 37 、 Figure 38 and Figure 39 As shown in , only one of these needs to be constrained, although generally, the protein has better expression when both linkers are constrained.

[0452] All constructs of version 1 (and other versions) also have a cleavable linker (CL) that is cleaved by human tumor proteases.

[0453] The present invention provides a prodrug protein comprising, from N-terminus to C-terminus, (sdABD-TTA1)-domain linker-constrained Fv domain-domain linker-(sdABD-TTA2)-CL-constrained pseudo-Fv domain-domain linker-sdABD-HSA.

[0454] As will be understood by those skilled in the art, in a constrained Fv domain or constrained pseudo-Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0455] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA.

[0456] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CCL-aVL-domain linker-(sdABD-TTA2)-CL-iVH-CCL-iVL-domain linker-sdABD-HSA.

[0457] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVL-CCL-aVH-domain linker-(sdABD-TTA2)-CL-iVL-CCL-iVH-domain linker-sdABD-HSA.

[0458] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVL-CCL-aVH-domain linker-(sdABD-TTA2)-CL-iVH-CCL-iVL-domain linker-sdABD-HSA.

[0459] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-NCL-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH and iVL have the sequences shown in Figure 5.

[0460] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains are bound to the same TTA, which can be EGFR, EpCAM, FOLR1, Trop2, CA9 or B7H3, the sequence of which is depicted in Figure 5.

[0461] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains are bound to different TTAs.

[0462] In some embodiments, the prodrug construct comprises sdABD (TTA1) -domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2) -CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EGFR and EpCAM, and sdABD-TTA has the sequence in Figure 5.

[0463] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EGFR and FOLR1, and sdABD-TTA has the sequence in Figure 5.

[0464] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EGFR and B7H3, and sdABD-TTA has the sequence in Figure 5.

[0465] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EpCAM and FOLR1, and sdABD-TTA has the sequence in Figure 5.

[0466] In some embodiments, the prodrug construct comprises sdABD (TTA1) -domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2) -CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EpCAM and B7H3, and sdABD-TTA has the sequence in Figure 5.

[0467] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to B7H3 and FOLR1, and sdABD-TTA has the sequence in Figure 5.

[0468] In some embodiments, the prodrug construct comprises sdABD (TTA1)-domain linker-aVH-CCL-aVL-domain linker-sdABD (TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD (1 / 2). In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains are bound to the same TTA, which can be EGFR, FOLR1, B7H3, Trop2, CA9 or EpCAM, the sequence of which is depicted in Figure 5, and CCL and CL are selected from linkers cleaved by MMP9 or meprin, and sdABD (1 / 2) has SEQ ID NO: 117 or SEQ ID NO: 121.

[0469] In Format 1, the preferred domain linker is SEQ ID NO: 233 (which also serves as the preferred constrained non-cleavable linker).

[0470] In Version 1, the preferred constructs are Pro140 and Pro140b.

[0471] B. Indestructible form

[0472] like Figure 2As shown in , the present invention provides a non-isomerizable, non-cleavable form. In this embodiment, it should be understood that "non-cleavable" only applies to the linkage of the constrained Fv domain because there is an activation cleavage site in the prodrug construct. In this embodiment, the constrained Fv domain comprises VH and VL domains connected using a constrained, non-cleavable linker, and the constrained pseudo-Fv domain uses a constrained, non-cleavable linker.

[0473] As will be understood by those skilled in the art, in a constrained Fv domain or constrained pseudo-Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0474] The present invention provides a prodrug protein comprising, from N-terminus to C-terminus, sdABD (TTA1)-domain linker-constrained Fv domain-domain linker-sdABD (TTA2)-cleavable linker-constrained pseudo-Fv domain-domain linker-sdABD-HSA.

[0475] As will be understood by those skilled in the art, in a constrained Fv domain or constrained pseudo-Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0476] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA.

[0477] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVH-CNCL-iVL-domain linker-sdABD-HSA.

[0478] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVL-CNCL-aVH-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA.

[0479] Thus, in one embodiment, the prodrug protein comprises from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVL-CNCL-aVH-domain linker-(sdABD-TTA2)-CL-iVH-CNCL-iVL-domain linker-sdABD-HSA.

[0480] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains are bound to the same TTA, which can be EGFR, EpCAM, FOLR1, Trop2, CA9, or B7H3, the sequence of which is depicted in Figure 5.

[0481] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to different TTAs.

[0482] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EGFR and EpCAM, and sdABD-TTA has the sequence in Figure 5. In this embodiment, preferred combinations of EGFR and EpCAM include:

[0483] cross EGFR2 EGFR2a EGFR2d EpCAM h13 In any orientation In any orientation In any orientation EpCAM h23 In any orientation In any orientation In any orientation EpCAM hVIB665 In any orientation In any orientation In any orientation EpCAM hVIB666 In any orientation In any orientation In any orientation

[0484] In this context, "either orientation" means that the EpCAM sdABD is located either N-terminally or C-terminally to the EGFR sdABD in the construct of the invention.

