Constrained conditionally activated binding proteins

Abstract

To provide many different protein compositions for the treatment of cancer.SOLUTION: The invention relates to Conditional Bispecific Redirected Activation constructs, or COBRAs, that are administered in an active pro-drug format. Upon exposure to tumor proteases, the constructs are cleaved and activated, such that they can bind both tumor target antigens (TTAs) as well as CD3, thus recruiting T cells expressing CD3 to the tumor, resulting in treatment. In some embodiments, the tumor target antigen is B7H3.SELECTED DRAWING: None

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Application No. 62 / 814,210, filed March 5, 2019, U.S. Provisional Application No. 62 / 814,744, filed March 6, 2019, and U.S. Provisional Application No. 62 / 826,523, filed March 29, 2019, the disclosures of which are incorporated herein by reference in their entireties. [Background technology]

[0002] The selective destruction of individual cells or specific cell types is often desirable in various clinical settings. For example, a major goal of cancer therapy is to specifically destroy tumor cells while leaving healthy cells and tissues as intact and undamaged as possible. One such method is to induce an immune response against the tumor, causing immune effector cells, such as natural killer (NK) cells or cytotoxic T lymphocytes (CTLs), to attack and destroy tumor cells.

[0003] The use of intact monoclonal antibodies (mAbs), which offer excellent binding specificity and affinity for tumor-associated antigens, has been successfully applied in the fields of cancer therapy and diagnosis. However, the large size of intact mAbs, their poor biodistribution, low potency, and long persistence in the blood pool have limited their clinical use. For example, intact antibodies can exhibit specific accumulation within tumors. In biodistribution studies, heterogeneous antibody distribution with primary accumulation in peripheral regions is evident when tumors are closely examined. Due to tumor necrosis, heterogeneous antibody distribution, and increased interstitial tissue pressure, it is impossible to reach the core of tumors with intact antibody constructs. In contrast, smaller antibody fragments exhibit rapid tumor localization, penetrate deeper into tumors, and are cleared relatively quickly from the bloodstream. However, many antibodies, including scFvs and other constructs, exhibit an "on-target / off-tumor" effect, in which the molecule is active on non-tumor cells, 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

[0004] The present invention provides many different protein compositions for the treatment of cancer. Accordingly, in one aspect, the present invention provides a constrained Fv domain 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), b) a domain linker, c) 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, and f) a cleavable linker (CL). g) a 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 human serum albumin, wherein the variable heavy domain and variable light domain are capable of binding to human CD3 but the constrained Fv domain does not bind to CD3, the variable heavy domain and pseudo-variable light domain associate intramolecularly to form an inactive Fv, and the variable light domain and pseudo-variable heavy domain associate intramolecularly to form an inactive Fv. In some embodiments, the human tumor target antigen is B7H3.

[0005] In a further aspect, the present invention provides a first single domain antigen binding domain (sdABD) (sdABD-TTA) that binds to a human tumor target antigen (TTA), comprising, from N-terminus to C-terminus, sdFR1-sdCDR1-sdFR2-sdCDR2-sdFR3-sdCDR3-sdFR4; b) a first domain linker; and c) a second domain linker comprising: i) vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCD a constrained Fv domain comprising: i) a variable heavy domain comprising vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4; 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); and g) a variable light domain comprising i) sdFR1-sdCDR1-sdFR2- a constrained pseudo-Fv domain comprising: ii) a pseudo-light variable domain comprising 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) a third sdABD that binds to human serum albumin, 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 to CD3, and the variable heavy domain and pseudo-variable light domain associate intramolecularly to form an inactive Fv, and the variable light domain and pseudo-variable heavy domain associate intramolecularly to form an inactive Fv. In some embodiments, the human tumor target antigen is B7H3.

[0006] In some embodiments of the Format 2 protein, the variable heavy domain is N-terminal to the variable light domain and the pseudo-light variable domain is N-terminal to the pseudo-variable heavy domain. In some embodiments, the variable heavy domain is N-terminal to the variable light domain and the pseudo-variable light domain is C-terminal to the pseudo-variable heavy domain. In some embodiments, the variable heavy domain is C-terminal to the variable light domain and the pseudo-variable light domain is N-terminal to the pseudo-variable heavy domain. In some embodiments, the variable heavy domain is C-terminal to the variable light domain and the pseudo-variable light domain is C-terminal to the pseudo-variable heavy domain.

[0007] In some embodiments of the Format 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 FIG. 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 NOs: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.

[0008] In some embodiments of a Format 2 protein, the pseudo-heavy variable domain of the constrained pseudo-Fv domain is SEQ ID NO: 146 (V Hi ), SEQ ID NO: 150 (V Hi2 ), and SEQ ID NO: 154 (VHiGL4). In some embodiments, the pseudo-light variable domain of the constrained pseudo-Fv domain is selected from the group of SEQ ID NO: 130 (V Li ), SEQ ID NO: 134 (V Li2 ), and SEQ ID NO: 138 (V LiGL ) is selected from the group

[0009] In a further aspect, the invention provides a constrained Fv domain comprising, from N-terminus to C-terminus, a) a first sdABD-TTA; b) a first domain linker; c) 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); and g) 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 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.In an additional aspect, the invention provides a constrained Fv domain comprising, from N-terminus to C-terminus: a) a single domain antigen binding domain (sdABD) that binds 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 (CL); e) a second sdABD that binds human serum albumin; and f and g) a 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 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.

[0010] In further embodiments of the above-listed Format 1, Format 2, and Format 4 proteins, the first variable heavy domain is N-terminal to the first variable light domain, and the pseudo-light variable domain is N-terminal to the pseudo-variable heavy domain.

[0011] In further embodiments of the above-listed Format 1, Format 2, and Format 4 proteins, the first variable heavy domain is N-terminal to the first variable light domain, and the pseudo-variable heavy domain is N-terminal to the pseudo-variable light domain.

[0012] In further embodiments of the above-listed Format 1, Format 2, and Format 4 proteins, the first variable light domain is N-terminal to the first variable heavy domain, and the pseudo-light variable domain is N-terminal to the pseudo-variable heavy domain.

[0013] In further embodiments of the above-listed Format 1, Format 2, and Format 4 proteins, the first variable light domain is N-terminal to the first variable heavy domain, and the pseudo-variable heavy domain is N-terminal to the pseudo-variable light domain.

[0014] In an additional aspect, the invention provides Format 1 and 2 proteins, wherein the first and second TTAs are the same.

[0015] In a further aspect, the invention provides Format 1 and 2 proteins, wherein the first and second TTAs are different.

[0016] In an additional aspect, the present invention provides Format 1, 2, and 4 proteins, wherein the first and second TTAs are selected from EGFR, EpCAM, FOLR1, Trop2, ca9, and B7H3. These sequences 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, SEQ ID NO: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.

[0017] In a further aspect, the invention provides Format 1, 2, and 4 proteins, wherein the half-life prolonging domain has SEQ ID NO: 117 (aHSA(10GE)) and SEQ ID NO: 121 (His-tagged aHSA).

[0018] In an additional aspect, the invention provides Format 1, 2, and 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, cathespin L, granzyme B, uPA, kallikrein 7, matriptase, and thrombin, or others shown in Figure 6.

[0019] In a further aspect, the present invention provides a method for the preparation of medicaments ... The present invention provides a protein selected from the group consisting of o695, 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.

[0020] In additional aspects, the invention provides nucleic acids encoding the Format 1, 2, or 4 proteins described herein, as well as expression vectors and host cells comprising the nucleic acids encoding the proteins.

[0021] In a further aspect, the invention provides methods of making the proteins of the invention and administering the same to patients in need of treatment.

[0022] In an additional aspect, the invention provides a protein comprising: a) a first protein, the pseudo-Fv domain 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, and 3) a first pseudo-variable light domain comprising iVLCDR1, iVLCDR2, and iVLCDR3; iv) a second domain linker; and v) an 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; 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; and v) s and a second protein comprising a "Format 3A" pair of prodrug proteins, wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3 when associated, 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 intermolecularly associate to form an inactive Fv, and the first and third sdABDs , 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, SEQ ID NO: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.

[0023] In a further aspect, the invention provides a method for producing a polypeptide 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, and 3) a first pseudo-variable light domain comprising iVLCDR1, iVLCDR2, and iVLCDR3, iv) a third domain linker, and v) an sdABD-HSA; and 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, and iv) a fifth sdABD-HSA. and a second protein comprising: a domain linker; and 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 sixth domain linker; and v) sdABD-HSA, wherein the first variable heavy domain and the first variable light domain are capable of binding to human CD3 when associated, the first variable heavy domain and the first pseudo-variable light domain associate intermolecularly to form an inactive Fv, and the first variable light domain and the first pseudo-variable heavy domain associate intermolecularly to form an inactive Fv.

[0024] In additional embodiments, the Format 3A and Format 3B proteins have an sdABD-HSA having SEQ ID NO:117 or SEQ ID NO:121.

[0025] In further aspects, Format 3A and Format 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 NOs: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.

[0026] In a further aspect, the invention provides an sdABD that binds to human Trop2 having a sequence selected from SEQ ID NO:77, SEQ ID NO:81, SEQ ID NO:85, SEQ ID NO:89, and SEQ ID NO:93.

[0027] In a further aspect, the invention provides an sdABD that binds to human B7H3 having a sequence selected from SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53, and SEQ ID NO:57.

[0028] In an additional aspect, the invention provides an sdABD that binds to human CA9 having a sequence selected from SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109, and SEQ ID NO:113.

[0029] In a further aspect, the invention provides an sdABD that binds to human EpCAM having a sequence selected from SEQ ID NO:69 and SEQ ID NO:73.

[0030] In a further aspect, the 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, as well as expression vectors and host cells containing the nucleic acids. [Brief explanation of the drawings]