[0485] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EGFR and FOLR1, and sdABD-TTA has the sequence in Figure 5. In this embodiment, preferred combinations of EGFR and FOLR1 include:

[0486] cross EGFR2 EGFR2a EGFR2d FOLR1h77-2 In any orientation In any orientation In any orientation FOLR1 h59.3 In any orientation In any orientation In any orientation FOLR h22-4 In any orientation In any orientation In any orientation

[0487] In this context, "either orientation" means that the FOLR1 sdABD is located at the N-terminus of the EGFR sdABD or at its C-terminus in the construct of the invention.

[0488] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EGFR and B7H3, and sdABD-TTA has the sequence in Figure 5. In this embodiment, preferred combinations of EGFR and B7H3 include:

[0489] cross EGFR2 EGFR2a EGFR2d B7H3 hF7 In any orientation In any orientation In any orientation B7H3 hF12 In any orientation In any orientation In any orientation B7H3 hF12(N57Q) In any orientation In any orientation In any orientation B7H3 hF12(N57E) In any orientation In any orientation In any orientation B7H3 hF12(N57D) In any orientation In any orientation In any orientation B7H3 hF12(S59A) In any orientation In any orientation In any orientation B7H3 hF12(S59Y) In any orientation In any orientation In any orientation

[0490] In this context, "either orientation" means that the B7H3 sdABD is located at the N-terminus or the C-terminus of the EGFR sdABD in the constructs of the invention.

[0491] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EpCAM and FOLR1, and sdABD-TTA has the sequence in Figure 5. In this embodiment, preferred combinations of EpCAM and FOLR1 include:

[0492] cross FOLR1h77-2 FOLR1 h59.3 FOLR h22-4 EpCAM h13 In any orientation In any orientation In any orientation EpCAM h23 In any orientation In any orientation In any orientation EpCAM hVIB665 In any orientation In any orientation In any orientation EpCAM hVIB666 In any orientation In any orientation In any orientation

[0493] In this context, "either orientation" means that the EpCAM sdABD is located at the N-terminus or the C-terminus of the FOLR1 sdABD in the construct of the invention.

[0494] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to EpCAM and B7H3, and sdABD-TTA has the sequence in Figure 5. In this embodiment, preferred combinations of EpCAM and B7H3 include:

[0495]

[0496] In this context, "either orientation" means that the B7H3 sdABD is located at the N-terminus or the C-terminus of the EGFR sdABD in the constructs of the invention.

[0497] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains bind to FOLR1 and B7H3, and sdABD-TTA has the sequence in Figure 5. In this embodiment, preferred combinations of FOLR1 and B7H3 include:

[0498] cross FOLR1h77-2 FOLR1 h59.3 FOLR h22-4 B7H3 hF7 In any orientation In any orientation In any orientation B7H3 hF12 In any orientation In any orientation In any orientation B7H3 hF12(N57Q) In any orientation In any orientation In any orientation B7H3 hF12(N57E) In any orientation In any orientation In any orientation B7H3 hF12(N57D) In any orientation In any orientation In any orientation B7H3 hF12(S59A) In any orientation In any orientation In any orientation B7H3 hF12(S59Y) In any orientation In any orientation In any orientation

[0499] In this context, "either orientation" means that the B7H3 sdABD is located at the N-terminus or the C-terminus of the FOLR1 sdABD in the constructs of the invention.

[0500] In some embodiments, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA1)-domain linker-aVH-CNCL-aVL-domain linker-(sdABD-TTA2)-CL-iVL-CNCL-iVH-domain linker-sdABD-HSA. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the two targeting domains are bound to the same TTA, which can be EGFR, FOLR1, B7H3, CA9, Trop2, or EpCAM, the sequence of which is depicted in Figure 5, and CCL and CL are selected from linkers cleaved by MMP9 or meprin, and sdABD (1 / 2) has SEQ ID NO: 117.

[0501] In Format 2, the preferred domain linker is SEQ ID NO: 233 (which also serves as the preferred constrained non-cleavable linker).

[0502] In Format 2, preferred dual targeting constructs (sometimes referred to herein as "iso-COBRA") include targeting EGFR and EpCAM, EGFR and Trop2, EGFR and FOLR1, EGF and B7H3, EpCAM and Trop2, EpCAM and FOLR1, EpCAM and B7H3, Trop2 and FOLR1, Trop2 and B7H3, and combinations of FOLR1 and B7H3, as described more fully below.

[0503] In Form 2, specific embodiments of use include, but are not limited to, Pro186, Pro225, Pro226, Pro233, Pro262, Pro311, Pro312, Pro313, Pro356, Pro359, Pro364, Pro388, Pro448, Pro449, Pro450, Pro451, Pro495, Pro246, Pro254, Pro255, Pro256, Pro420, Pro421, Pro432, Pro479, Pro480, Pro187, Pro221, Pro222, Pro223, Pro224, Pro393, Pro394, Pro395, Pro396, Pro429, Pro430, Pro431, Pro601, Pro602, V3 and V4, Pro664, Pro665, 6, Pro551, Pro552, Pro623, Pro624, Pro698, Pro655, Pro656, Pro657, Pro658, Pro516, Pro517, Pro518, and Pro519.

[0504] C. Single TTA construct

[0505] like Figure 4 Also included in the compositions of the invention are "Form 4" constructs, as shown in Figure 2, which are similar to Form 2 constructs but lack the second TTA ABD. In this embodiment, it is understood that "non-cleavable" applies only to the linkage of the constrained Fv domain, as an activation cleavage site is present in the prodrug construct. In this embodiment, the constrained Fv domain comprises VH and VL domains connected using a constrained non-cleavable linker, and the constrained pseudo-Fv domain uses a constrained non-cleavable linker.

[0506] As will be understood by those skilled in the art, in a constrained Fv domain or constrained pseudo-Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0507] The present invention provides a prodrug protein comprising, from the N-terminus to the C-terminus, an sdABD (TTA)-domain linker-constrained Fv domain-cleavable linker-sdABD-HSA-constrained pseudo-Fv domain. (Note that for all constructs of this format, the sdABD-HSA typically lacks a His6 tag, but may contain one).

[0508] As will be understood by those skilled in the art, in a constrained Fv domain or constrained pseudo-Fv domain, the order of VH and VL can be (N-terminus to C-terminus) VH-linker-VL or VL-linker-VH.

[0509] Thus, in one embodiment, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA)-domain linker-aVH-CNCL-aVL-CL-(sdABD-HSA)-domain linker-iVL-CNCL-iVH.

[0510] Thus, in one embodiment, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA)-domain linker-aVH-CNCL-aVL-CL-(sdABD-HSA)-domain linker-iVH-CNCL-iVL.

[0511] Thus, in one embodiment, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA)-domain linker-aVL-CNCL-aVH-CL-(sdABD-HSA)-domain linker-iVH-CNCL-iVL.

[0512] Thus, in one embodiment, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA)-domain linker-aVL-CNCL-aVH-CL-(sdABD-HSA)-domain linker-iVL-CNCL-iVH.

[0513] Thus, in one embodiment, the prodrug protein comprises, from N-terminus to C-terminus: (sdABD-TTA)-domain linker-aVL-CNCL-aVH-CL-(sdABD-HSA)-domain linker-iVL-CNCL-iVH. In this embodiment, aVH, aVL, iVH, iVL have the sequences shown in Figure 5. In this embodiment, the targeting domain is bound to TTA, which can be EGFR, EpCAM, FOLR1, Trop2, CA9, or B7H3, the sequence of which is depicted in Figure 5.

[0514] In Format 4, the preferred domain linker is SEQ ID NO: 233 (which also serves as the preferred constrained non-cleavable linker).

[0515] In form 4, the preferred sdABD-HSA is the sdABD-HSA of SEQ ID NO: 121 or 117.

[0516] D. Two protein compositions

[0517] In some embodiments, the compositions of the invention comprise two different molecules, sometimes referred to as "half-COBRA TM In the absence of a cleavage site, the two distinct molecules associate intramolecularly to form a pseudo-Fv. In the presence of a protease, the cleavage site is cleaved, releasing the inert variable domain, and the protein pair then forms an active antigen-binding domain to CD3, as generally depicted in FIG3 .

[0518] It is important in the design of the half-construct that the active variable domain and sdABD-TTA remain together after cleavage so that the two cleaved parts are held together by tumor antigen receptors on the tumor surface and can then form an active anti-CD3 binding domain.

[0519] There are two different general form 3 constructs, those in which each member of the pair has a single sdABD-TTA ( Figure 3A ) and those with two different sdABD-TTAs, each targeting a different TTA ( Figure 3B ).

[0520] 1. Half-COBRA with a single TTA-binding domain TM Construct (Format 3A)

[0521] In some embodiments, the first half-COBRA TM From the N-terminus to the C-terminus, there is a sdABD (TTA1) domain linker-aVH-CL-iVL domain linker-sdABD (1 / 2), and the second has sdABD (1 / 2)-domain linker-iVH-CL-aVL-domain linker-sdABD (TTA2). In this embodiment, aVH, aVL, iVH, iVL and sdABD (1 / 2) have the sequences shown in Figure 5, and sdABD-TTAa binds to human EGFR, EpCAM, Trop2, CA9 FOLR1 and / or B7H3 and has the sequence depicted in Figure 5.

[0522] 2. Semi-COBRA with double TTA ABD TM Construct

[0523] In some embodiments, the paired prodrug constructs may have two sdABD-TTA binding domains per construct, such as Figure 3B In this embodiment, the first member of the pair comprises, from N-terminus to C-terminus, sdABD-TTA1-domain linker-sdABD-TTA2-domain linker-aVH-CL-iVL-domain linker-sdABD(HAS), and the second member comprises, from N-terminus to C-terminus, sdABD-TTA1-domain linker-sdABD-TTA2-aVL-CL-iVH-domain linker-sdABD-HSA.