[0031] [Figure 1] This example illustrates a "Form 1" type of protease activation of the present invention, referred to herein as a "constrained-cleavage construct" or "cc construct." In this embodiment, the representative construct is Pro140, and there is an ABD for each of the two TTAs (both of which are the same as shown in FIG. 1 , but they can be different as described herein). Upon cleavage, the prodrug construct splits into three components: one containing the α-TTA domain linked to the active VH of αCD3 via a domain linker, a second containing the α-TTA domain linked to the active VL of αCD3 via a domain linker, and the "remaining" piece containing a half-life extending 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. Note that in "Format 1" embodiments, the resulting active component is trivalent, with one binding to CD3 and two binding to TTA, resulting in a bispecific binding protein; in some cases, this trivalent may be trispecific, with one binding to CD3, one binding to a first TTA, and one binding to a second TTA. Figure 1 also shows a half-life prolonging domain, in many embodiments an anti-human serum albumin (HSA) domain, as defined herein as sdABD, although as discussed herein, this is optional and / or may be replaced by other half-life prolonging domains; in addition, the half-life prolonging domain may also be at the N-terminus of the construct or may be internal as well. 1 also shows the VH and VL of the Fv and the iVH and iVL of the pseudo-Fv in a particular order, e.g., VH-linker-VL (and iVL-linker-iVH) from N-terminus to C-terminus, although as will be understood by those skilled in the art, these may be reversed (VL-linker-VH and iVH-linker-iVL). Alternatively, one of these Fvs may be in one orientation and the other in another, although expression of the protein in the orientation shown here was surprisingly higher than in other orientations. [Figure 2]These constructs, referred to herein as "constrained non-cleavable constructs" or "CNCL constructs," and sometimes referred to herein as "dimerization constructs" as discussed herein, represent a "Form 2" type of protease activation of the present invention. These constructs do not isomerize as discussed herein. Upon cleavage, the two prodrug constructs split into four components: two half-life extending domains (in this case, sdABDs for HSA) linked to pseudodomains (which may or may not be capable of self-association, depending on the linker length and inactivating mutations), and two active moieties that self-assemble into a dimeric active moiety containing four anti-TTA domains (which may all be the same, or two are the same and the other two are different). Note that in "Format 2" embodiments, the resulting active component is hexavalent, with a divalent that binds CD3 and a tetravalent that binds TTA, resulting in a bispecific binding protein; in some cases, this hexavalent may be trispecific, with a divalent that binds CD3, a divalent that binds a first TTA, and a divalent that binds a second TTA. Figure 2 also shows a half-life prolonging domain, in many embodiments an anti-human serum albumin (HSA) domain, as defined herein as sdABD, although as discussed herein, this is optional and / or can be replaced by other half-life prolonging domains; in addition, the half-life prolonging domain may also be at the N-terminus of the construct or internal as well. 2 also shows the VH and VL of the Fv and the iVH and iVL of the pseudo-Fv in a particular order, e.g., VH-linker-VL (and iVL-linker-iVH) from N-terminus to C-terminus, although as will be appreciated by those skilled in the art, these may be reversed (VL-linker-VH and iVH-linker-iVL). Alternatively, one of these Fvs may be in one orientation and the other in another, although expression of the protein in the orientation shown here was surprisingly higher than in other orientations. [Figure 3A]Figures 3A-3B show a "Format 3" type construct, sometimes referred to as a "hemi-construct" or "hemi-COBRA™" as outlined herein, because these are two distinct polypeptide chains that together constitute the MCE therapeutic, as discussed further herein. In this embodiment, the constructs are delivered in pairs, and intramolecular self-assembly prior to cleavage results in an inactive anti-CD3 Fv domain. Upon cleavage, the inactive variable domain is released, and the two active variable domains then intermolecularly assemble to form the active anti-CD3 binding domain. The two sdABD-TTAs bind to their corresponding receptors on the tumor cell surface, and cleavage is carried out by a protease. This allows for intermolecular assembly, as the molecules are physically held in place and favors assembly of the active anti-CD3 domains. As described above for Formats 1 and 2, in this embodiment, the N- to C-terminal order of the variable domains may be reversed or mixed as well. Additionally, the sdABD(HSA) may be at either the N- or C-terminus of each hemi-construct. Pro16 has an sdABD(HSA) at the C-terminus, and Pro17 has an sdABD(HSA) at the N-terminus (see Pro19, SEQ ID NO:XX has an sdABD(HSA) at the C-terminus). Figure 3A shows a Format 3 construct with a single sdABD-TTA domain per hemiconstruct, and Figure 3B shows a Format 3 construct with two sdABD-TTAs per hemiconstruct in a "dual-targeting" or "hetero-targeting" format. Note that Figure 3B uses FOLR1 and EGFR as the two TTAs, but other combinations as outlined herein can also be used. [Figure 3B] Same as above. [Figure 4]Figure 1 shows a "Format 4" type construct similar to the "Format 2" construct, but with only a single sdABD-TTA. The diagram shows the sdABD-TTA to EGFR, but as will be understood by those skilled in the art, other TTAs can be used. Upon cleavage, the prodrug construct splits into two components: a half-life extending domain (in this case, an sdABD to HSA) linked to a pseudo-Fv, and an active moiety that 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. Note that in the "Format 4" embodiment, the resulting active moiety is tetravalent, with one binding to CD3 and one binding to TTA, resulting in a bispecific binding protein. Figure 4 also shows a half-life-prolonging domain, in many embodiments an anti-human serum albumin (HSA) domain, as defined herein, but as discussed herein, this is optional and / or can be replaced by other half-life-prolonging domains. Additionally, the half-life-prolonging domain may also be at the N-terminus of the construct or may be internal as well. Figure 4 also shows the VH and VL of the Fv and the iVH and iVL of the pseudo-Fv in a specific order, e.g., VH-linker-VL (and iVL-linker-iVH) from N-terminus to C-terminus, although as will be understood by those skilled in the art, these may also be reversed (VL-linker-VH and iVH-linker-iVL). Alternatively, one of these Fvs may be in one orientation and the other in another, although expression of the protein in the orientation shown here was surprisingly higher than in other orientations. [Figure 5A] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5B]A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5C] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5D] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5E] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5F] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5G] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5H] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5I] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 5J] A number of sequences of the invention are shown. For antigen-binding domains, the CDRs are underlined. As more fully outlined herein, these domains can be assembled in a wide variety of configurations within the invention, including "Form 1," "Form 2," "Form 3," and "Form 4" orientations. [Figure 6A] A number of suitable protease cleavage sites are shown. As will be appreciated by those of skill in the art, these cleavage sites can be used as cleavable linkers. In some embodiments, for example, if a more flexible cleavable linker is required, there can be additional amino acids (generally glycine and serine) at either or both the N- and C-terminus of these cleavage sites. [Figure 6B] A number of suitable protease cleavage sites are shown. As will be appreciated by those of skill in the art, these cleavage sites can be used as cleavable linkers. In some embodiments, for example, if a more flexible cleavable linker is required, there can be additional amino acids (generally glycine and serine) at either or both the N- and C-terminus of these cleavage sites. [Figure 6C] A number of suitable protease cleavage sites are shown. As will be appreciated by those of skill in the art, these cleavage sites can be used as cleavable linkers. In some embodiments, for example, if a more flexible cleavable linker is required, there can be additional amino acids (generally glycine and serine) at either or both the N- and C-terminus of these cleavage sites. [Figure 6D]A number of suitable protease cleavage sites are shown. As will be appreciated by those of skill in the art, these cleavage sites can be used as cleavable linkers. In some embodiments, for example, if a more flexible cleavable linker is required, there can be additional amino acids (generally glycine and serine) at either or both the N- and C-terminus of these cleavage sites. [Figure 7A] We present some data relating to the "Format 3" or "Hemi-COBRA™" construct, which shows that the Format 3 construct binds cooperatively to CD3 after cleavage by a protease (in this case, EK protease, but any of the protease cleavage sites outlined herein and shown in Figures 5 and 6 can be used), creating a CD3 binding site, as shown by sandwich FACS analysis. [Figure 7B] We present some data relating to the "Format 3" or "Hemi-COBRA™" construct, which shows that the Format 3 construct binds cooperatively to CD3 after cleavage by a protease (in this case, EK protease, but any of the protease cleavage sites outlined herein and shown in Figures 5 and 6 can be used), creating a CD3 binding site, as shown by sandwich FACS analysis. [Figure 7C] We present some data relating to the "Format 3" or "Hemi-COBRA™" construct, which shows that the Format 3 construct binds cooperatively to CD3 after cleavage by a protease (in this case, EK protease, but any of the protease cleavage sites outlined herein and shown in Figures 5 and 6 can be used), creating a CD3 binding site, as shown by sandwich FACS analysis. [Figure 7D]We present some data relating to the "Format 3" or "Hemi-COBRA™" construct, which shows that the Format 3 construct binds cooperatively to CD3 after cleavage by a protease (in this case, EK protease, but any of the protease cleavage sites outlined herein and shown in Figures 5 and 6 can be used), creating a CD3 binding site, as shown by sandwich FACS analysis. [Figure 8A] These results demonstrate that protease cleavage cooperatively activates complementary hemi-COBRA™ pairs and T cell killing of EGFR+ target cells. Figures 8A and 8B show that constructs cleaved with different concentrations of protease, in isolation, do not affect target cell viability. However, Figure 8C shows that in combination, target cell viability is significantly reduced in the presence of protease. Figure 8D illustrates the general mechanism. [Figure 8B] These results demonstrate that protease cleavage cooperatively activates complementary hemi-COBRA™ pairs and T cell killing of EGFR+ target cells. Figures 8A and 8B show that constructs cleaved with different concentrations of protease, in isolation, do not affect target cell viability. However, Figure 8C shows that in combination, target cell viability is significantly reduced in the presence of protease. Figure 8D illustrates the general mechanism. [Figure 8C] These results demonstrate that protease cleavage cooperatively activates complementary hemi-COBRA™ pairs and T cell killing of EGFR+ target cells. Figures 8A and 8B show that constructs cleaved with different concentrations of protease, in isolation, do not affect target cell viability. However, Figure 8C shows that in combination, target cell viability is significantly reduced in the presence of protease. Figure 8D illustrates the general mechanism. [Figure 8D]These results demonstrate that protease cleavage cooperatively activates complementary hemi-COBRA™ pairs and T cell killing of EGFR+ target cells. Figures 8A and 8B show that constructs cleaved with different concentrations of protease, in isolation, do not affect target cell viability. However, Figure 8C shows that in combination, target cell viability is significantly reduced in the presence of protease. Figure 8D illustrates the general mechanism. [Figure 9] Several non-targeting controls are shown for use in assays to test the efficacy of Format 1 constructs. [Figure 10A] This demonstrates that generation of an active CD3-binding domain depends on target binding of both "arms," ​​e.g., the sdABD-TTA domains (one of which is present on each of the two constructs). ELISA assays were performed as described in the Examples. [Figure 10B] This demonstrates that generation of an active CD3-binding domain depends on target binding of both "arms," ​​e.g., the sdABD-TTA domains (one of which is present on each of the two constructs). ELISA assays were performed as described in the Examples. [Figure 10C] This demonstrates that generation of an active CD3-binding domain depends on target binding of both "arms," ​​e.g., the sdABD-TTA domains (one of which is present on each of the two constructs). ELISA assays were performed as described in the Examples. [Figure 10D] This demonstrates that generation of an active CD3-binding domain depends on target binding of both "arms," ​​e.g., the sdABD-TTA domains (one of which is present on each of the two constructs). ELISA assays were performed as described in the Examples. [Figure 10E] This demonstrates that generation of an active CD3-binding domain depends on target binding of both "arms," ​​e.g., the sdABD-TTA domains (one of which is present on each of the two constructs). ELISA assays were performed as described in the Examples. [Figure 10F]This demonstrates that generation of an active CD3-binding domain depends on target binding of both "arms," ​​e.g., the sdABD-TTA domains (one of which is present on each of the two constructs). ELISA assays were performed as described in the Examples. [Figure 11] A schematic diagram of a suitable hemi-COBRA™ pair is shown, where "Mep" represents a meprin protease cleavage site, "His-6" is a tag as more fully discussed herein, ST14 is a matriptase protease cleavage site, and "Thb" is a thrombin protease cleavage site. [Figure 12A] TDCC data associated with the constructs in Figure 11 are shown. Figure 12A shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against OvCAR8 cells, Figure 12B shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against HCT116 cells, and Figure 12C shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against LoVo cells, all of which are cancer cell lines. [Figure 12B] TDCC data associated with the constructs in Figure 11 are shown. Figure 12A shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against OvCAR8 cells, Figure 12B shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against HCT116 cells, and Figure 12C shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against LoVo cells, all of which are cancer cell lines. [Figure 12C] TDCC data associated with the constructs in Figure 11 are shown. Figure 12A shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against OvCAR8 cells, Figure 12B shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against HCT116 cells, and Figure 12C shows that the addition of a pre-cleaved hemi-COBRA pair confers efficacy against LoVo cells, all of which are cancer cell lines. [Figure 13]Figures 13A-13B show that the MMP9 linker is stable in vivo. NSG mice were administered a single intravenous bolus dose of either Pro40 (MMP9-cleavable) or Pro74 (non-cleavable) via the tail vein at a dose level of 0.5 mg / kg. Dosing solutions of each compound were 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 via orbital or submandibular bleeding near the beginning of the study, and one via cardiac puncture at the terminal time point. Blood collection times 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 with MAbs specific for the anti-HSA sdABD, detected in the EGFR extracellular domain. [Figure 14] A schematic diagram of the Format 3A hemi-COBRA™ constructs used in the experiments shown in Figure 15 is shown. Pro51 is a positive control because it is "always on" since it forms an active anti-CD3 Fv. Pro98 is a negative control because its sdABD is directed against hen egg white lysozyme, which is not expressed by tumors. Pro77 and Pro53 are prodrug Format 3A pairs using an sdABD against EGFR and an MMP9 cleavage site. Pro74 and Pro72 are negative control Format 3A pairs because they do not have a cleavage site. [Figure 15] We demonstrate that Format 1 constructs are effective in causing tumor regression in vivo using two different tumor cell lines implanted in mice using the protocol in the Examples. Antitumor activity with the hemi-COBRA constructs (Pro77 and Pro53) was dependent on the inclusion of both the anti-EGFR sdABD and MMP9-cleavable linker along with the active anti-CD3 Fv. [Figure 16]

[0023] Figure 1 shows a schematic diagram of a next-generation full-length construct having two pseudo-Fv domains with a cleavable site between them, as generally described in US2018 / 0134789, which is incorporated herein by reference. However, as shown in the diagram below, this first-generation full-length construct does not exhibit very good conditioning, as it can isomerize to form both active and inactive constructs. [Figure 17] We show that the Format 3A construct pair indeed exhibits better conditioning than the first generation full-length construct of Pro100. [Figure 18] Figure 19 shows additional first generation full-length constructs that were tested. [Figure 19] It is shown that the first generation constructs exhibit high activity even in the uncleaved form, eg, under poor conditions. [Figure 20] Figure 1 shows that the first generation full-length construct exhibits two monomer peaks on analytical SEC. [Figure 21] A schematic diagram of the reason for the non-cleavable activity is shown, which is that the full-length first generation construct isomerizes to form two conformations: one that is inactive due to the absence of an active anti-CD3 Fv formed ("bivalent scFv"), and one that, in the absence of proteases, forms a "single-chain diabody" type of configuration. See PEDS 23(8):667-677 (2010). [Figure 22]