[0524] The two sdABD-TTAs on each member of the pair are different, but usually both members (half-COBRA TM ) all have the same two sdABD-TTAs, for example, all have EGFR and FOLR1 or EGFR and B7H3, etc.

[0525] In some embodiments, the two sdABD-TTAs are selected from the sdABD-TTAs shown in FIG. 5 .

[0526] IV. Methods of Preparing the Compositions of the Invention

[0527] The prodrug compositions of the present invention are prepared as generally understood by those skilled in the art and as outlined below.

[0528] The invention provides nucleic acid compositions encoding the prodrug compositions of the present invention. As will be appreciated by those skilled in the art, the nucleic acid compositions will depend on the form of the prodrug polypeptide. Therefore, for example, when the form requires two amino acid sequences, such as a "form 3" construct, two nucleic acid sequences can be incorporated into one or more expression vectors for expression. Similarly, a prodrug construct as a single polypeptide (form 1, 2, and 4) requires a single nucleic acid in a single expression vector to produce.

[0529] As known in the art, the nucleic acid encoding the components of the present invention can be incorporated into expression vectors as known in the art, and depends on the host cell for producing the prodrug composition of the present invention. Generally speaking, nucleic acid can be operably linked to any number of regulatory elements (promoter, replication origin, selective marker, ribosome binding site, inducer, etc.). The expression vector can be extrachromosomal or integrating vector.

[0530] The nucleic acids and / or expression vectors of the invention are then transformed into any number of different types of host cells well known in the art, including mammalian, bacterial, yeast, insect, and / or fungal cells, with mammalian cells (e.g., CHO cells, 293 cells) being useful in many embodiments.

[0531] As is well known in the art, the prodrug compositions of the present invention are prepared by culturing host cells containing the expression vector.Once produced, conventional antibody purification steps are performed, including protein A affinity chromatography and / or ion exchange chromatography.

[0532] V. Formulation and Administration of Prodrug Compositions of the Invention

[0533] Formulations of the prodrug compositions used in accordance with the present invention are prepared for storage by mixing the prodrug (a single protein in the case of Forms 1, 2, and 4, and two proteins in the case of Form 3) having the desired degree of purity with optional pharmaceutically acceptable carriers, excipients, or stabilizers (as generally summarized in Remington's Pharmaceutical Sciences 16th edition, Osol, A. Ed.

[1980] ), in the form of a lyophilized formulation or an aqueous solution.

[0534] The prodrug compositions of the present invention are administered to a subject according to known methods, such as intravenous administration as a bolus or by continuous infusion over a period of time.

[0535] The prodrug compositions of the present invention can be used to treat cancer.

[0536] Example

[0537] A. Example 1: Construction and purification of Pro constructs

[0538] Transfection

[0539] From independent expression vector (pcdna3.4 derivative) express every kind of protein (for example, form 1,2 and 4 single protein) or paired construct (form 3).According to the transfection scheme of manufacturer, the plasmid DNA of equal amount of coding a pair of half-cobra or single-chain construct is mixed and transfected into Expi293 cells.5 days after transfection, by centrifugation (6000rpm x 25 ') and filtration (0.2uM filter) harvest conditioned medium.Confirm protein expression by SDS-PAGE.Purification construct, and final buffer composition is: 25mM citrate, 75mM arginine, 75mM NaCl, 4% sucrose, pH 7.Final preparation is stored at-80 ℃.

[0540] Activation of MMP9

[0541] Recombinant human (rh) MMP9 was activated according to the following protocol. Recombinant human MMP-9 (R&D #911-MP-010) was at 0.44 mg / ml (4.7 uM). P-aminophenylmercuric acetate (APMA) (Sigma) was prepared in DMSO at a stock concentration of 100 mM. The assay buffer was 50 mM Tris pH 7.5, 10 mM CaCl2, 150 mM NaCl, and 0.05% Brij-35.

[0542] - Dilute rhMMP9 to approximately 100ug / ml with assay buffer (25ul hMMP9 + 75uL assay buffer)

[0543] - Add p-aminophenylmercuric acetate (APMA) to a final concentration of 1 mM from a 100 mM stock solution in DMSO (1 uL to 100 uL)

[0544] - Incubate at 37°C for 24 hours

[0545] - Dilute MMP9 to 10 ng / ul (add 900 ul assay buffer to 100 ul activation solution)

[0546] The concentration of activated rhMMP9 was approximately 100 nM.