[0033] Figure 1 shows the results of a TDCC assay performed at 37°C for 2 days using first generation single-chain constructs. The results show that the uncleaved constructs exhibit potent killing. These results lead to the generation of Format 1 constructs. [Figure 23A] 1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 23B]1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 23C] 1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 23D] 1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 23E] 1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 23F] 1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 23G] 1 shows Format 1 constructs for use in the present invention. As will be understood by those skilled in the art and as described herein, these are illustrated with sdABD-EGFR or EpCAM targeting moieties, although sdABDs for other TTAs can also be used. [Figure 24] It is shown that the Format 1 construct (in this case Pro140) forms a single isomer that is stable at 37°C. [Figure 25]Figure 1 shows that the Format 1 construct has very low binding to human CD3 in the 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 held at either 4°C or 37°C for 3 days. [Figure 26] It is also shown that the Format 1 construct has very low TDCC activity in the uncleaved form. [Figure 27] The particular Format 1 construct, Pro140, used in in vivo studies using sdABD-EGFR as the targeting moiety and an MMP9 cleavage site is shown. [Figure 28] Figures 28A-B show tumor regression using the Format 1 construct. [Figure 29] We show that the cleavage site of the constrained Fv can generate several different fragments, partial and complete cleavage fragments. Surprisingly, the partial cleavage form is more active than the complete cleavage form, resulting in the generation of form 2. [Figure 30] Schematics of a number of Format 2 are shown, all of which use the sdABD-EGFR targeting domain, although sdABDs for other TTAs can be used as outlined herein and listed in the sequences. Pro51 and Pro201 are positive controls (in the active "hemi" and active dimer configurations, respectively), and Pro214 is a full-length negative control, as it has no cleavage site. [Figure 31] Figure 1 shows the TDCC activity of the Format 2 construct Pro187, which uses a meprin cleavage site. Pro187 was 1200-fold more active in the TDCC assay when pre-cleaved and added than when added uncleaved. Pre-cleaved Pro187 demonstrated activity positioned between the positive controls Pro51 and Pro201. Uncleaved Pro187 demonstrated similar activity to Pro214, which does not contain a protease-cleavable linker. [Figure 32]Figure 1 shows the TDCC activity of the Format 2 construct Pro186, which uses the MMP9 cleavage site. In the TDCC assay, Pro186 was 18-fold more active when pre-cleaved than when pre-cleaved. Pre-cleaved Pro186 demonstrated activity positioned between the positive controls Pro51 and Pro201. Uncleaved Pro186 demonstrated greater activity than Pro214, which does not contain a protease-cleavable linker, likely due to the MMP activity generated by the cells during the 48-hour assay period. [Figure 33] We demonstrate that the Pro186 construct binds to cells with different levels of EGFR receptor but not to CHO cells that do not express EGFR on the cell surface. Pro186 saturates cells expressing different levels of EGFR at similar COBRA concentrations. [Figure 34] A schematic diagram of the Format 2 constructs used in the in vivo studies is shown in Figure 35, all of which use the sdABD-EGFR targeting domain. [Figure 35] The format 2 construct Pro186 is shown to be highly effective at both dose levels and better than the format 1 construct Pro140 at the lower doses. [Figure 36] A number of Format 2 constructs are shown that are based on Pro186 but have different protease cleavage sites. All of these constructs use sdABD-EGFR for both targeting domains, but other sdABDs can be used for different TTAs, either the same or different. That is, both homo-targeting (both targeting sdABDs to the same TTA) or hetero-targeting (one sdABD to one TTA and another sdABD to a different TTA) can be achieved. [Figure 37]Schematic diagrams of different type 2 constructs are shown, with different linker lengths depending on the Fv domain.These are shown using MMP9 cleavage sites, but as outlined herein, other can be used.Similarly, all of these constructs use sdABD-EGFR for both targeting domains, but other sdABDs can be used for different TTAs, and can be the same or different. [Figure 38] We show that the linker length of the pseudo-Fvs can be varied, e.g., a Format 2 construct with a short linker between active Fvs ("short active") and a longer linker between pseudo-Fvs ("long inactive") exhibits similar activity to the "short active" with "short inactive" construct. Thus, the conditioning of COBRA constructs does not depend on 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 favored over the bivalent scFv fold. [Figure 39] This shows that the linker length of the active Fv can be varied, e.g., a Format 2 construct with a "long active" and a "short inactive" behaves similarly to a "short active" and a "short inactive" construct. Thus, the conditioning of the COBRA construct does not depend on 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 favored over the bivalent scFv fold. [Figure 40A]Schematics of the various constructs are shown. Pro188 is a format 1 construct similar to Pro140, except for the longer linker (16-mer) in the pseudo-Fv. Pro189 and Pro190 (format 2 constructs) are similar to Pro186 and Pro187, except for the longer linker (16-mer) in the pseudo-Fv. Pro191 and Pro192 (also format 2 constructs) are similar to Pro189 and Pro190, except for the additional cleavage site upstream of the sdABD(1 / 2). Pro193 (format 4) has a single EGFR targeting domain, iVH and iVL rearranged to be in reverse order, and an additional cleavage site upstream of the sdABD(1 / 2). Pro195 is a format 2 construct similar to Pro186, in which the targeting domain binds to the same TTA, EGFR, but to a different epitope. Pro196, Pro197, and Pro198 are Type 2 constructs with rearranged variable domains. [Figure 40B] Schematics of the various constructs are shown. Pro188 is a format 1 construct similar to Pro140, except for the longer linker (16-mer) in the pseudo-Fv. Pro189 and Pro190 (format 2 constructs) are similar to Pro186 and Pro187, except for the longer linker (16-mer) in the pseudo-Fv. Pro191 and Pro192 (also format 2 constructs) are similar to Pro189 and Pro190, except for the additional cleavage site upstream of the sdABD(1 / 2). Pro193 (format 4) has a single EGFR targeting domain, iVH and iVL rearranged to be in reverse order, and an additional cleavage site upstream of the sdABD(1 / 2). Pro195 is a format 2 construct similar to Pro186, in which the targeting domain binds to the same TTA, EGFR, but to a different epitope. Pro196, Pro197, and Pro198 are Type 2 constructs with rearranged variable domains. [Figure 40C]Schematics of the various constructs are shown. Pro188 is a format 1 construct similar to Pro140, except for the longer linker (16-mer) in the pseudo-Fv. Pro189 and Pro190 (format 2 constructs) are similar to Pro186 and Pro187, except for the longer linker (16-mer) in the pseudo-Fv. Pro191 and Pro192 (also format 2 constructs) are similar to Pro189 and Pro190, except for the additional cleavage site upstream of the sdABD(1 / 2). Pro193 (format 4) has a single EGFR targeting domain, iVH and iVL rearranged to be in reverse order, and an additional cleavage site upstream of the sdABD(1 / 2). Pro195 is a format 2 construct similar to Pro186, in which the targeting domain binds to the same TTA, EGFR, but to a different epitope. Pro196, Pro197, and Pro198 are Type 2 constructs with rearranged variable domains. [Figure 41] We demonstrate that different sdABD clones directed against human FOLR1 exhibit differential killing. We compared the Pro22-type construct (Pro51 with a FLAG sequence instead of NCL) that binds to human FOLR1 with the Pro22-EGFR construct against a number of cell line families. [Figure 42] Schematics of four sdABD-FOLR1 constructs are shown, including a Pro201 active domain dimer as a positive control using sdABD-EGFR2 (two molecules intermolecularly associate to form two active Fvs against CD3), and two Format 2 test articles, Pro311 using h77.2 sdABD and Pro311 using h59.3 sdABD, and two negative controls, Pro299 using h77.2 sdABD and Pro303 using h59.3 sdABD. [Figure 43] 1 shows a schematic diagram of the Format 2 construct of FOLR / MMP9 in vivo design. [Figure 44] We show the efficacy of the Pro312 construct in vivo and demonstrate that the MMP9-cleavable linker is required for antitumor activity. [Figure 45]Schematics of several formats using the sdABD against human B7H3 (sdABD-B7H3) are shown, including Pro244, a positive control (using sdABD-B7H3(hF7)), and two format 2 test articles, a format 2 construct of Pro225, and a negative control lacking the cleavage site at Pro295 (sdABD-B7H3). [Figure 46] Pro225 has greater conditioning than the control Pro295. [Figure 47] 1 shows that the format 2 construct using the meprin linker, Pro373, shows greater conditionality compared to Pro295. [Figure 48] A number of sdABD-B7H3 (using the hF12 sequence) constructs are shown, showing a positive control of Pro51 using sdABD-EGFR, a positive control of Pro244 using sdABD-hF12 B7H3, Pro226 from the test construct, and a negative control of no cleavage site, Pro296. [Figure 49] Figure 1 shows good conditioning of the Pro226 construct in the TDCC assay. [Figure 50] 1 shows humanization of sdABD into human EpCAM. [Figure 51] A schematic of the various formats is shown. Pro244 is a standard T cell engager positive control, Pro205 is an activation domain dimer positive control, Pro199 is a Format 2 construct, and Pro175 is a negative control. [Figure 52] Figure 1 shows the TDCC activity of sdABD-EpCAM constructs showing good conditioning. [Figure 53] Figures 53A-B show the TDCC activity of the sdABD-EpCAM Pro199 construct, which shows good conditioning in HT29 and LoVo cell models. [Figure 54] Figures 54A-B show the TDCC activity of the sdABD-EpCAM Pro200 construct, which shows good conditioning in HT29 and LoVo cell models. [Figure 55]Schematic diagram of Pro255, which uses two different sdABD-TTAs, one against EGFR (sdABD-EGFR) and one against EpCAM (sdABD-EpCAM), compared to Pro199, which has a dual EpCAM sdABD. These are sometimes referred to herein as "hetero-targeting" constructs, in this case, Format 2 constructs. [Figure 56] 1 shows that a Pro255 dual targeting molecule with an MMP9 cleavage site exhibits good conditioning. [Figure 57A] Results of experiments on three different cell types are shown. First, Raji transfectants were generated with similar expression levels of EpCAM, EGFR, and EpCAM + EGFR (data not shown). Pro255, which targets both EpCAM and EGFR, was then tested in TDCC assays using each cell type. Figure 57A shows the parental Raji line, which does not express either receptor. Figure 57B shows conditioning on the EpCAM line. Figure 57C shows conditioning on the EGRF line. Figure 57D shows conditioning on the EpCAM / EGFR line. [Figure 57B] Results of experiments on three different cell types are shown. First, Raji transfectants were generated with similar expression levels of EpCAM, EGFR, and EpCAM + EGFR (data not shown). Pro255, which targets both EpCAM and EGFR, was then tested in TDCC assays using each cell type. Figure 57A shows the parental Raji line, which does not express either receptor. Figure 57B shows conditioning on the EpCAM line. Figure 57C shows conditioning on the EGRF line. Figure 57D shows conditioning on the EpCAM / EGFR line. [Figure 57C]Results of experiments on three different cell types are shown. First, Raji transfectants were generated with similar expression levels of EpCAM, EGFR, and EpCAM + EGFR (data not shown). Pro255, which targets both EpCAM and EGFR, was then tested in TDCC assays using each cell type. Figure 57A shows the parental Raji line, which does not express either receptor. Figure 57B shows conditioning on the EpCAM line. Figure 57C shows conditioning on the EGRF line. Figure 57D shows conditioning on the EpCAM / EGFR line. [Figure 57D] Results of experiments on three different cell types are shown. First, Raji transfectants were generated with similar expression levels of EpCAM, EGFR, and EpCAM + EGFR (data not shown). Pro255, which targets both EpCAM and EGFR, was then tested in TDCC assays using each cell type. Figure 57A shows the parental Raji line, which does not express either receptor. Figure 57B shows conditioning on the EpCAM line. Figure 57C shows conditioning on the EGRF line. Figure 57D shows conditioning on the EpCAM / EGFR line. [Figure 58] 1 shows a schematic diagram of Pro258 in the Format 4 construct. [Figure 59] Figures 59A-B show that Pro258 is conditional in both FBS and human serum. The conditioning of the MMP9 linker is underestimated by MMP9 activity in culture. Interestingly, while Pro51 TDCC activity is inhibited by HSA binding, Pro258 TDCC activity is similar to Pro51 in the presence of HSA. Finally, Pro258 conditioning is slightly enhanced in the presence of HSA by 6X. [Figure 60] Figures 60A-60C. Figure 60A shows the cleavage of an MMP9 substrate by other MMPs. Figures 60B and 60C show the cleavage of a FRET probe containing an MMP9 linker sequence. [Figure 61A] 61A-61B show some exemplary constructs and their formats. [Figure 61B] Same as above. [Figure 62A] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62B] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62C] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62D] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62E] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62F] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62G] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62H] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62I] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62J] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62K] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62L] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62M] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62N] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62O] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62P] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62Q] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62R] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62S] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62T] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62U] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 62V] Although many sequences of the present invention are shown, many additional sequences can also be found in the sequence listing. CDRs are underlined and bold, linkers are double underlined (cleavable linkers are italicized and double underlined), and domain separation is indicated by " / ". All His6 tags are optional, as they can be used to reduce immunogenicity in humans and can be purification tags. [Figure 63A] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63B]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63C] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63D]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63E] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63F]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63G] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63H]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63I] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63J]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63K] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63L]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63M] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63N]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63O] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63P]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63Q] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63R]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63S] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63T]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63U] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63V]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63W] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63X]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63Y] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63Z]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63AA] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63BB]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63CC] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63DD]Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 63EE] Figure 63A shows the amino acid sequences of exemplary Format 2 constructs containing multiple sdABD-B7H3 and pseudo-Fv domains (e.g., Vli2 / Vhi2 domains). Figure 63A shows the amino acid sequences of Pro601 and Pro602. Figure 63B shows the amino acid sequences of V3 and V4. Pro601 contains two identical sdAbs that bind to B7H3 (e.g., aB7H3 hF7 sdAb). Pro602 contains two identical sdAbs that bind to B7H3 (e.g., anti-B7H3 hF12 sdAb). V3 contains two different sdAbs that bind to B7H3 (e.g., anti-B7H3 hF7 sdAb and anti-B7H3 hF12 sdAb). V4 contains two different sdAbs that bind to B7H3 (eg, anti-B7H3 hF7 sdAb and anti-B4H3 hF12 sdAb). [Figure 64] Figures 64A-64C show the design and predicted folding mechanism of COBRA. Figure 64A shows a schematic of Pro186 COBRA (SEQ ID NO: 145 in Figure 62B). Figure 64B shows the predicted COBRA fold. Figure 64C shows an analytical size exclusion chromatogram of Pro186. [Figure 65A] 1 shows an exemplary embodiment of a construct described herein, including Pro186 (pre-cleaved), Pro186 cleavage product, and a PRO186 active dimer. [Figure 65B] 1 shows an exemplary embodiment of a construct described herein, including Pro186 (pre-cleaved), Pro186 cleavage product, and a PRO186 active dimer. [Figure 65C] 1 shows an exemplary embodiment of a construct described herein, including Pro186 (pre-cleaved), Pro186 cleavage product, and a PRO186 active dimer. [Figure 66] 1 provides an illustration of COBRA conversion to an active dimer upon protease cleavage. [Figure 67] Figures 67A-67B provide characterization of COBRA binding. Figure 67A shows binding activity to human, cynomolgus monkey, and mouse antibodies. Figure 67B shows PRO186 binding to human CD3 epsilon, active PRO186 binding of human CD3 epsilon, and active PRO186 binding of human EGFR. [Figure 68] Figures 68A-68B show the cleavage of the PRO186 linker by MMP2 and MMP9. Figure 68A shows a Western blot of active binding product molecules upon cleavage. Figure 68B shows the accumulation of active binding product molecules versus cleavage time. [Figure 69] Figure 69 shows the in vitro activity of conditional PRO186 constructs. The left panel of Figure 69 shows the results of a T cell killing assay. The right panel of Figure 69 shows the level of IFN-gamma release in relation to the concentration of the test article. [Figure 70] EGFR expression compared to activity in three tumor cell lines: LoVo (a colorectal cancer (CRC) cell line), HT-29 (a colorectal cancer (CRC) cell line), and SCC25 (a head and neck cancer cell line). [Figure 71] Figures 71A and 71B show the expression of EGFR, MMP2, and MMP9 on tumor cells and tumor xenografts. Figure 71A shows the EGFR cell surface density on three cancer cell lines: LoVo, HT-29, and SCC25. Figure 71B shows immunohistochemical staining of EGFR, MMP2, and MMP9 on tumor xenografts. [Figure 72] 1 provides a schematic diagram of the experimental procedure of the adoptive human T cell transfer model in tumor-bearing mice. [Figure 73] Figure 73 shows regression of established solid tumors in mice by PRO186. The left panel of Figure 73 shows regression of a LoVo-derived tumor. The center panel of Figure 73 shows regression of a HT-29-derived tumor. The right panel of Figure 73 shows regression of a SCC25-derived tumor. [Figure 74] Figures 74A-B show that cleaved PRO186 is cleared more rapidly than intact (uncleaved) PRO186. Figure 74A shows the pharmacokinetics of the test article in the plasma of non-tumor-bearing mice. Figure 74B shows the tumor volume of LoVo-derived tumors in mice administered the test article. [Figure 75]

[0023] Figure 1 shows the results of a T-cell dependent cytotoxicity (TDCC) assay using Pro233 and MMP9-cleaved Pro233. Pro233 is dependent on the humanized anti-EGFR binding domain. The results show that cleaved Pro233 exhibits efficacy compared to the uncleaved form on EGFR-expressing HT29 cells. [Figure 76]

[0039] Figure 1 shows 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 potent than the uncleaved version and the uncleaved control on EpCAM expressing HT29 tumor cells. [Figure 77] Figure 1 shows the results of a TDCC assay using Pro225, Pro225 cleaved by MMP9, and Pro295 (uncleavable control). Pro225 is dependent on the hF7 anti-B7H3 binding domain. The results show that cleaved Pro566 is more potent than the uncleavable version and the uncleavable control on B7H3 expressing HT29 tumor cells (note that HT29 expresses EGFR, B7H3, and EpCAM). [Figure 78]

[0039] Figure 1 shows 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 potent than the uncleaved version and the uncleaved control on EpCAM expressing HT29 tumor cells. [Figure 79]

[0023] Figure 1 shows the results of a TDCC assay using a dual-targeting construct for EGFR and EpCAM using the EGFR2-binding domain and the hVIB664 EpCAM-binding domain (sometimes referred to herein as a "hetero-COBRA"), demonstrating that MMP9-cleaved Pro623 was more potent than either uncleaved Pro623 or the uncleaved control, Pro625. [Figure 80] 1 shows the results of a TDCC assay using a dual-targeting construct for EGFR and EpCAM using the EGFR2-binding domain and the hVIB665 EpCAM-binding domain, demonstrating that MMP9-cleaved Pro624 was more potent than either uncleaved Pro624 or the uncleaved control, Pro626. [Figure 81]

[0039] Figure 1 shows the results of a TDCC assay using a dual-targeting construct for EGFR and EpCAM (reverse orientation from Pro624) using the hEGFR2 binding domain and the hVIB665 EpCAM binding domain. The results show that MMP9-cleaved Pro698 was more potent than either uncleaved Pro698 or the uncleaved control Pro699. [Figure 82] 1 shows 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-cleaved Pro665 was more potent than either uncleaved Pro665 or the uncleaved control, Pro659. [Figure 83]1 shows the results of a TDCC assay using a dual-targeting construct for B7H3 and EpCAM (reverse orientation from Pro655) using the hF7 B7H3-binding domain and the hVIB664 EpCAM-binding domain at Pro657. The results show that MMP9-cleaved Pro657 was more potent than either uncleaved Pro657 or the uncleavable control, Pro661. [Figure 84] 1 shows 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-cleaved Pro656 was more potent than either uncleaved Pro656 or the uncleavable control, Pro660. [Figure 85] 1 shows the results of a TDCC assay using a dual-targeting construct for B7H3 and EpCAM (reverse orientation from Pro656) using the hF7 B7H3-binding domain and the hVIB665 EpCAM-binding domain at Pro658. The results show that MMP9-cleaved Pro658 was more potent than either uncleaved Pro658 or the uncleavable control, Pro662. [Figure 86] Experiments are shown here demonstrating that the dual-targeting constructs Pro656 and Pro658 (with different orientations of the two domains), which bind both B7H3 and EpCAM, kill all three cell types: those expressing B7H3 but not EpCAM, those expressing EpCAM but not B7H3, and those expressing both. In contrast, Pro225 (with two anti-B7H3 domains) kills only two cell types: those expressing B7H3 and those expressing both B7H3 and EpCAM. Similarly, Pro566 (with two anti-EpCAM domains) kills only two cell types: those expressing EpCAM and those expressing both B7H3 and EpCAM. Raji F cell lines transiently expressing the appropriate proteins were used. [Figure 87]

[0033] Figure 1 shows the results of a TDCC assay using the B7H3-targeting construct Pro226. The results show that Pro226 and MMP9-cleaved Pro226 exhibit greater activity than the uncleaved control, Pro296. Pre-cleaved Pro226 exhibits an EC50 of approximately 8 pM. [Figure 88] Figure 1 shows the results of a TDCC assay using Pro226 expressed in the presence of tunicamycin, which improves the potency of Pro226. Tunicamycin prevents glycosylation and increases the EC50 of the cleavage product to 1 pM, indicating that glycosylation decreases the EC50 (increasing potency). [Figure 89] 1 shows the results of a TDCC assay using Pro664, which contains the anti-B7H3 hF12 domain with an amino acid mutation (S59Y) to remove the glycosylation site, compared to Pro226 (the same construct but without the amino acid mutation). Pro664 shows similar activity to Pro226. [Figure 90] 1 shows the results of a TDCC assay using Pro665, which contains the anti-B7H3 hF12 domain with an amino acid mutation (N57Q) to remove the glycosylation site, compared to Pro226 (the same construct but without the amino acid mutation). Pro665 shows higher potency than Pro226. [Figure 91] 1 shows the results of a TDCC assay using Pro667, which contains the anti-B7H3 hF12 domain with an amino acid mutation (N57E) to remove the glycosylation site, compared to Pro226 (the same construct but without the amino acid mutation). Pro667 shows lower potency than Pro226. [Figure 92] 1 shows the results of a TDCC assay using Pro694, which contains the anti-B7H3 hF12 domain with an amino acid mutation (S59A) to remove the glycosylation site, compared to Pro226 (the same construct but without the amino acid mutation). Pro694 shows higher potency than Pro226. [Figure 93]1 shows the results of a TDCC assay using Pro695, which contains the anti-B7H3 hF12 domain with an amino acid mutation (N57D) to remove the glycosylation site, compared to Pro226 (the same construct but without the amino acid mutation). Pro695 shows similar potency to Pro226. [Figure 94] Figure 1 shows a TDCC assay comparing two different inactivators in the inactive domain: Pro186 has Flag inactivation in both Vhi and Vli, and Pro476 has i2 inactivation, which show similar potency. [Figure 95] A TDCC assay comparing Pro186, which contains an MMP9 cleavage site, with Pro393, which contains an S9 cleavage site, is shown, and the results show that both constructs behave similarly with different protease cleavage sites. [Figure 96] A TDCC assay comparing Pro186, which contains an MMP9 cleavage site, and Pro394, which contains an ST14 MV cleavage site, is shown, and the results show that both constructs behave similarly with different protease cleavage sites. [Figure 97] A TDCC assay comparing Pro186, which contains an MMP9 cleavage site, with Pro395, which contains a CathS cleavage site, is shown, and the results show that both constructs behave similarly with different protease cleavage sites. [Figure 98] A TDCC assay comparing Pro186, which contains the MMP9 cleavage site, with Pro396, which contains the MMP9v cleavage site, is shown, and the results indicate that both constructs behave similarly with different protease linker sequences. Note that the MMP9 site is cleaved by both MMP9 and CathS, whereas MMP9v is MMP9-specific. [Figure 99] A TDCC assay comparing Pro186, which contains the MMP9 cleavage site, and Pro429, which contains the MepGzb cleavage site, is shown, and the results show that both constructs behave similarly with different protease cleavage sites. [Figure 100]A TDCC assay comparing Pro186, which contains the MMP9 cleavage site, and Pro430, which contains the MMP9-2 cleavage site, is shown, which behave similarly despite the different MMP9 cleavage sites. [Figure 101] A TDCC assay comparing Pro186, which contains the MMP9 cleavage site, with Pro431, which contains the ST14 MS cleavage site, is shown, and the results show that both constructs behave similarly with different protease cleavage sites. [Figure 102] TDCC assay using the Trop2-containing construct, Pro676, shows that Pro677 and its active dimer (Pro684AD) exhibit greater activity against BXPC3 tumor cells than the uncleavable control, Pro680. The active dimer is generated by expressing a self-dimerizing active domain. [Figure 103] 1 shows a TDCC assay using the Trop2-containing construct, Pro677, demonstrating that Pro677 and its active dimer (Pro685AD) exhibit greater activity against BXPC3 tumor cells than the uncleavable control, Pro681. [Figure 104] Figure 1 shows a TDCC assay using the Trop2-containing construct, Pro677, demonstrating that Pro677 and its active dimer (Pro685AD) exhibit greater activity against HT29 tumor cells than the uncleavable control, Pro681. Pro677 exhibits less activation by HT29 cells than BXPC3 cells. [Figure 105] 1 shows a TDCC assay using the Trop2-containing construct, Pro678, demonstrating that Pro678 and its active dimer (Pro686AD) exhibit greater activity against BXPC3 tumor cells than the uncleavable control, Pro682. [Figure 106] 1 shows a TDCC assay using the Trop2-containing construct Pro679, demonstrating that Pro679 and its active dimer (Pro687 active domain, AD) exhibit greater activity against BXPC3 tumor cells than the uncleavable control, Pro683. [Figure 107] 1 shows a TDCC assay using the Trop2-containing construct, Pro808, demonstrating that Pro808, its cleaved form and the active dimer (Pro810AD) exhibit greater activity against BXPC3 tumor cells than the uncleaved control, Pro809. [Figure 108] Figure 1 shows a TDCC assay using Pro808, a construct containing Trop2, in HT29 tumor cells. Pro810AD (active domain) shows greater activity in the TDCC assay than the uncleaved control, Pro809. Full-length Pro808 shows less activation in the assay with HT29 cells. [Figure 109] 1 shows a TDCC assay using the Trop2-containing construct, Pro819, demonstrating that Pro819 and its active dimer (Pro821AD) exhibit greater activity against BXPC3 tumor cells than the uncleavable control, Pro820. [Figure 110] Figure 13 shows a TDCC assay using the Trop2-containing construct, Pro819, demonstrating that the active dimer (Pro821AD) exhibits greater activity than the uncleaved control. Full-length Pro819 exhibits less activation by HT29 cells when compared to the uncleaved control, Pro820. For both pairs in Figures 107 and 108, and Figures 109 and 110, the data, combined with data showing that addition of an MMP9 inhibitor reduces activity, suggest that different target cell types produce different levels of MMP activity (see Figure 133). [Figure 111] 1 shows a TDCC assay using the FOLR1 construct, Pro311, demonstrating that Pro311 and MMP9-cleaved Pro311 exhibit greater activity against H292 tumor cells than the Pro299 uncleavable control. [Figure 112] 1 shows a TDCC assay using the FOLR1 construct, Pro312, demonstrating that Pro312 and MMP9-cleaved Pro312 exhibit greater activity against H292 tumor cells than the Pro303 uncleavable control. [Figure 113] 1 shows the results of a TDCC assay using the dual-targeting construct Pro420 with FOLR1 and EGFR targeting. The results show that Pro420 and MMP9-cleaved Pro420 exhibit greater activity than the uncleavable control, Pro299. [Figure 114] 1 shows the results of a TDCC assay using the dual-targeting construct Pro421 (reverse orientation from Pro420) with FOLR1 and EGFR targeting. The results show that Pro421 and MMP9-cleaved Pro421 exhibit greater activity than the uncleavable control, Pro299. [Figure 115] 1 shows the results of a TDCC assay using the dual-targeting construct Pro551, which targets EGFR and FOLR1. The results show that Pro551 and MMP9-cleaved Pro551 exhibit greater activity than the uncleavable control, Pro550. [Figure 116] 1 shows the results of a TDCC assay using the dual-targeting construct Pro552 (reverse orientation from Pro551) with FOLR1 and EGFR targeting. The results show that Pro522 and MMP9-cleaved Pro522 exhibit greater activity than the uncleavable control, Pro303. [Figure 117] 1 shows the results of a TDCC assay using the dual targeting construct Pro254, where each targeting domain binds to a different epitope on EGFR and is potent against EGFR-expressing HT29 cells. [Figure 118] TDCC assay results are shown using two different dual-targeting constructs that both bind to B7H3 but also use two different targeting domains: Pro479 and Pro480, which are identical except for the orientation of the two binding domains, and both are active after cleavage by MMP9 on B7H3-expressing HT29 cells. [Figure 119] 1 shows the results of a TDCC assay using Pro233 and cleaved Pro233 on EGFR-expressing HT-29 cells. Pro233 shows good conditioning and good potency when cleaved by MMP9. [Figure 120] Results of a tumor regression study using Pro225, a B7H3-targeting and MMP9-linker construct, are shown. The study was conducted using a human PBMC engraftment model. NSG-β2M- / - mice (Jackson) were intravenously engrafted with human PBMCs. Three days after engraftment, the mice were subcutaneously implanted with tumor cell lines. Once tumor growth was established, mice were randomized based on tumor volume and administered test articles intravenously as indicated. Tumor volume was assessed by caliper measurement. Results demonstrate that Pro225 regresses established solid tumors compared to the non-cleavable control, Pro295. [Figure 121]

[00130] Figure 120 shows the results of a tumor regression study using Pro226, a B7H3-targeted and MMP9-linked construct. The study was performed as shown in Figure 120, and the results show that Pro226 regresses established solid tumors compared to the non-cleavable control, Pro295. [Figure 122]

[0023] Figure 1 shows the results of a tumor regression study using Pro565 and Pro566, both with EpCAM targeting and an MMP9 linker. The results show that both Pro565 and Pro566 exhibit anti-tumor responses compared to the non-cleavable control, Pro568. [Figure 123] Figure 123 shows the results of a tumor regression study using Pro393 utilizing EGFR targeting and an S9 linker using the protocol described in Figure 122. The results show that Pro393 regresses established solid tumors compared to the non-cleavable control, Pro214. [Figure 124] Figure 122 shows the results of a tumor regression study using Pro394 targeting EGFR and utilizing the ST14 MV linker, using the protocol described in Figure 122. The results show that Pro394 exhibits a smaller anti-tumor response compared to the non-cleavable control, Pro214. [Figure 125]Figure 123 shows the results of a tumor regression study using Pro395 utilizing EGFR targeting and a CathS linker using the protocol described in Figure 122. The results show that Pro395 regresses established solid tumors compared to the non-cleavable control, Pro214. [Figure 126] Figure 123 shows the results of a tumor regression study using Pro396 utilizing EGFR targeting and an MMP9v linker using the protocol described in Figure 122. The results demonstrate that Pro396 exhibits an anti-tumor response compared to the non-cleavable control, Pro214. [Figure 127] Figure 123 shows the results of a tumor regression study using Pro430, which utilizes EGFR targeting and an MMP9-2 linker, using the protocol described in Figure 122. The results show that Pro430 regresses established solid tumors compared to the non-cleavable control, Pro214. [Figure 128] Figure 122 shows the results of a tumor regression study using Pro431, which utilizes EGFR targeting and the ST14 MS linker, using the protocol described in Figure 122. The results show that Pro431 exhibits an anti-tumor response compared to the non-cleavable control, Pro214. [Figure 129] Figure 123 shows the results of a tumor regression study using Pro476, which utilizes EGFR targeting, an MMP9-2 linker, and the inactivation domains Vli2 and VHi2, using the protocol described in Figure 122. The results show that Pro476 regresses established solid tumors compared to the non-cleavable control, Pro214. [Figure 130] Figure 122 shows the results of a tumor regression study using Pro517 utilizing EGFR targeting and an MMP9-2 linker, using the protocol described in Figure 122. The results demonstrate that Pro517 regresses established solid tumors compared to the non-cleavable control, Pro513. [Figure 131]Figure 120 shows the results of a tumor regression study using Pro664, which utilizes B7H3 targeting and an MMP9 linker, using the protocol described in Figure 120. The results show that Pro664 regresses established solid tumors compared to the non-cleavable control, Pro766. [Figure 132] Figure 123 shows the results of a tumor regression study using Pro676, which utilizes Trop2 targeting and an MMP9 linker, using the protocol described in Figure 122. The results show that Pro676 regresses established solid tumors compared to the non-cleavable control, Pro681. [Figure 133] Results from studies using Pro225 (an anti-B7H3 construct with an MMP9 linker) and increasing amounts of the MMP-specific inhibitor batimastat are presented, demonstrating that the potency of uncleaved Pro225 is reduced by batimastat, demonstrating in cellular in-assay COBRA cleavage and activation. DETAILED DESCRIPTION OF THE INVENTION

[0032] introduction The present invention relates to a method 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 by targeting normal tissues. Therefore, there is a need to activate the binding capacity of therapeutic molecules only near affected tissues to avoid normal tissue interactions. Therefore, the present invention relates to multivalent, conditionally effective ("MCE") proteins that have multiple functional protein domains. Generally, one of these domains is an antigen-binding domain (ABD) that binds to a target tumor antigen (TTA), and another is an ABD that binds to a T cell antigen, such as CD3, under certain conditions. In addition, the MCE protein also contains one or more protease cleavage sites. In other words, the therapeutic molecule is produced in a "prodrug"-like format that is inactive until the CD3-binding domain is exposed to the tumor environment. The tumor environment contains proteases, and upon exposure to the protease, the prodrug is cleaved and becomes active.

[0033] This is generally achieved herein 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, that constrains the CD3 Fv of the MCE into an inactive form as discussed herein. The TTA targets the MCE in the vicinity of the tumor, thus exposing it to proteases. Upon cleavage, the active variable heavy and light domains can now pair to form one or more active ABDs against CD3, thus recruiting T cells to the tumor and leading to therapy.

[0034] Generally, CD3 binding domains ("Fv") are constrained, and the linker between the active variable heavy domain and the active variable light domain that traditionally forms an Fv is too short to allow the two active variable domains to bind to each other; this is referred to as a "constrained linker," which can be constrained and cleavable (CCL, as used in Format 1) or not (CNCL, as used in Format 2). Rather, in the prodrug (e.g., uncleaved) format, the prodrug polypeptide also contains a "pseudo-Fv domain." The pseudo-Fv domain contains variable heavy and light domains with standard framework regions but "inactive" or "inactive" CDRs. The pseudo-Fv domain also has a constrained linker between the inactive variable heavy domain and the inactive variable light domain. Because neither the Fv domain nor the pseudo-Fv domain can self-assemble due to steric hindrance, intramolecular assembly occurs, pairing the aVL with the iVH and the aVH with the iVL, depending on the affinity of their respective framework regions. However, due to the "inactive" CDRs of the pseudodomain, the resulting ABD does not bind to CD3, thus preventing toxicity outside the affected tissue, such as the tumor. However, in the presence of a protease within or near the tumor, the prodrug construct is cleaved to release the pseudo-Fv domain from the surface, thus allowing the "true" variable heavy and variable light domains to associate intermolecularly (e.g., the two cleaved constructs come together), thus causing active CD3 binding and resulting tumor efficacy. These constructs are generally referred to herein as conditional bispecific redirected activation constructs or "COBRA™." The stability of the intramolecular assembly is demonstrated herein by conditioning experiments, where, in the absence of proteases, the uncleaved construct has no activity (e.g., no active CD3-binding domain is formed).

[0035] Interestingly, while these constructs are all referred to herein as "constrained" for ease of explanation, further studies have shown that intramolecular assembly is preferred even when one of the Fv domains is unconstrained, e.g., one of the domains can have a longer, more flexible linker. That is, as shown in Figures 37-39, intramolecular assembly still occurs when only one of the Fv domains, either the domain with active VL and VH, or the pseudo-Fv domain, is constrained (e.g., the uncleaved construct is inactive in the absence of protease cleavage). However, in this system, the protein has better expression when both linkers are constrained. However, as will be understood by those skilled in the art, any of the Format 1, Format 2, or Format 4 constructs herein can have one of these Fv domains with an "unconstrained" or "flexible" linker. For ease of reference, the constructs are shown in formats in which both Fv domains are constrained.

[0036] The constructs and formats of the present invention are variations of the invention described in WO2017 / 156178, which is expressly incorporated herein by reference in its entirety. As shown in Figures 17-21, the previous constructs have the ability to isomerize to form both bivalent scFvs and single-chain diabodies due to the presence of two sets of VH and VL domains in a single polypeptide. Even after purification of each isoform, the bivalent constructs can still reach equilibrium with the diabody isoforms. The single-chain diabodies retain the ability to bind to CD3 in the absence of protease cleavage, reducing the utility of the constructs.

[0037] To solve this problem, the present invention provides four separate types of constructs to achieve this conditional activation. Prodrug activation can occur in one of four general ways, as generally shown in the figures. A "Form 1" mechanism is shown in Figure 1. In this embodiment, the prodrug construct has two cleavage sites: one between the VH and vL domains of the constrained Fv, thus freeing the two variable domains to associate; and a second cleavage site that releases the pseudo-Fv domain from the prodrug construct, leaving two molecules that associate via the inherent self-assembly of the variable heavy and variable light domains, each of which similarly contains an antigen-binding domain for a tumor antigen, thus enabling recruitment of T cells to tumor sites.

[0038] In an alternative embodiment, the prodrug construct is shown in the "Format 2" scheme in Figure 2. 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 format, the constrained pseudo-Fv domains, inactive VH and VL, associate with the VH and VL of the constrained Fv domain such that there is no CD3 binding. However, once cleavage occurs in the tumor environment, two different activating proteins, each containing the active variable heavy and light domains, associate to form two anti-CD3 binding domains. This Format 2 has two target tumor antigen binding domains ("TTA-ABD"), which can be identical (e.g., "homo-COBRA") or different (e.g., "hetero-COBRA"), as described more fully below. If different, they can each be directed against a different tumor antigen or against the same tumor antigen but different epitopes, as described more fully below.

[0039] In addition to the "single-chain protein" COBRA format discussed above, in which all of the components are contained on a single amino acid sequence, there are also constructs that rely on two proteins acting in pairs, "hemi-COBRA," as shown in Figure 3. In this embodiment, each protein has one active variable domain and one non-active 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 non-active domain is cleaved and the two active variable domains self-assemble to form the anti-CD3 binding domain.

[0040] Additionally, the present invention also provides "Format 4" constructs, as shown in Figure 4. These are similar to the "Format 2" designs, except that a single ABD is used for the TTA, which upon cleavage forms a tetravalent bispecific construct, where two of the prodrug molecules now contain two active anti-CD3 domains, as further described below.

[0041] Thus, the formats and constructs of the present invention are used in the treatment of disease.

[0042] definition In order that this application may be more fully understood, certain definitions are set forth below. Such definitions are intended to encompass grammatical equivalents.

[0043] As used herein, "amino acid" and "amino acid identity" refer to one of the 20 naturally occurring amino acids or any non-natural analogue that may be present at a particular, defined position. In many embodiments, "amino acid" refers to one of the 20 naturally occurring amino acids. "Protein" herein refers to at least two covalently attached amino acids, including proteins, polypeptides, oligopeptides, and peptides.

[0044] As used herein, "amino acid modification" refers to an amino acid substitution, insertion, and / or deletion in a polypeptide sequence, or a modification to a moiety chemically linked to a protein. For example, the modification may be a modified carbohydrate or PEG structure attached to the protein. For clarity, unless otherwise specified, the amino acid modification is always to an amino acid encoded by DNA, e.g., the 20 amino acids for which codons exist in DNA and RNA. A preferred amino acid modification herein is a substitution.