[0547] Cleavage of constructs for TDCC assay

[0548] To cleave the construct, add CaCl2 to 10 mM of a 100 μl protein sample at a concentration of 1 mg / ml (10.5 μM) in formulation buffer (25 mM citric acid, 75 mM L-arginine, 75 mM NaCl, 4% sucrose). Add activated rhMMP9 to a concentration of 20-35 nM. Incubate the sample overnight (16-20 hours) at room temperature. Verify the integrity of the cleavage using SDS PAGE (10%-20% TG, TG running buffer, 200 v, 1 hour). Samples are typically 98% lysed.

[0549] B. Example 2: T cell-dependent cellular cytotoxicity (TDCC) assay

[0550] HT-29 cells transduced with firefly luciferase were grown to approximately 80% confluence and detached with Versene (0.48 mM EDTA–Ca-Mg in PBS). The cells were centrifuged and resuspended in TDCC medium (5% heat-inactivated FBS in RPMI 1640 supplemented with HEPES, GlutaMax, sodium pyruvate, non-essential amino acids, and β-mercaptoethanol). Purified human P pan-T cells were thawed, centrifuged, and resuspended in TDCC medium.

[0551] Co-cultures of HT-29_Luc cells and T cells were added to 384-well cell culture plates. Serial dilutions of COBRA were then added to the co-cultures and incubated at 37°C for 48 hours. Finally, an equal volume of SteadyGlo luciferase assay reagent was added to the plate and incubated for 20 minutes. Plates were read on Perkin Elmer Envision with an exposure time of 0.1s / well. Total fluorescence was read and data were analyzed on GraphPad Prism 7 or version 8.3.1 (depending on timing).

[0552] C. Example 3: General Design of an In Vivo Adoptive T Cell Transfer Efficacy Model

[0553] These protocols were used for many of the experiments in the Figures. Tumor cells were implanted subcutaneously (SC) into the right flank of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog number 005557) and allowed to grow until they reached an average volume of approximately 200 mm 3 Established tumors. Human T cells were cultured in parallel in T cell culture medium (X-VIVO 15 [Lonza, catalog number 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01mM 2-mercaptoethanol) in G-Rex100M breathable flasks (Wilson Wolf catalog number 81100S) containing MACSiBead from a T cell activation / expansion kit (Miltenyi catalog number 130-091-441) for about 10 days and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion of mice were coordinated so that on day 0 of the study, mice were randomized into groups (N=6) based on tumor size; each was then injected intravenously (IV) with 2.5x10 6 Cultured human T cells were administered with a first dose of COBRA or a control molecule. Mice were given seven doses every three days (days 0, 3, 6, 9, 12, 15, and 18) and then followed for another 2-3 weeks until tumor volumes reached >2000 mm 3 Or terminate the study. Tumor volume was measured every 3 days.

[0554] D. Example 4: In vivo activity of EGFR / MMP9 semi-COBRA on Pro77 and Pro53.

[0555] 5x 10 6 LoVo cells or 5 x 10 6HT29 cells were subcutaneously implanted into the right flank of NSG (NOD.Cg-PrkdcscidIl2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog number 005557) and allowed to grow until tumors formed. Parallel human T cells were cultured for 10 days in T cell culture medium (X-VIVO15 [Lonza, catalog number 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M breathable flasks (Wilson Wolf catalog number 81100S) containing MACSiBeads from the T cell activation / expansion kit (Miltenyi catalog number 130-091-441) and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion in mice were coordinated such that on day 0 of the study, mice were randomized into groups (N=6) based on tumor size; each was then injected intravenously (IV) with 2.5×10 6 Cultured human T cells were administered with a first dose of COBRA or a control molecule. Mice were given seven doses every three days (days 0, 3, 6, 9, 12, 15, and 18) and then followed until tumor volume reached >2000 mm 3 Or terminate the study. Each group received 0.2 mg / kg (mpk) of the anti-EGFR x CD3 positive control Pro51 bispecific antibody (bsAb), 0.5 mpk of the negative control anti-hen egg lysozyme (HEL) x CD3 bsAb Pro98, 0.5 mpk each of the anti-EGFR half-COBRA pair Pro77 and Pro53 with an MMP9-cleavable linker, or 0.5 mpk each of the anti-EGFR half-COBRA pair Pro74 and Pro72 with a non-cleavable (NCL) linker. Tumor volume was measured every 3 days.

[0556] E. Example 5: In vivo activity of EGFR / MMP9 COBRA Pro140.

[0557] 5x 10 6 LoVo cells or 5 x 10 6HT29 cells were subcutaneously implanted into the right flank of NSG (NOD.Cg-PrkdcscidIl2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog number 005557) and allowed to grow until tumors formed. Human T cells were cultured in parallel for 10 days in T cell culture medium (X-VIVO15 [Lonza, catalog number 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M breathable flasks (Wilson Wolf catalog number 81100S) containing MACSiBeads from the T cell activation / expansion kit (Miltenyi catalog number 130-091-441) and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion in mice were coordinated such that on day 0 of the study, mice were randomized into groups (N=6) based on tumor size; each was then injected intravenously (IV) with 2.5×10 6 Cultured human T cells were administered with a first dose of COBRA or a control molecule. Mice were given seven doses every three days (days 0, 3, 6, 9, 12, 15, and 18) and then followed until tumor volume reached >2000 mm 3 Each group received 0.2 mpk of the anti-EGFR x CD3 positive control Pro51 bispecific antibody (bsAb), 0.5 mpk of the negative control anti-hen egg lysozyme (HEL) x CD3 bsAb Pro98, or 0.5 mpk of the anti-EGFR COBRA Pro140 with an MMP9-cleavable linker. Tumor volume was measured every 3 days.