[0045] As used herein, an "amino acid substitution" or "substitution" refers to the replacement of an amino acid at a particular position in a parent polypeptide sequence with a different amino acid sequence. Specifically, in some embodiments, the substitution is for an amino acid that does not naturally occur at the particular position, either in that organism or in any organism. For clarity, an engineered protein that changes the nucleic acid coding sequence but does not change the starting amino acid (e.g., replacing CGG (which encodes arginine) with CGA (which still encodes arginine) to increase host organism expression levels) is not an "amino acid substitution"; that is, if the protein has the same amino acid at the particular position from which it started, despite the generation of a new gene encoding the same protein, it is not an amino acid substitution.

[0046] As used herein, "amino acid insertion" or "insertion" refers to the addition of an amino acid sequence at a particular position in a parent polypeptide sequence.

[0047] As used herein, an "amino acid deletion" or "deletion" refers to the removal of an amino acid sequence at a particular position in a parent polypeptide sequence.

[0048] The polypeptides of the present invention specifically bind to CD3 and target tumor antigens (TTA), such as target cell receptors, as outlined herein. "Specific binding" to an antigen or epitope, or "specifically binding to" or "specific for" it, refers to binding that is measurably different from non-specific interactions. Specific binding can be measured, for example, by determining the binding of a molecule compared to the binding of a control molecule, which is generally a molecule of similar structure that does not have binding activity. For example, specific binding can be determined by competition with a control molecule that is similar to the target.

[0049] Specific binding to a particular antigen or epitope is, for example, at least about 10 -4 M, at least about 10 -5 M, at least about 10 -6 M, at least about 10 -7 M, at least about 10 -8 M, at least about 10 -9 M, or at least about 10 -10 M, at least about 10 -11 M, at least about 10 -12 This can be demonstrated by an antibody having a KD for an antigen or epitope of M or greater, where KD refers to the dissociation rate of a particular antigen-antibody interaction. Typically, an antibody that specifically binds to an antigen has a KD that is 20, 50, 100, 500, 1000, 5,000, 10,000, or more times greater than that of a control molecule compared to the antigen or epitope.

[0050] Specific binding to a particular antigen or epitope can also be exhibited, for example, by an antibody having a K or K for the antigen or epitope that is at least 20, 50, 100, 500, 1000, 5,000, 10,000, or more times greater than a control, where K or K refers to the association rate of a particular antibody-antigen interaction. Binding affinity is generally measured using a Biacore assay or Octet, as known in the art.

[0051] As used herein, "parent polypeptide" or "precursor polypeptide" (including Fc parent or precursor) refers to a polypeptide that is subsequently modified to generate a variant. The parent polypeptide may be a naturally occurring polypeptide or a variant or engineered version of a naturally occurring polypeptide. Parent polypeptide can refer to the polypeptide itself, a composition comprising the parent polypeptide, or the amino acid sequence that encodes it. Thus, as used herein, "parent Fc polypeptide" refers to an unmodified Fc polypeptide that is modified to generate a variant, and as used herein, "parent antibody" refers to an unmodified antibody that is modified to generate a variant antibody.

[0052] As used herein, "position" means a location in the sequence of a protein. Positions may be numbered sequentially or according to an established format, such as the EU index for antibody numbering.

[0053] As used herein, "target antigen" means a molecule that is specifically bound by the variable region of a given antibody. A target antigen may be a protein, carbohydrate, lipid, or other compound. A wide range of suitable exemplary target antigens are described herein.

[0054] As used herein, "target cell" refers to a cell that expresses a target antigen. Generally, for purposes of the present invention, a target cell is either a tumor cell that expresses a TTA or a T cell that expresses the CD3 antigen.

[0055] As used herein, "Fv" or "Fv domain" or "Fv region" generally refers to a polypeptide comprising the VL and VH domains of the antigen-binding domain from an antibody. Fv domains typically form an "antigen-binding domain" or "ABD" as discussed herein when they contain active VH and VL domains (although in some cases, Fvs containing a constrained linker are used, whereby an active ABD is not formed prior to cleavage). As discussed below, Fv domains can be organized in many ways within the present invention and can be "active" or "inactive," such as in scFv, constrained Fv, or pseudo-Fv formats. It should be understood that in some cases, an Fv domain is composed of VH and VL domains on a single polypeptide chain, as shown in Figures 1 and 2, but with a constrained linker such that an intramolecular ABD cannot form. In these embodiments, two active ABDs are formed after cleavage. In some cases, an Fv domain is composed of VH and VL domains, one of which is inactive, whereby an intermolecular ABD is formed only after cleavage. As discussed below, Fv domains can be organized in many ways in the present invention and can be "active" or "inactive", in scFv format, constrained Fv format, pseudo-Fv format, etc. Additionally, as discussed herein, Fv domains containing a VH and a VL can be / form an ABD, and other ABDs that do not contain VH and VL domains can be formed using sdABDs.

[0056] By "variable domain" herein is meant a region of an immunoglobulin that comprises one or more Ig domains substantially encoded by any of the Vκ, Vλ, and / or VH genes that comprise the kappa, lambda, and heavy chain immunoglobulin loci, respectively. In some cases, a single variable domain may be used, such as an sdFv (also referred to herein as an sdABD).

[0057] In embodiments utilizing both variable heavy (VH) and variable light (VL) domains, each VH and VL consists of three hypervariable regions ("complementarity determining regions," "CDRs") and four "framework regions" or "FRs," arranged from amino-terminus to carboxy-terminus in the following order: FR1-CDR1-FR2-CDR2-FR3-CDR3-FR4. Thus, a VH domain has the structure vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCDR3-vhFR4, and a 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, in the prodrug formats of the present invention, there are "constrained Fv domains" in which the VH and VL domains are unable to self-associate, and "pseudo-Fv domains" in which the CDRs do not form an antigen-binding domain when self-associated.

[0058] The hypervariable regions confer antigen-binding specificity and generally comprise approximately amino acid residues 24-34 (LCDR1, "L" indicates the light chain), 50-56 (LCDR2), and 89-97 (LCDR3) in the light chain variable region, and approximately 31-35B (HCDR1, "H" indicates the heavy chain), 50-65 (HCDR2), and 95-102 (HCDR3) in the heavy chain variable region (Kabat et al., SEQUENCES OF PROTEINS OF IMMUNOLOGICAL INTEREST, 5th Ed. Public Health Service, National Institutes of Health, Health, Bethesda, Md. (1991)), and / or the amino acid residues forming the hypervariable loops in the light chain variable region (e.g., residues 26-32 (LCDR1), 50-52 (LCDR2), and 91-96 (LCDR3) in the light chain variable region, and residues 26-32 (HCDR1), 53-55 (HCDR2), and 96-101 (HCDR3) in the heavy chain variable region (Chothia and Lesk (1987) J. Mol. Biol. 196:901-917). Particular CDRs of the invention are described below.

[0059] As will be understood by those skilled in the art, the exact numbering and arrangement of CDRs may vary among different numbering systems. However, it should be understood that the disclosure of a variable heavy and / or variable light sequence includes the disclosure of the associated (unique) CDRs. Thus, the disclosure of each variable heavy region is a disclosure of the vhCDRs (e.g., vhCDR1, vhCDR2, and vhCDR3), and the disclosure of each variable light region is a disclosure of the vlCDRs (e.g., vlCDR1, vlCDR2, and vlCDR3).

[0060] For a useful comparison of CDR numbering, see Lafranc et al., Dev. Comp. Immunol. 27(1):55-77 (2003). Table 1 [Table 1]

[0061] Throughout this specification, the Kabat numbering system is generally used when referring to variable domain residues (approximately residues 1-107 in light chain variable regions and residues 1-113 in heavy chain variable regions), with the EU numbering system being used for the Fc region (see, e.g., Kabat et al., supra (1991)).

[0062] The present invention provides many different CDR sets. In this case, a "complete CDR set" in the context of an anti-CD3 component includes 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, the VH and VL CDRs, are capable of binding antigen both individually and as a set. For example, in a constrained Fv domain, the vhCDRs can bind, for example, to CD3, and the vlCDRs can bind to CD3, but in the constrained format, they cannot bind to CD3.

[0063] In the context of a single domain ABD ("sdABD") as generally used herein to bind a target tumor antigen (TTA), the CDR set is only three CDRs, which are sometimes also referred to in the art as a "VHH" domain.

[0064] These CDRs may, by design, be part of a larger variable light or variable heavy domain. In addition, as more fully outlined herein, the variable heavy and variable light domains may be on separate polypeptide chains, or, in the case of scFv sequences, on a single polypeptide chain, depending on the format and organization of the moieties herein.

[0065] CDR contributes to the formation of antigen binding site, more specifically epitope binding site. "Epitope" refers to the determinant that interacts with specific antigen binding site in variable region known as paratope. Epitope is a group of molecules such as amino acids or sugar side chains, and usually has specific structural characteristics and specific charge characteristics. A single antigen can have two or more epitopes.

[0066] An epitope may include amino acid residues that are directly involved in binding (also called the immunodominant component of the epitope) and other amino acid residues that are not directly involved in binding, e.g., amino acid residues that are effectively blocked by a particular antigen-binding peptide, in other words, amino acid residues that lie within the footprint of a particular antigen-binding peptide.

[0067] Epitopes can be either conformational or linear. Conformational epitopes are produced by spatially juxtaposed amino acids from different segments of a linear polypeptide chain. Linear epitopes are produced by adjacent amino acid residues in the polypeptide chain. Conformational and nonconformational epitopes can be distinguished in that the binding to the former, but not the latter, is lost in the presence of denaturing solvents.

[0068] An epitope typically contains at least three, and usually at least five or eight to ten amino acids in a unique spatial conformation. Antibodies that recognize the same epitope can be identified by simple immunoassays, such as "binning," that demonstrate the ability of one antibody to block the binding of another antibody to a target antigen. As outlined below, the present invention encompasses not only the antigen-binding domains and antibodies enumerated herein, but also those that compete for binding to the epitope bound by the enumerated antigen-binding domains.

[0069] The variable heavy and variable light domains of the present invention may be "active" or "inactive."

[0070] As used herein, "inactive VH" ("iVH") and "inactive VL" ("iVL") refer to pseudo-Fv domain components that, when paired with their cognate VL or VH partner, respectively, form the resulting VH / VL pair that does not specifically bind to the antigen that an "active" VH or "active" VL would bind if bound to a similar VL or VH that is not "inactive." Exemplary "inactive VH" and "inactive VL" domains are formed by mutation of a wild-type VH or VL sequence, as outlined more fully below. Exemplary mutations are within CDR1, CDR2, or CDR3 of the 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, when paired with its "active" cognate partner, i.e., VL or VH, respectively, is capable of specifically binding to its target antigen. Thus, it should be understood that the pseudo-Fv can be a VH / iVL pair, an iVH / VL pair, or an iVH / iVL pair.

[0071] In contrast, as used herein, the term "active" refers to a CD3-binding domain that is capable of specifically binding to CD3. This term is used in two contexts: (a) when referring to a single member of an Fv binding pair (i.e., VH or VL) that is paired with its cognate partner and is capable of specifically binding to CD3, and (b) a cognate pair of sequences (i.e., VH and VL) that are capable of specifically binding to CD3. Exemplary "active" VH, VL, or VH / VL pairs are wild-type or parental sequences.

[0072] "CD-x" refers to a cluster of differentiation (CD) protein. In an exemplary embodiment, CD-x is selected from CD proteins involved in the recruitment or activation of T cells in a subject to which a polypeptide construct of the present invention has been administered. In an exemplary embodiment, CD-x is CD3, the sequence of which is shown in Figure 5.

[0073] The term "binding domain" in the context of the present invention characterizes a domain that (specifically) binds to / interacts with / recognizes a given target epitope or a given target site on a target molecule (antigen), for example, EGFR and CD-3, respectively. The structure and function of the target antigen-binding domain (which recognizes EGFR), and preferably the structure and / or function of the CD-3-binding domain (which recognizes CD3), are based on the structure and / or function of an antibody, for example, a full-length or entire immunoglobulin molecule containing the sdABD. According to the present invention, the target antigen-binding domain is generally characterized by the presence of three CDRs that bind to the target tumor antigen (generally referred to in the art as variable heavy domains, although no corresponding light chain CDRs are present). Alternatively, the 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 CD3-binding domain preferably also comprises at least the minimal structural requirements of an antibody to enable target binding. More preferably, the CD3-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). In exemplary embodiments, it is contemplated that the target antigen and / or CD3-binding domain is produced or obtainable by phage display or library screening methods.

[0074] As used herein, "domain" refers to a protein sequence having a function, as outlined herein. 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 extending domains.

[0075] By "domain linker" herein is meant an amino acid sequence that connects two domains as outlined herein. A domain linker can be a cleavable linker, a constrained cleavable linker, a non-cleavable linker, a constrained non-cleavable linker, an scFv linker, etc.

[0076] By "cleavable linker" ("CL") herein is meant an amino acid sequence that can be cleaved by a protease, preferably a human protease in affected tissue, as outlined herein. Cleavable linkers are generally at least three amino acids in length, with 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, or more amino acids being of use in the present invention, depending on the flexibility required. A number of cleavable linker sequences can be found in Figures 6 and 5.

[0077] By "non-cleavable linker" ("NCL") herein is meant an amino acid sequence that cannot be cleaved by human proteases under normal physiological conditions.

[0078] As used herein, a "constrained cleavable linker" ("CCL") refers to a short polypeptide containing a protease cleavage site (as defined herein) that links two domains as outlined herein in such a way that the two domains cannot significantly interact with each other until they are on different polypeptide chains, e.g., after cleavage. When a CCL links VH and VL domains as defined herein, the VH and VL cannot self-assemble to form a functional Fv prior to cleavage due to intramolecular steric hindrance (although they can assemble intermolecularly into a pseudo-Fv domain). Upon cleavage by the relevant protease, the VH and VL can assemble to form an active antigen-binding domain intermolecularly. Generally, CCLs are less than 10 amino acids in length, with 9, 8, 7, 6, 5, and 4 amino acids being of use in the present invention. Generally, protease cleavage sites are generally at least 4+ amino acids in length to confer sufficient specificity, as shown in FIG. 6.

[0079] By "constrained non-cleavable linker" ("CNCL") herein is meant a short polypeptide that connects two domains as outlined herein in a way that does not allow the two domains to significantly interact with each other, and that is not significantly cleaved by human proteases under physiological conditions.

[0080] By "constrained Fv domain" herein is meant an Fv domain comprising an active variable heavy domain and an active variable light domain covalently linked by a constrained linker as outlined herein in such a way that the active heavy and light variable domains cannot interact intramolecularly to form an active Fv that binds to an antigen, such as CD3. Constrained Fv domains are thus similar to scFvs, but are not capable of binding to antigen due to the presence of the constrained linker (although they can assemble intermolecularly with inactive variable domains to form pseudo-Fv domains).

[0081] By "pseudo-Fv domain" herein is meant a domain comprising a pseudo or inactive variable heavy domain or a pseudo or inactive variable light domain, or both, linked using a domain linker (which may be cleavable, constrained, non-cleavable, non-constrained, etc.). The iVH and iVL domains of the pseudo-Fv domain do not bind to human antigens either when associated with each other (iVH / iVL) or when associated with an active VH or VL; therefore, the iVH / iVL, iVH / VL, and iVL / VH Fv domains do not significantly bind to human proteins, and thus these domains are inactive in the human body.

[0082] By "single-chain Fv" or "scFv" herein is generally meant a variable heavy (VH) domain covalently linked to a variable light (VL) domain using a domain linker as discussed herein to form an scFv or scFv domain. The scFv domain can be in either an N-terminal to C-terminal orientation (VH-linker-VL or VL-linker-VH).

[0083] As used herein, "single-domain Fv," "sdFv," or "sdABD" refers to an antigen-binding domain having only three CDRs, generally 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 sdABDs are used herein: sdABDs that bind to TTA and are annotated as such (commonly referred to as sdABD-TTA, or sdABD-EGFR for those that bind to EGFR, sdABD-FOLR1 for those that bind to FOLR1, etc.), and sdABDs that bind to HSA ("sdABD-HSA" or "sdABD(1 / 2)")

[0084] "Protease cleavage site" refers to an amino acid sequence that is recognized and cleaved by a protease. Suitable protease cleavage sites are outlined below and shown in Figures 5 and 6.

[0085] As used herein, a "protease cleavage domain" refers to a "protease cleavage site," as well as the interplay between individual protease cleavage sites and between protease cleavage site(s) and other functional components of the constructs of the invention (e.g., V H , V L The term "protease cleavage domain" refers to a peptide sequence incorporating any linker between the protease cleavage domain(s), iVH, iVL, target antigen binding domain(s), half-life extending domain, etc. As outlined herein, the protease cleavage domain may also include additional amino acids as needed, for example, to provide flexibility.

[0086] The terms "COBRA™" and "conditional bispecific redirected activation" refer to a bispecific, conditionally effective protein having multiple 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 "hemi-COBRA™" refers to a conditionally effective protein that can bind to a T cell antigen when the variable heavy chain of the hemi-COBRA can associate with the variable light chain of another hemi-COBRA™ (a complementary hemi-COBRA™) due to its unique self-assembly when concentrated on the surface of a target-expressing cell.

[0087] Detailed Description of the Embodiments I. Fusion Proteins of the Invention The fusion proteins of the present invention have many different components, generally referred to herein as domains, linked together in various 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, a prodrug construct of the present invention can bind to TTA and CD3 and is therefore "bispecific." Proteins can also have higher specificity; for example, if one αTTA binds to EGFR and a second αTTA binds to EpCAM, and an anti-CD3 binding domain is present, this is a "trispecific" molecule. Similarly, the addition of an anti-HSA binding domain to this construct is "tetraspecific," as shown in Figure 3B.

[0088] As will be appreciated by those skilled in the art, the proteins of the present invention can have different valencies and can be multispecific, i.e., they can bind a target with more than one binding site, e.g., Pro140 is bivalent for EGFR.

[0089] The proteins of the invention may comprise a CD3 antigen-binding domain arranged in a variety of ways, such as with a tumor-targeting antigen-binding domain, a half-life extending domain, a linker, etc., as outlined herein.

[0090] A. CD3 antigen-binding domain The specificity of T cell responses is mediated by the recognition of antigens (presented in the major histocompatibility complex, MHC) by the T cell receptor complex. As part of the T cell receptor complex, CD3 is a protein complex present on the cell surface that includes the CD3γ (gamma) chain, the CD3δ (delta) chain, two CD3e (epsilon) chains, and two CD3ζ (zeta) chains. The CD3 molecule associates with the α (alpha) and β (beta) chains of the T cell receptor (TCR) to comprise the TCR complex. Clustering of CD3 on T cells, such as by Fv domains that bind to CD3, results in T cell activation, similar to T cell receptor engagement but independent of the specificity typical of that clone.

[0091] However, as is known in the art, CD3 activation can cause many toxic side effects, and therefore the present invention relates 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 that subsequently cleave the prodrug polypeptides of the present invention to provide active CD3-binding domains. Thus, in the present invention, binding of the anti-CD3 Fv domain to CD3 is regulated by a protease cleavage domain that limits binding of the CD3 Fv domain to CD3 only in the microenvironment of diseased cells or tissues that have high levels of proteases, for example, in the tumor microenvironment as described herein.

[0092] Thus, the present invention provides two sets of VH and VL domains, an active set (VH and VL) and an inactive set (iVH and iVL), all four of which are present in the prodrug construct. The construct is formatted so that the VH and VL sets cannot self-associate, but rather associate with inactive partners, e.g., iVH and VL and iVL and VH as shown herein.

[0093] 1. Active anti-CD3 variable heavy and variable light domains There are many suitable active CDR sets and / or VH and VL domains known in the art that can be used in the present invention, for example, CDRs and / or VH and VL domains from, for example, muromonab-CD-3 (OKT3), otelixizumab (TRX4), teplizumab (MGA031), visilizumab (Nuvion), SP34 or I2C, TR-66 or X35-3, VIT3, BMA030 (BW264 / 56), CLB-T3 / 3, CRIS7, YTH12 These antibodies are derived from known anti-CD3 antibodies such as .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.

[0094] In one embodiment, the VH and VL sequences that form the active Fv domain that binds 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.

[0095] 2. Inactive anti-CD3 variable heavy and variable light domains The inactive iVH and iVL domains contain "regular" framework regions (FRs) that allow them to associate with the active variable domains, rendering the pair inactive, e.g., unable to bind to CD3.

[0096] 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 human framework regions that allows self-assembly with another variable domain can be used, regardless of what amino acids are in the CDR positions within the variable region. For clarity, an inactive domain is said to contain the CDRs, although technically, an inactive variable domain does not confer binding capacity.

[0097] As is understood in the art, generating an inactive VH or VL domain is generally straightforward and can be accomplished in a variety of ways. In some embodiments, generating an inactive variable domain is generally accomplished by modifying one or more of the CDRs of an active Fv, including altering one or more of the three CDRs of the active variable domain. This can be accomplished by making one or more amino acid substitutions at functionally important residues in one or more CDRs, replacing some or all CDR residues with random sequences, replacing one or more CDRs with tag or flag sequences, and / or exchanging the CDRs and / or variable regions with those from an unrelated antibody (e.g., directed against a protein from a different organism).

[0098] In some cases, only one of the CDRs in a variable region may be modified to render it inactive, although other embodiments involve modification of 1, 2, 3, 4, 5, or 6 CDRs.

[0099] In some cases, the inactive domains can be engineered to promote selective binding in a prodrug format, favoring intramolecular iVH-VL and VH-iVL domain formation (e.g., rather than intermolecular pairing) prior to cleavage. See, e.g., Igawa et al., Protein Eng. Des. Selection 23(8):667-677 (2010), expressly incorporated herein by reference in its entirety, particularly with regard to interface residue amino acid substitutions.

[0100] In certain embodiments, the CD3-binding domain of the polypeptide constructs described herein not only exhibits strong CD3-binding affinity with human CD3, but also exhibits excellent cross-reactivity with individual cynomolgus monkey CD3 proteins. In some cases, the CD3-binding domain of the polypeptide construct is cross-reactive with CD3 from cynomolgus monkeys. In certain cases, the human:cynomolgus monkey KD ratio for CD3 is between 5 and 0.2.

[0101] In some embodiments, the CD3-binding domain of the antigen-binding protein can be any domain that binds to CD3, 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 CD3-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 CD3-binding domain of the antigen-binding protein to contain human or humanized residues from the antigen-binding domain of an antibody or antibody fragment.

[0102] Thus, in one aspect, the antigen-binding domain comprises a humanized or human binding domain. In one embodiment, the humanized or human anti-CD3 binding domain comprises one or more (e.g., all three) light chain complementarity determining region 1 (LC CDR1), light chain complementarity determining region 2 (LC CDR2), and light chain complementarity determining region 3 (LC CDR3) of a humanized or human anti-CD3 binding domain described herein, and / or one or more (e.g., all three) heavy chain complementarity determining region 1 (HC CDR1), heavy chain complementarity determining region 2 (HC CDR2), and heavy chain complementarity determining region 3 (HC CDR3) of a humanized or human anti-CD3 binding domain described herein, e.g., a humanized or human anti-CD3 binding domain comprising one or more, e.g., all three LC CDRs and one or more, e.g., all three HC CDRs.

[0103] In some embodiments, the humanized or human anti-CD3 binding domain comprises a humanized or human light chain variable region specific for CD3, wherein the light chain variable region specific for CD3 comprises human or non-human light chain CDRs within a human light chain framework region. In certain cases, the light chain framework region is a λ (lambda) light chain framework. In other cases, the light chain framework region is a κ (kappa) light chain framework.

[0104] In some embodiments, one or more CD-3 binding domains are humanized or fully human. In some embodiments, one or more activating CD-3 binding domains have a KD binding to CD-3 of 1000 nM or less on CD-3-expressing cells. In some embodiments, one or more activating CD-3 binding domains have a KD binding to CD-3 of 100 nM or less on CD-3-expressing cells. In some embodiments, one or more activating CD-3 binding domains have a KD binding to CD-3 of 10 nM or less on CD-3-expressing cells. In some embodiments, one or more CD-3 binding domains are cross-reactive with cynomolgus monkey CD-3. In some embodiments, one or more CD-3 binding domains comprise an amino acid sequence provided herein.

[0105] In some embodiments, the humanized or human anti-CD3 binding domain comprises a humanized or human heavy chain variable region specific for CD3, which heavy chain variable region specific for CD3 comprises human or non-human heavy chain CDRs within human heavy chain framework regions.

[0106] In one embodiment, the anti-CD3 binding domain is an Fv comprising a light chain and a heavy chain of the amino acid sequence provided herein. In one embodiment, the anti-CD3 binding domain comprises a light chain variable region comprising an amino acid sequence having at least one, two, or three modifications (e.g., substitutions) but not more than 30, 20, or 10 modifications (e.g., substitutions) of the amino acid sequence of a light chain variable region provided herein, or a sequence having 95-99% identity to an amino acid sequence provided herein, and / or a heavy chain variable region comprising an amino acid sequence having at least one, two, or three modifications (e.g., substitutions) but not more than 30, 20, or 10 modifications (e.g., substitutions) of the amino acid sequence of a heavy chain variable region provided herein, or a sequence having 95-99% identity to an amino acid sequence provided herein. In one embodiment, the humanized or human anti-CD3 binding domain is an scFv, in which a light chain variable region comprising an amino acid sequence described herein is attached to a heavy chain variable region comprising an amino acid sequence described herein via an scFv linker. The light and heavy chain variable regions of the scFv can be, for example, in either 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.

[0107] In some embodiments, the CD3 binding domain of the antigen binding protein has affinity for CD3 on CD3-expressing cells 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 CD3 binding domain of the antigen binding protein has affinity for CD3ε 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 further embodiments, the CD3 binding domain of the antigen binding protein has low affinity for CD3, i.e., about 100 nM or more.

[0108] Binding affinity to CD-3 can be determined 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., generally using, for example, Biacore or Octet assays, as known in the art. The binding activity of the antigen-binding protein of the present disclosure or its CD-3-binding domain to CD-3 can be assayed by immobilizing a ligand (e.g., CD-3) or the antigen-binding protein itself or its CD-3-binding domain to beads, a substrate, cells, etc. The agent can be added in an appropriate buffer, and the binding partner can be incubated for a period of time at a given temperature. After washing to remove unbound material, the binding protein can be released, for example, with a high pH SDS buffer, and analyzed, for example, by surface plasmon resonance (SPR).

[0109] In many embodiments, preferred active and inactive binding domains are those shown in Figure 5. Figure 5 shows one active VH and VL, and three inactive VHi and three inactive VLis, which have been inactivated in different ways.

[0110] As shown in Figure 5, a particularly useful pair of active anti-CD3 VL and VH domains has a VL with a vlCDR1 of SEQ ID NO: 127, a vlCDR2 of SEQ ID NO: 128, and a vlCDR3 of SEQ ID NO: 129, and a VH with a vhCDR1 of SEQ ID NO: 143, a vhCDR2 of SEQ ID NO: 144, and a vhCDR3 of SEQ ID NO: 145.

[0111] As shown in FIG. 5, a particularly useful pair of active anti-CD3 VL and VH domains has the VL of SEQ ID NO:126 and the VH of SEQ ID NO:142.

[0112] B. Antigen-binding domain for tumor target antigen In addition to the CD3 and half-life-extending domains described herein, the polypeptide constructs described herein also include targeting domains that bind 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 that each target antigen-binding domain binds to the target antigen on the target cell, thereby activating the CD3-binding domain to bind to T cells. Generally, the TTA-binding domains can bind to their targets before protease cleavage, so they can "wait" on the target cell to be activated as T cell guides. At least one target antigen is involved in and / or associated with a disease, disorder, or condition. Exemplary target antigens include those associated with proliferative diseases, neoplastic diseases, inflammatory diseases, immunological 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 tumor cells. 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 to healthy tissue.

[0113] In some embodiments, the target antigen is a cell surface molecule such as a protein, lipid, or polysaccharide, hi some embodiments, the target antigen is on a tumor cell, a virus-infected cell, a bacteria-infected cell, a damaged red blood cell, an arterial plaque cell, or a fibrotic cell.

[0114] Preferred embodiments of the present invention utilize sdABDs as targeting domains. These are preferred over scFv ABDs because the addition of other VH and VL domains to the constructs of the present invention can complicate the formation of pseudo-Fv domains.

[0115] In some embodiments, the prodrug constructs of the invention utilize a single TTA binding domain, generally as shown in Figure 3A as the sdABD-TTA pair and in Figure 4 as the "Format 4" configuration. While Figure 4 shows the use of a single anti-EGFR ABD, other TTA binding domains can be used.

[0116] In some embodiments, particularly in Format 1 and Format 2 constructs, the prodrug constructs of the present invention utilize two TTA ABDs, again preferably in an sdABD-TTA format. When dual targeting domains are used, they can bind to the same epitope on the same TTA. For example, as discussed herein, many of the constructs herein utilize two identical targeting domains. In some embodiments, two targeting domains that bind to different epitopes on the same TTA can be used, e.g., as shown in Figure 5, two EGFR sdABDs bind to different epitopes on human EGFR. In some embodiments, the two targeting domains bind to different TTAs, see e.g., the figure.

[0117] Polypeptide constructs contemplated herein comprise at least one antigen-binding domain, which 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.

[0118] (a) EGFR sdABD As shown in Figure 5, there are a number of particularly useful sdABDs that bind to human EGFR, referred to herein as "sdABD-EGFR" or "EGFRABD."

[0119] In one useful embodiment, the sdABD-EGFR1 has an sdCDR1 of SEQ ID NO: 10, an sdCDR2 of SEQ ID NO: 11, and an sdCDR3 of SEQ ID NO: 12. Optionally, the sdABD-EGFR has SEQ ID NO: 9.

[0120] In one useful embodiment, the sdABD-EGFR2a has an sdCDR1 of SEQ ID NO: 14, an sdCDR2 of SEQ ID NO: 15, and an sdCDR3 of SEQ ID NO: 16. Optionally, the sdABD-EGFR has SEQ ID NO: 13.

[0121] In one useful embodiment, sdABD-EGFR2d has an sdCDR1 of SEQ ID NO: 18, an sdCDR2 of SEQ ID NO: 19, and an sdCDR3 of SEQ ID NO: 20. Optionally, sdABD-EGFR has SEQ ID NO: 17.

[0122] (b) EpCAM sdABD As shown in Figure 5, there are a number of particularly useful sdABDs that bind to human EpCAM, referred to herein as "sdABD-EpCAM" or "EpCAMABD."

[0123] In one useful embodiment, sdABD-EpCAM h13 has an sdCDR1 of SEQ ID NO: 62, an sdCDR2 of SEQ ID NO: 63, and an sdCDR3 of SEQ ID NO: 64. In some cases, sdABD-EpCAM has SEQ ID NO: 61.

[0124] In one useful embodiment, sdABD-EpCAM h23 has an sdCDR1 of SEQ ID NO: 66, an sdCDR2 of SEQ ID NO: 67, and an sdCDR3 of SEQ ID NO: 68. Optionally, sdABD-EpCAM has SEQ ID NO: 65.

[0125] In one useful embodiment, sdABD-EpCAM hVIB665 has an sdCDR1 of SEQ ID NO: 70, an sdCDR2 of SEQ ID NO: 71, and an sdCDR3 of SEQ ID NO: 72. In some cases, the sdABD-EpCAM has SEQ ID NO: 69. Note that in contrast to the h13 and h23 EpCAM sdABDs, hVIB665 (also referred to as "acEpCAM hVIB665") binds to both the cleaved and uncleaved forms of EpCAM (which is known to undergo cleavage in vivo).

[0126] In one useful embodiment, sdABD-EpCAM hVIB666 has an sdCDR1 of SEQ ID NO: 74, an sdCDR2 of SEQ ID NO: 75, and an sdCDR3 of SEQ ID NO: 76. In some cases, the sdABD-EpCAM has SEQ ID NO: 73. Note that in contrast to the h13 and h23 EpCAM sdABDs, hVIB666 (also referred to as "acEpCAM hVIB666") binds to both the cleaved and uncleaved forms of EpCAM (which is known to undergo cleavage in vivo).

[0127] (c)B7H3 sdABD As shown in Figure 5, there are a number of particularly useful sdABDs that bind to human B7H3, referred to herein as "sdABD-B7H3" or "B7H3-ABD."

[0128] In one useful embodiment, sdABD-B7H3 hF7 has an sdCDR1 of SEQ ID NO: 34, an sdCDR2 of SEQ ID NO: 35, and an sdCDR3 of SEQ ID NO: 36. Optionally, sdABD-B7H3 has SEQ ID NO: 33.

[0129] In one useful embodiment, sdABD-B7H3 hF12 has an sdCDR1 of SEQ ID NO: 38, an sdCDR2 of SEQ ID NO: 39, and an sdCDR3 of SEQ ID NO: 40. Optionally, sdABD-B7H3 has SEQ ID NO: 37.