[0558] F. Example 6: In vivo activity of EGFR / MMP9 COBRA Pro186.

[0559] 5x 10 6HT29 cells were implanted subcutaneously into the right flank of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog number 005557) and allowed to grow until tumors formed. Human T cells were cultured in parallel for 10 days in T cell culture medium (X-VIVO 15 [Lonza, catalog number 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M breathable flasks (Wilson Wolf catalog number 81100S) containing MACSiBeads from the T cell activation / expansion kit (Miltenyi catalog number 130-091-441) and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion in mice were coordinated such that on day 0 of the study, mice were randomized into groups (N=6) based on tumor size; each was then injected intravenously (IV) with 2.5×10 6 Cultured human T cells were administered with a first dose of COBRA or a control molecule. Mice were given seven doses every three days (days 0, 3, 6, 9, 12, 15, and 18) and then followed until tumor volume reached >2000 mm 3 Or terminate the study. Each group received 0.1 mg / kg (mpk) of the anti-EGFR x CD3 positive control Pro51 bispecific antibody (bsAb), 0.3 mpk of anti-EGFR COBRA Pro214 with a non-cleavable (NCL) control linker, 0.1 or 0.3 mpk of anti-EGFR COBRA Pro140 with an MMP9-cleavable linker, or 0.1 or 0.3 mpk of anti-EGFR COBRA Pro186 with an MMP9-cleavable linker. Tumor volume was measured every 3 days.

[0560] G. Example 7: Successful humanization of anti-EGFR sequences

[0561] The results are shown below.

[0562]

[0563] These results indicate that humanization of the EGFR binding domain was successful and that there is strong affinity for the target EGFR when there are two binding sites on the molecule.

[0564] Example: Successful humanization of EpCAM sdABD

[0565] The results are shown below.

[0566]

[0567] These results indicate that humanization of the EpCAM-binding domain was successful.

[0568] H. Example 8: COBRA TM : A novel conditionally active bispecific antibody that regresses established solid tumors in mice

[0569] Despite the clinical success of bispecific antibodies (bsAbs) targeting hematological malignancies (e.g., blinatumomab, a CD19xCD3 bsAb), efficacy in solid tumor indications remains a significant challenge. Because T cell redirecting bsAbs are so effective, even very low levels of cell surface target antigen expression on normal tissues can quickly become a safety concern and severely limit the dose levels that can be achieved in patients. This limits the likelihood of achieving effective concentrations and reduces the therapeutic potential of these highly active molecules. Furthermore, identifying "clean" target antigens that are uniquely expressed on tumors but not on normal tissues is difficult at best.

[0570] To overcome these challenges, a system called COBRA was developed TM (conditional bispecific redirected activation) is a new recombinant bsAb platform. COBRA is engineered to be able to target more widely expressed and validated tumor cell surface antigens by concentrating T cell engagement in the tumor microenvironment. COBRA molecules are designed to bind to target antigens that can be expressed on tumors and normal cells, but do not engage T cells unless exposed to a proteolytic microenvironment, which is common in tumors but not common in normal healthy tissues. Once bound to the tumor target antigen, protease-dependent linker cleavage allows COBRA to convert inactive anti-CD3 scFv into an active CD3 scFv binding domain. After conversion, COBRA is then able to simultaneously engage T cells and target antigens, thereby generating an effective cytolytic T cell response against tumor cells. In addition, COBRA is designed with a half-life extension portion that is removed from the active molecule after proteolytic cleavage. This allows the inactive COBRA to persist in the circulation before tumor target binding and to clear unbound active COBRA molecules more quickly, thereby reducing the possibility of cytotoxic activity in normal tissues.

[0571] Here, we reveal a novel design of the COBRA molecule and demonstrate its ability to engage CD3 and epidermal growth factor receptor (EGFR) to elicit potent cytotoxic activity in T cell cultures and in human T cell-implanted tumor-bearing mice. We report sub-picomolar T cell activation and cytotoxicity in vitro, as well as COBRA linker cleavage-dependent T cell-mediated regression of established solid tumor xenografts in NSG mice in vivo.

[0572] Figures 64A-64CThe COBRA design and predicted folding mechanism are shown. Figure 64A Schematic diagram depicting PRO186 COBRA. Figure 64B The predicted COBRA fold is shown. COBRA comprises inactive VH and VL paired with anti-CD3 VH and VL domains. Uncleaved PRO186 COBRA binds to EGFR, has impaired CD3 binding, and binds serum albumin. Figure 64C Shown is an analytical size exclusion chromatogram of PRO186. The data indicate that uncleaved PRO186 folds into a single structure.

[0573] Figures 65A-65C Depicts exemplary embodiments of constructs described herein, including PRO186 (pre-cleaved PRO186), PRO186 cleavage products, and PRO186 active dimers. One cleavage product comprises the anti-CD3 VH and VL domains, and it binds to EGFR and has impaired CD3 binding. Another cleavage product comprises the anti-CD3 inactive VH and VL domains and binds to serum albumin. Active PRO186 dimers comprise active anti-CD3 agonists (dimers of anti-CD3 VH and VL) and bind to CD3 and EGFR.

[0574] Figure 66 Provides an illustration of the conversion of COBRA to the active dimer following protease cleavage.

[0575] Figure 67A-Figure 67B Characterization of COBRA binding is provided. Figure 67A Binding activity is shown for human, cynomolgus monkey and mouse preparations. Figure 67B PRO186 binding to human CD3ε is shown; active PRO186 binding to human CD3ε and active PRO186 binding to human EGFR. Binding kinetics were assessed by Octet (Forte Bi) with EGFR (Acro Biosystems), serum albumin (Athens Research Technology) and CD3ε (Creative Biomart).

[0576] Figures 68A-68B Cleavage of the PRO186 linker by MMP2 and MMP9 is shown. Figure 68A Western blot depicting active binding product molecules after cleavage. Figure 68B The accumulation of active binding product molecules is shown relative to the cleavage time.

[0577] Figure 69 In vitro activity of the conditional PRO186 construct is shown. Figure 69 - Left panel shows the results of T cell killing assay. Figure 69- The right panel shows the level of IFN-γ release in relation to the concentration of the test article.

[0578] Figure 70 EGFR expression relative to activity is shown in three tumor cell lines - LoVo (colorectal cancer (CRC) cell line), HT-29 (colorectal cancer (CRC) cell line) and SCC25 (head and neck cancer cell line). For in vitro EGFR expression, bound antibody / cell was measured using 1:1 PE-labeled anti-EGFR mAb #EGFR.1 and BD QuantiBrite Bead. For in vivo EGFR expression, IHC staining was performed using anti-EGFR mAb #WP84 and MACH4-HRP detection (Ensigna). For T cell killing assays, luciferase-expressing tumor cells were co-cultured with human T cells at an E:T of 10:1 for 48 hours and measured by Steady-Glo (Promega). For IFNγ release assays, IFNγ was measured at 24 hours at an E:T of 10:1 using a Meso Scale Discovery V-Pex.

[0579] Figure 71A and Figure 71B EGFR, MMP2, and MMP9 expression on tumor cells and tumor xenografts is shown. Figure 71A Shown are the cell surface densities of EGFR on three cancer cell lines - LoVo, HT-29 and SCC25. Figure 71B Immunohistochemical staining of tumor xenografts for EGFR, MMP2, and MMP9 is shown.

[0580] Figure 72 A schematic diagram of the experimental procedure for the adoptive human T cell transfer model in tumor-bearing mice is provided. The experiment was used to measure in vivo anti-tumor efficacy and pharmacokinetics (PK). The procedure included (1) subcutaneous implantation of tumors in the right flank of NSG mice, (2) allowing established tumors, such as tumors of approximately 200 mm3, to form, (3) giving the mice q3dx7 starting on day 0, (4) administering the last dose on day 18, and (5) terminating the study. The procedure also included the following, which were performed simultaneously with the in vivo experiments: (a) activating and expanding human T cells in culture for 10 days so that expansion began over the same time course as tumor implantation, and (b) harvesting T cells on day 0.

[0581] Figure 73 PRO186 is shown to cause regression of established solid tumors in mice. Figure 73 - Left panel shows regression of LoVo derived tumors. Figure 73 - Middle panel shows regression of HT-29 derived tumors. Figure 73- Right panel shows regression of SCC25-derived tumors.

[0582] Figure 74A-Figure 74B Cleaved PRO186 was shown to be cleared faster than intact (uncleaved) PRO186. Figure 74A Shown are the pharmacokinetics of the test articles in the plasma of non-tumor-bearing mice. Figure 74B Shown are the tumor volumes of LoVo-derived tumors in mice administered with the test articles.

[0583] Conclusions: Multivalent sdAb-diabody fusions were designed that were converted into highly potent bispecific redirecting T cell therapeutics upon proteolytic activation. In vitro assays demonstrated that protease-dependent linker cleavage enhanced T cell-mediated killing by 200-fold, resulting in a therapeutic with sub-picomolar potency. Administration of PRO186 (Pro186) in mice with established xenografts resulted in protease-cleavage-dependent T cell-mediated tumor regression in multiple tumor models. PRO186 exhibited (1) prolonged in vivo half-life following administration and (2) rapid clearance upon proteolytic activation, demonstrating that PRO186 is a therapeutic with an improved safety profile over conventional T cell redirecting bispecifics.