[0130] In one useful embodiment, sdABD-B7H3 hF12(N57Q) has an sdCDR1 of SEQ ID NO: 42, an sdCDR2 of SEQ ID NO: 43, and an sdCDR3 of SEQ ID NO: 44. Optionally, sdABD-B7H3 has SEQ ID NO: 41. In contrast to the hF7 and hF12 B7H3 sdABD, the amino acid substitution N57Q removes a glycosylation site.

[0131] In one useful embodiment, sdABD-B7H3 HF12(N57E) has an sdCDR1 of SEQ ID NO: 46, an sdCDR2 of SEQ ID NO: 47, and an sdCDR3 of SEQ ID NO: 48. Optionally, sdABD-B7H3 has SEQ ID NO: 45. In contrast to the hF7 and hF12 B7H3 sdABD, the amino acid substitution N57E removes a glycosylation site.

[0132] In one useful embodiment, sdABD-B7H3 hF12(N57D) has an sdCDR1 of SEQ ID NO: 50, an sdCDR2 of SEQ ID NO: 51, and an sdCDR3 of SEQ ID NO: 52. Optionally, sdABD-B7H3 has SEQ ID NO: 49. In contrast to the hF7 and hF12 B7H3 sdABD, the amino acid substitution N57D removes a glycosylation site.

[0133] In one useful embodiment, sdABD-B7H3 hF12(S59A) has an sdCDR1 of SEQ ID NO: 54, an sdCDR2 of SEQ ID NO: 55, and an sdCDR3 of SEQ ID NO: 56. Optionally, sdABD-B7H3 has SEQ ID NO: 53. In contrast to the hF7 and hF12 B7H3 sdABD, the amino acid substitution S59A removes a glycosylation site.

[0134] In one useful embodiment, sdABD-B7H3 hF12(S59Y) has an sdCDR1 of SEQ ID NO: 58, an sdCDR2 of SEQ ID NO: 59, and an sdCDR3 of SEQ ID NO: 60. Optionally, sdABD-B7H3 has SEQ ID NO: 57. In contrast to the hF7 and hF12 B7H3 sdABD, the amino acid substitution NS59Y removes a glycosylation site.

[0135] (d) FOLR1 sdABD As shown in Figure 5, there are a number of particularly useful sdABDs that bind to human FOLR1, referred to herein as "sdABD-FOLR1" or "FOLR1-ABD."

[0136] In one useful embodiment, sdABD-FOLR1 h77-2 has an sdCDR1 of SEQ ID NO: 22, an sdCDR2 of SEQ ID NO: 23, and an sdCDR3 of SEQ ID NO: 24. Optionally, sdABD-FOLR1 has SEQ ID NO: 21.

[0137] In one useful embodiment, sdABD-FOLR1 h59.3 has an sdCDR1 of SEQ ID NO: 26, an sdCDR2 of SEQ ID NO: 27, and an sdCDR3 of SEQ ID NO: 28. Optionally, sdABD-FOLR1 has SEQ ID NO: 25.

[0138] In one useful embodiment, sdABD-FOLR1 h22-4 has an sdCDR1 of SEQ ID NO: 30, an sdCDR2 of SEQ ID NO: 31, and an sdCDR3 of SEQ ID NO: 32. Optionally, sdABD-FOLR1 has SEQ ID NO: 29.

[0139] (e) Trop2 sdABD As shown in Figure 5, there are a number of particularly useful sdABDs that bind to human Trop2, referred to herein as "sdABD-Trop2" or "Trop2-ABDs."

[0140] In one useful embodiment, sdABD-Trop2 hVIB557 has an sdCDR1 of SEQ ID NO: 78, an sdCDR2 of SEQ ID NO: 79, and an sdCDR3 of SEQ ID NO: 80. Optionally, sdABD-Trop2 has SEQ ID NO: 77.

[0141] In one useful embodiment, sdABD-Trop2 hVIB565 has an sdCDR1 of SEQ ID NO: 82, an sdCDR2 of SEQ ID NO: 83, and an sdCDR3 of SEQ ID NO: 84. Optionally, sdABD-Trop2 has SEQ ID NO: 81.

[0142] In one useful embodiment, sdABD-Trop2 hVIB575 has an sdCDR1 of SEQ ID NO: 86, an sdCDR2 of SEQ ID NO: 87, and an sdCDR3 of SEQ ID NO: 88. Optionally, sdABD-Trop2 has SEQ ID NO: 85.

[0143] In one useful embodiment, sdABD-Trop2 hVIB578 has an sdCDR1 of SEQ ID NO: 90, an sdCDR2 of SEQ ID NO: 01, and an sdCDR3 of SEQ ID NO: 92. Optionally, sdABD-Trop2 has SEQ ID NO: 89.

[0144] In one useful embodiment, sdABD-Trop2 hVIB609 has an sdCDR1 of SEQ ID NO: 94, an sdCDR2 of SEQ ID NO: 95, and an sdCDR3 of SEQ ID NO: 96. Optionally, sdABD-Trop2 has SEQ ID NO: 93.

[0145] In one useful embodiment, sdABD-Trop2 hVIB619 has an sdCDR1 of SEQ ID NO: 98, an sdCDR2 of SEQ ID NO: 99, and an sdCDR3 of SEQ ID NO: 100. Optionally, sdABD-Trop2 has SEQ ID NO: 97.

[0146] (f) CA9 sdABD As shown in Figure 5, there are a number of particularly useful sdABDs that bind to human CA9, referred to herein as "sdABD-CA9" or "CA9-ABD."

[0147] In one useful embodiment, sdABD-CA9 hVIB456 has an sdCDR1 of SEQ ID NO: 102, an sdCDR2 of SEQ ID NO: 103, and an sdCDR3 of SEQ ID NO: 104. Optionally, sdABD-Trop2 has SEQ ID NO: 101.

[0148] In one useful embodiment, sdABD-CA9 hVIB476 has an sdCDR1 of SEQ ID NO: 106, an sdCDR2 of SEQ ID NO: 107, and an sdCDR3 of SEQ ID NO: 108. Optionally, sdABD-Trop2 has SEQ ID NO: 105.

[0149] In one useful embodiment, sdABD-CA9 hVIB407 has an sdCDR1 of SEQ ID NO: 110, an sdCDR2 of SEQ ID NO: 111, and an sdCDR3 of SEQ ID NO: 112. Optionally, sdABD-Trop2 has SEQ ID NO: 109.

[0150] In one useful embodiment, sdABD-CA9 hVIB445 has an sdCDR1 of SEQ ID NO: 114, an sdCDR2 of SEQ ID NO: 115, and an sdCDR3 of SEQ ID NO: 116. Optionally, sdABD-Trop2 has SEQ ID NO: 113.

[0151] In some embodiments, the protein prior to cleavage of the protease cleavage domain is less than about 100 kDa. In some embodiments, the protein following cleavage of the protease cleavage domain is about 25 to about 75 kDa. In some embodiments, the protein prior to protease cleavage has a size above the renal threshold for first-pass clearance. In some embodiments, the protein prior to protease cleavage has an elimination half-life of at least about 50 hours. In some embodiments, the protein prior to 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 against the same target antigen. In some embodiments, the protein has increased tissue distribution compared to IgG against the same target antigen.

[0152] C. Half-life extending domain The MCE proteins of the present invention (again, also referred to herein as "COBRA™" proteins or constructs) optionally comprise a half-life prolonging domain. Such domains are contemplated to include, but are not limited to, HSA binding domains, Fc domains, small molecules, and other half-life prolonging domains known in the art.

[0153] Human serum albumin (HSA) (molecular mass approximately 67 kDa) is the most abundant protein in plasma, present at approximately 50 mg / mL (600 μM), and has a half-life of approximately 20 days in humans. HSA helps maintain plasma pH, contributes to colloid pressure, functions as a carrier for many metabolites and fatty acids, and serves as the major drug transport protein in plasma.

[0154] Noncovalent association with albumin extends the elimination half-life of short-lived proteins. For example, recombinant fusion of an albumin-binding domain to a Fab fragment resulted in a 25- to 58-fold reduced in vivo clearance and a 26- to 37-fold increased half-life when administered intravenously to mice and rabbits, respectively, compared with administration of the Fab fragment alone. In another example, when insulin was acylated with fatty acids to promote association with albumin, a sustained effect was observed when injected subcutaneously into rabbits or pigs. Together, these studies demonstrated a link between albumin binding and sustained action.

[0155] In one aspect, the antigen-binding protein described herein comprises a half-life extending domain, e.g., a domain that specifically binds to HSA. In other embodiments, the HSA-binding domain is a peptide. In further embodiments, the HSA-binding domain is a small molecule. It is contemplated that the HSA-binding domain of the antigen-binding protein is fairly small, in some embodiments, 25 kD or less, 20 kD or less, 15 kD or less, or 10 kD or less. In certain cases, when the HSA-binding domain is a peptide or a small molecule, it is 5 kD or less.

[0156] In many embodiments, the half-life extending 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 "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 Figure 5.

[0157] A half-life prolonging domain of an antigen-binding protein provides altered pharmacodynamics and pharmacokinetics of the antigen-binding protein itself. As described above, a half-life prolonging domain extends the elimination half-life. A half-life prolonging domain also alters the pharmacological properties of the antigen-binding protein, including altering tissue distribution, penetration, and diffusion. In some embodiments, a half-life prolonging domain provides improved tissue (including tumor) targeting, tissue penetration, tissue distribution, diffusion within tissues, and enhanced efficacy compared to a protein without a half-life prolonging binding domain. In one embodiment, a therapy effectively and efficiently utilizes a reduced amount of the antigen-binding protein, resulting in reduced side effects, such as reduced cytotoxicity to non-tumor cells.

[0158] Further characteristics of a half-life extending domain, e.g., an HSA-binding domain, include the binding affinity of the HSA-binding domain to HSA. The affinity of the HSA-binding domain can be selected to target a specific terminal half-life of 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 a moderate binding affinity. In still other embodiments, the HSA-binding domain has a low or no binding affinity. Exemplary binding affinities include KD concentrations of 10 nM or less (high), 10 nM to 100 nM (medium), and greater than 100 nM (low). As noted above, binding affinity to HSA is determined by known methods, such as surface plasmon resonance (SPR).

[0159] D. Protease Cleavage Site The protein compositions of the present invention, particularly the prodrug constructs, comprise one or more protease cleavage sites, generally located within a cleavable linker, as outlined herein.

[0160] As described herein, the prodrug constructs of the present invention contain at least one protease cleavage site comprising an amino acid sequence cleaved by at least one protease. In some cases, the MCE proteins described herein contain 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 cleaved by at least one protease. As discussed more fully herein, when two or more protease cleavage sites are used in a prodrug construct, they 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 appreciated 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.

[0161] The amino acid sequence of the protease cleavage site depends on the protease being targeted. As is known in the art, there are many human proteases that are found in the body and that may be associated with disease states.

[0162] Proteases are known to be secreted by some diseased cells and tissues, such as tumor or cancer cells, creating a protease-rich or protease-enriched microenvironment. In some cases, a subject's blood is rich in proteases. In some cases, tumor-surrounding cells secrete proteases into the tumor microenvironment. Tumor-surrounding cells that secrete proteases include, but are not limited to, tumor stromal cells, myofibroblasts, blood cells, mast cells, B cells, NK cells, regulatory T cells, macrophages, cytotoxic T cells, dendritic cells, mesenchymal stem cells, polymorphonuclear cells, and other cells. In some cases, proteases, such as proteases that target amino acid sequences found in microbial peptides, are present in the subject's blood. This characteristic allows targeted therapeutics, such as antigen-binding proteins, to have additional specificity because T cells are not bound by the antigen-binding protein except in the protease-rich microenvironment of the target cell or tissue.

[0163] Proteases are proteins that cleave proteins, sometimes in a sequence-specific manner. Proteases include serine proteases, cysteine ​​proteases, aspartic acid proteases, threonine proteases, glutamic acid proteases, metalloproteases, asparagine peptide lyases, 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 proteases, trypsin-like proteases, elastase-like proteases, subtilisin-like proteases, actinidin, bromelain, calpain, and caspases (e.g., caspase- 3), Mir1-CP, papain, HIV-1 protease, HSV protease, CMV protease, chymosin, renin, pepsin, matriptase, legumain, plasmepsin, nepenthesin, metalloexopeptidase, metalloendopeptidase, matrix metalloproteinase (MMP), MMP1, MMP2, MMP3, MMP8, MMP9, MMP13, MMP11, MMP14, meprin, urokinase-type 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).

[0164] Some suitable proteases and protease cleavage sequences are shown in FIGS.

[0165] E. Linker As discussed herein, different domains of the present invention are generally linked together using amino acid linkers, which can confer functionality, including flexibility or inflexibility (e.g., steric hindrance), as well as the ability to be cleaved using in situ proteases. These linkers can be classified in a number of ways.

[0166] The present invention provides "domain linkers" that are used to link two or more domains (e.g., VH and VL, target tumor antigen binding domains (TTABDs, sometimes referred to herein as "αTTA" (for "anti-TTA")) to VH or VL, half-life extending domains to another component, etc.). Domain linkers can be, for example, non-cleavable (NCL), cleavable ("CL"), constrained and cleavable (CCL), and constrained and non-cleavable (CNCL).

[0167] 1. Non-cleavable linker In some embodiments, domain linkers are non-cleavable. Generally, they can be one of two types: non-cleavable and flexible, which allows the "upstream" and "downstream" components of the linker of the construct to self-assemble intramolecularly in a specific way, or non-cleavable and constrained, which prevents the two components separated by the linker from self-assembling intramolecularly. Note, however, that in the latter case, the two component domains separated by the non-cleavable constraining linker will not self-assemble intramolecularly, while other intramolecular components will self-assemble to form a pseudo-Fv domain.

[0168] (i) Non-cleavable but flexible linkers In this embodiment, a linker is generally used to connect the domains and maintain their functionality through a longer, flexible domain that is not cleaved by in situ proteases in patients. Examples of internal, non-cleavable linkers suitable for linking domains in the polypeptides of the 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 may be about 15 amino acids in length.

[0169] (ii) Non-cleavable and constrained linkers In some cases, the linker does not contain a cleavage site and is too short to allow the protein domains separated by the linker to self-assemble intramolecularly, making it a "constrained non-cleavable linker" or "CNCL." For example, in Pro186, the active VH and active VL are separated by eight amino acids (an "8-mer") that do not allow the VH and VL to self-assemble into an active antigen-binding domain. In some embodiments, the linker is still flexible, e.g., (GGGS)n (n=2). In other embodiments, more rigid linkers, such as those containing proline or bulky amino acids, may be used, although these are generally less preferred.

[0170] 2. Cleavable Linker All of the prodrug constructs herein include at least one cleavable linker. Thus, in one embodiment, the domain linker is a 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 the protease cleavage site. Optionally, depending on the length of the cleavage recognition site, there can be several more linking amino acids at either or both the N-terminus or C-terminus of the CL, for example, 1, 2, 3, 4, or 5 amino acids at either or both the N-terminus or C-terminus of the cleavage site. Thus, the cleavable linker can also be constrained (e.g., an 8-mer) or flexible.

[0171] 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.

[0172] II. Domains of the Invention The present invention provides many different formats for the prodrug polypeptides of the invention. The present invention provides constrained Fv domains and constrained pseudo-Fv domains. In addition, the present invention provides multivalent conditionally effective ("MCE") proteins that contain two Fv domains but are non-isomerizing constructs. As outlined herein, these can be non-isomerizing, cleavable, or non-isomerizing, non-cleavable formats, but all constructs contain at least one protease cleavage domain.

[0173] Importantly, both of these domains (Fv domain and pseudo-Fv domain) are referred to herein as "constrained," meaning that only one of them needs to be constrained, as discussed above and shown in Figures 36, 37, and 38, but generally the protein has better expression when both linkers are constrained.

[0174] Those skilled in the art will understand that with respect to formats 1, 2, and 4, there are four possibilities for the N- 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; and aVL-aVH and iVH-iVL. All four have been tested, and while all four are active, the first order, aVH-aVL and iVL-iVH, shows better expression than the other three. Thus, while the description herein will generally be presented in this aVH-aVL and iVL-iVH format, the entire disclosure herein encompasses other orders of these domains.