Claims

1. A fusion protein comprising, from N-terminus to C-terminus: a) first sdABD-TTA; b) a first domain linker; c) a constrained Fv domain, said constrained Fv domain comprising: i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3; ii) a constrained non-cleavable linker (CNCL); and iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3; d) a second domain linker; e) second sdABD-TTA; f) a cleavable linker (CL); g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising: i) a first pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a first pseudo heavy variable domain; h) a third domain linker; and i) a third sdABD that binds to human serum albumin; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3; The first variable heavy domain and the first pseudo variable light domain intermolecularly associate to form an inactive Fv; The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv; and wherein at least one of the sdABD-TTAs is a sdABD-B7H3 having a sequence selected from the group consisting of SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53, and SEQ ID NO:

57.

2. The fusion protein of claim 1, which is Pro664 and has SEQ ID NO:

282.

3. A fusion protein comprising, from N-terminus to C-terminus: a) a first single domain antigen binding domain (sdABD) that binds to a human tumor target antigen (TTA) (sdABD-TTA); b) a first domain linker; c) a constrained Fv domain, said constrained Fv domain comprising: i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3; ii) a constrained non-cleavable linker (CNCL); and iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3; d) a second domain linker; e) second sdABD-TTA; f) a cleavable linker (CL); g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising: i) a first pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a first pseudo heavy variable domain; h) a third domain linker; and i) a third sdABD that binds to human serum albumin; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3; The first variable heavy domain and the first pseudo variable light domain intermolecularly associate to form an inactive Fv; The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv; and wherein at least one of the sdABD-TTAs is a sdABD-EpCAM having a sequence selected from the group consisting of SEQ ID NO: 69 and SEQ ID NO:

73.

4. A fusion protein comprising, from N-terminus to C-terminus: a) first sdABD-TTA; b) a first domain linker; c) a constrained Fv domain, said constrained Fv domain comprising: i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3; ii) a constrained non-cleavable linker (CNCL); and iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3; d) a second domain linker; e) second sdABD-TTA; f) a cleavable linker (CL); g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising: i) a first pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a first pseudo heavy variable domain; h) a third domain linker; and i) a third sdABD that binds to human serum albumin; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3; The first variable heavy domain and the first pseudo variable light domain intermolecularly associate to form an inactive Fv; The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv; and wherein at least one of the sdABD-TTA is sdABD-Trop2 having a sequence selected from the group consisting of SEQ ID NO:77, SEQ ID NO:81, SEQ ID NO:85, SEQ ID NO:89, SEQ ID NO:93, and SEQ ID NO:

97.

5. A fusion protein comprising, from N-terminus to C-terminus: a) first sdABD-TTA; b) a first domain linker; c) a constrained Fv domain, said constrained Fv domain comprising: i) a first variable heavy domain comprising vhCDR1, vhCDR2 and vhCDR3; ii) a constrained non-cleavable linker (CNCL); and iii) a first variable light domain comprising vlCDR1, vlCDR2 and vlCDR3; d) a second domain linker; e) second sdABD-TTA; f) a cleavable linker (CL); g) a constrained pseudo-Fv domain, said constrained pseudo-Fv domain comprising: i) a first pseudo light variable domain; ii) a non-cleavable linker (NCL); and iii) a first pseudo heavy variable domain; h) a third domain linker; and i) a third sdABD that binds to human serum albumin; wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3, but the constrained Fv domain does not bind to CD3; The first variable heavy domain and the first pseudo variable light domain associate intramolecularly to form an inactive Fv; and The first variable light domain and the first pseudo variable heavy domain associate intramolecularly to form an inactive Fv. wherein at least one of the sdABD-TTAs is sdABD-CA9 having a sequence selected from the group consisting of SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109, and SEQ ID NO:

113.

6. The fusion protein of any one of claims 1 and 3 to 5, wherein the first variable heavy domain is located at the N-terminus of the first variable light domain, and the pseudo light variable domain is located at the N-terminus of the pseudo variable heavy domain.

7. The fusion protein of any one of claims 1 and 3 to 5, wherein the first variable heavy domain is located at the N-terminus of the first variable light domain, and the pseudo variable heavy domain is located at the N-terminus of the pseudo variable light domain.

8. The fusion protein of any one of claims 1 and 3 to 5, wherein the first variable light domain is located at the N-terminus of the first variable heavy domain, and the pseudo light variable domain is located at the N-terminus of the pseudo variable heavy domain.

9. The fusion protein of any one of claims 1 and 3 to 5, wherein the first variable light domain is located at the N-terminus of the first variable heavy domain, and the pseudo variable heavy domain is located at the N-terminus of the pseudo variable light domain.

10. The fusion protein of any one of claims 1 and 3 to 9, wherein the first and second TTAs are identical.

Images