[0175] It should be noted that generally, the N- to C-terminal order of the full-length constructs of the present invention is based on the aVH-aVL and iVL-iVH orientations.

[0176] In addition, it is known in the art that immunogenicity in humans may result from the C-terminal sequence of certain ABDs. Therefore, in general, a histidine tag (either His6 or His10) can be used, especially when the C-terminus of the construct terminates with an sdABD (e.g., the sdABD-HSA domain of many of the constructs). Although many or most of the sequences herein were generated using a His6 C-terminal tag for purification reasons, these sequences can also be used to reduce immunogenicity in humans, as shown by Holland et al., DOI 10.1007 / s10875-013-9915-0 and WO2013 / 024059.

[0177] A. Constrained Fv Domain The present invention provides constrained Fv domains comprising active VH and active VL domains covalently linked using a constrained linker, which can be cleavable (Format 1) or non-cleavable (Formats 2 and 4), as outlined herein. 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 domains, where vhCDR1, vhCDR2, and vhCDR3 of the VH bind human CD-3, and vlCDR1, vCDR2, and vlCDR3 of the VL bind human CD-3; however, in a prodrug format (e.g., uncleaved), the VH and VL are unable to sterically associate to form an active binding domain, preferring instead to pair intramolecularly with the pseudo-Fv.

[0178] The constrained Fv domains can 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, as described herein.

[0179] As will be appreciated by one of skill in the art, the order of VH and VL in a constrained Fv domain can be either VH-linker-VL or VL-linker-VH (from N-terminus to C-terminus).

[0180] As outlined herein, for Format 1 constructs, the constrained Fv domain can comprise a VH and VL linked using a cleavable linker, 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. Generally, a constrained Fv domain contains active VH and VL domains (e.g., capable of binding to 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.

[0181] As outlined herein, for Format 2 constructs, the constrained Fv domain can comprise a VH and VL linked using a non-cleavable linker. In this embodiment, the constrained Fv domain has the structure (N- to C-terminus): vhFR1-vhCDR1-vhFR2-vhCDR2-vhFR3-vhCDR3-vhFR4-CNCL-vlFR1-vlCDR1-vlFR2-vlCDR2-vlFR3-vlCDR3-vlFR4. Generally, a constrained Fv domain contains active VH and VL domains (e.g., capable of binding to 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.

[0182] Of particular use in the present invention is a constrained non-cleavable Fv domain having an aVH with SEQ ID NO: 142, an aVL with SEQ ID NO: 126 and a domain linker with SEQ ID NO: 233.

[0183] B. Constrained pseudo-Fv domain The present invention provides constrained pseudo-Fv domains comprising inactive or pseudo-iVH and iVL domains covalently linked using a constrained linker (which can be cleavable or non-cleavable, as outlined herein). The constrained linker prevents intramolecular association between the iVH and iVL in the absence of cleavage. Thus, constrained pseudo-Fv domains generally comprise iVH and iVL domains with framework regions that allow the iVH and iVL to associate (when in their unconstrained forms), but the resulting pseudo-Fv domains do not bind to human proteins. The iVH domain can assemble with the aVL domain, and the iVL domain can assemble with the aVH domain, but the resulting structure does not bind to CD3.

[0184] The constrained pseudo-Fv domains comprise an inactive VH and VL (iVH and iVL).

[0185] As will be appreciated by one of skill in the art, the order of VH and VL in the constrained pseudo-Fv domain can be either VH-linker-VL or VL-linker-VH (from N-terminus to C-terminus).

[0186] As outlined herein, constrained pseudo-Fv domains can comprise iVH and iVL linked using a non-cleavable linker as shown in Formats 1, 2, and 4, or with a cleavable linker as shown in Format 3.

[0187] Generally, a constrained Fv domain contains inactive VH and VL domains (e.g., capable of binding to CD3 when associated) and therefore has the structure (N-terminus to C-terminus): vhFR1-ivlCDR1-vhFR2-ivlCDR2-vhFR3-ivlCDR3-vhFR4-CNCL-vlFR1-ivhCDR1-vlFR2-ivhCDR2-vlFR3-ivhCDR3-vlFR4.

[0188] Constrained non-cleavable pseudo-Fv domains having an iVH with SEQ ID NO: 146, SEQ ID NO: 150, and SEQ ID NO: 154, an iVL with SEQ ID NO: 130, SEQ ID NO: 134, or SEQ ID NO: 138, and a domain linker with SEQ ID NO: 233 are of particular use in the present invention.

[0189] III. Form of the Invention As discussed herein, the prodrug constructs of the present invention can take many different forms, including cleavable forms with dual TTA binding domains, non-cleavable forms with dual TTA binding domains (either of which can have the same TTA binding domain or different binding domains), and non-cleavable forms with a single targeting domain.

[0190] A. Cleavable Format with Dual Targeting The present invention provides a non-isomerizing cleavable format of the "Format 1" type of Figure 1. In this embodiment, the constrained Fv domain comprises VH and VL domains linked using a constrained cleavable linker, and the constrained pseudo-Fv domain uses a constrained non-cleavable linker. For ease of discussion, both of these are referred to herein as "constrained," although as discussed above and shown in Figures 37, 38, and 39, only one of these need be constrained, although generally, proteins have better expression when both linkers are constrained.

[0191] All constructs of Format 1 (as well as other formats) also have a cleavable linker (CL) that is cleaved by human tumor proteases.

[0192] 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.

[0193] As will be appreciated by one of skill in the art, the order of VH and VL of either the constrained Fv domain or constrained pseudo-Fv domain can be (from N-terminus to C-terminus) either VH-linker-VL or VL-linker-VH.

[0194] 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.

[0195] 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.

[0196] 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.

[0197] 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.

[0198] 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), where aVH, aVL, iVH, and iVL have the sequences shown in Figure 5.

[0199] 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, the aVH, aVL, iVH, iVL have the sequence shown in Figure 5. In this embodiment, the two targeting domains bind to the same TTA, which may be EGFR, EpCAM, FOLR1, Trop2, CA9, or B7H3, the sequence of which is shown in Figure 5.

[0200] 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 different TTAs.

[0201] 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, the aVH, aVL, iVH, and 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 of Figure 5.

[0202] 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, the aVH, aVL, iVH, and 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 of Figure 5.

[0203] 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, the aVH, aVL, iVH, and 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 of Figure 5.

[0204] 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 of Figure 5.

[0205] 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, the aVH, aVL, iVH, and 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 of Figure 5.

[0206] 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, the aVH, aVL, iVH, and 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 of Figure 5.

[0207] 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, the aVH, aVL, iVH, and iVL have the sequence shown in Figure 5. In this embodiment, the two targeting domains bind to the same TTA, which may be EGFR, FOLR1, B7H3, Trop2, CA9, or EpCAM, the sequence of which is shown in Figure 5, where CCL and CL are selected from linkers that are cleaved by MMP9 or meprin, and sdABD(1 / 2) has SEQ ID NO: 117 or SEQ ID NO: 121.

[0208] In Format 1, a preferred domain linker is SEQ ID NO: 233 (which also serves as a preferred constrained non-cleavable linker).

[0209] In format 1, the preferred constructs are Pro140 and Pro140b.

[0210] B. Non-cleavable form As shown in Figure 2, the present invention provides a non-isomerizing, non-cleavable format. In this embodiment, it is understood that "non-cleavable" applies only to the linkage of the constrained Fv domains, due to the presence of an activated cleavage site in the prodrug construct. In this embodiment, the constrained Fv domain comprises a VH and VL domain linked using a constrained non-cleavable linker, and the constrained pseudo-Fv domain uses a constrained non-cleavable linker.

[0211] As will be appreciated by one of skill in the art, the order of VH and VL of either the constrained Fv domain or constrained pseudo-Fv domain can be (from N-terminus to C-terminus) either VH-linker-VL or VL-linker-VH.

[0212] 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.

[0213] As will be appreciated by one of skill in the art, the order of VH and VL of either the constrained Fv domain or constrained pseudo-Fv domain can be (from N-terminus to C-terminus) either VH-linker-VL or VL-linker-VH.

[0214] 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.

[0215] 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.

[0216] 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.

[0217] 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.

[0218] In some embodiments, the prodrug protein comprises, from N- 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, the aVH, aVL, iVH, iVL have the sequence shown in Figure 5. In this embodiment, the two targeting domains bind to the same TTA, which may be EGFR, EpCAM, FOLR1, Trop2, CA9, or B7H3, the sequence of which is shown in Figure 5.

[0219] In some embodiments, the prodrug protein comprises, from N- 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.

[0220] 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 of Figure 5. In this embodiment, preferred combinations of EGFR and EpCAM include: [Table 2]

[0221] In this case, "either orientation" means that the EpCAM sdABD is either N-terminal to the EGFR sdABD or C-terminal to the EGFR sdABD in the constructs of the invention.

[0222] 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 of Figure 5. In this embodiment, preferred combinations of EGFR and FOLR1 include the following: [Table 3]

[0223] In this case, "either orientation" means that the FOLR1 sdABD is either N-terminal to the EGFR sdABD or C-terminal to the EGFR sdABD in the constructs of the invention.

[0224] 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, the 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 of Figure 5. In this embodiment, preferred combinations of EGFR and B7H3 include: [Table 4]

[0225] In this case, "either orientation" means that the B7H3 sdABD is either N-terminal to the EGFR sdABD or C-terminal to the EGFR sdABD in the constructs of the invention.

[0226] 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 of Figure 5. In this embodiment, a preferred combination of EpCAM and FOLR1 includes the following: [Table 5]

[0227] In this case, "either orientation" means that the EpCAM sdABD is either N-terminal to the FOLR1 sdABD or C-terminal to the FOLR1 sdABD in the constructs of the invention.

[0228] 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 of Figure 5. In this embodiment, preferred combinations of EpCAM and B7H3 include: [Table 6]

[0229] In this case, "either orientation" means that the B7H3 sdABD is either N-terminal to the EGFR sdABD or C-terminal to the EGFR sdABD in the constructs of the invention.

[0230] 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 of Figure 5. In this embodiment, preferred combinations of FOLR1 and B7H3 include the following: [Table 7]

[0231] In this case, "either orientation" means that the B7H3 sdABD is either N-terminal to the FOLR1 sdABD or C-terminal to the FOLR1 sdABD in the constructs of the invention.

[0232] 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, the aVH, aVL, iVH, and iVL have the sequence shown in Figure 5. In this embodiment, the two targeting domains bind to the same TTA, which may be EGFR, FOLR1, B7H3, CA9, Trop2, or EpCAM, the sequence of which is shown in Figure 5, where CCL and CL are selected from linkers that are cleaved by MMP9 or meprin, and sdABD(1 / 2) has SEQ ID NO: 117.

[0233] In Format 2, a preferred domain linker is SEQ ID NO: 233 (which also serves as a preferred constrained non-cleavable linker).

[0234] In Format 2, preferred dual targeting constructs (sometimes referred to herein as "hetero-COBRAs") include combinations targeting EGFR and EpCAM, EGFR and Trop2, EGFR and FOLR1, EGFR and B7H3, EpCAM and Trop2, EpCAM and FOLR1, EpCAM and B7H3, Trop2 and FOLR1, Trop2 and B7H3, and FOLR1 and B7H3, as described more fully below.

[0235] In format 2, specific use embodiments include 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, Pro667, Pro Protease inhibitors include, but are not limited to, 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.

[0236] C. Single TTA Constructs As shown in Figure 4, "Format 4" constructs, which are similar to the Format 2 constructs but lack the second TTA ABD, are also included in the compositions of the invention. In this embodiment, it is understood that "non-cleavable" applies only to the linkage of the constrained Fv domains due to the presence of an activation cleavage site in the prodrug construct. In this embodiment, the constrained Fv domain comprises VH and VL domains linked using a constrained non-cleavable linker, and the constrained pseudo-Fv domain uses a constrained non-cleavable linker.

[0237] As will be appreciated by one of skill in the art, the order of VH and VL of either the constrained Fv domain or constrained pseudo-Fv domain can be (from N-terminus to C-terminus) either VH-linker-VL or VL-linker-VH.

[0238] The present invention provides prodrug proteins comprising, from N- to C-terminus, sdABD(TTA)-domain linker-constrained Fv domain-cleavable linker-sdABD-HSA-constrained pseudo-Fv domain (note that for all constructs of this type, sdABD-HSA generally does not have His6, but it can be included).

[0239] As will be appreciated by one of skill in the art, the order of VH and VL of either the constrained Fv domain or constrained pseudo-Fv domain can be (from N-terminus to C-terminus) either VH-linker-VL or VL-linker-VH.

[0240] 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.

[0241] 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.

[0242] 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.

[0243] 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.

[0244] Thus, in one embodiment, the prodrug protein comprises, from N- to C-terminus, (sdABD-TTA)-domain linker-aVH-CNCL-aVL-CL-(sdABD-HSA)-domain linker-iVL-CNCL-iVH. In this embodiment, the aVH, aVL, iVH, and iVL have the sequence shown in Figure 5. In this embodiment, the targeting domain binds to a TTA, which can be EGFR, EpCAM, FOLR1, Trop2, CA9, or B7H3, the sequence of which is shown in Figure 5.

[0245] In Format 4, a preferred domain linker is SEQ ID NO: 233 (which also serves as a preferred constrained non-cleavable linker).

[0246] In Format 4, preferred sdABD-HSAs are those of SEQ ID NOs: 121 or 117.

[0247] D. Two Protein Compositions In some embodiments, the compositions of the invention comprise two distinct molecules, sometimes referred to as "hemi-COBRA™" or "hemi-constructs," that associate intramolecularly to form a pseudo-Fv in the absence of cleavage. In the presence of a protease, the cleavage site is cleaved, releasing the non-active variable domain, and the protein pair then forms an active antigen-binding domain against CD3, as generally shown in Figure 3.

[0248] The key to the hemi-construct design is that the active variable domain and sdABD-TTA remain together after cleavage, allowing the two cleaved portions to be held together by tumor antigen receptors on the tumor surface and then form the active anti-CD3 binding domain.

[0249] There are two different general format 3 constructs: one with a single sdABD-TTA in each member of the pair (Figure 3A), and one with two different sdABD-TTAs, each for a different TTA (Figure 3B).

[0250] 1. Hemi-COBRA™ Construct with a Single TTA Binding Domain (Format 3A) In some embodiments, a first hemi-COBRA™ has, from N-terminus to C-terminus, sdABD(TTA1)-domain linker-aVH-CL-iVL-domain linker-sdABD(1 / 2), and a second hemi-COBRA™ has sdABD(1 / 2)-domain linker-iVH-CL-aVL-domain linker-sdABD(TTA2). In this embodiment, the 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 shown in Figure 5.

[0251] 2. Hemi-COBRA™ Construct with Double TTA ABD In some embodiments, paired prodrug constructs can have two sdABD-TTA binding domains per construct, as shown in Figure 3B. In this embodiment, the first member of the pair comprises, from N- 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- to C-terminus, sdABD-TTA1-domain linker-sdABD-TTA2-aVL-CL-iVH-domain linker-sdABD-HSA.

[0252] The two sdABD-TTAs on each member of the pair are different, but generally both members (hemi-COBRA™) have the same two sdABD-TTAs, such as both having EGFR and FOLR1 or EGFR and B7H3.

[0253] The two sdABD-TTAs are, in some embodiments, selected from those shown in FIG.

[0254] IV. Methods of Making the Compositions of the Invention The prodrug compositions of the present invention are generally made as understood by those of skill in the art and as outlined below.

[0255] The present invention provides nucleic acid compositions encoding the prodrug compositions of the present invention. As will be understood by those skilled in the art, the nucleic acid composition will depend on the format of the prodrug polypeptide(s). Thus, for example, if a format requires two amino acid sequences, such as a "Format 3" construct, the two nucleic acid sequences can be incorporated into one or more expression vectors for expression. Similarly, single-polypeptide prodrug constructs (Formats 1, 2, and 4) require a single nucleic acid in a single expression vector for production.

[0256] As is known in the art, nucleic acids encoding the components of the present invention can be incorporated into expression vectors, as is known in the art and depending on the host cell used to produce the prodrug compositions of the present invention. Generally, the nucleic acid is operably linked to any number of regulatory elements (promoter, origin of replication, selectable marker, ribosome binding site, inducer, etc.). Expression vectors can be extrachromosomal or integrating vectors.

[0257] The nucleic acids and / or expression vectors of the invention are then transformed into any number of different types of host cells as are well known in the art, including mammalian, bacterial, yeast, insect, and / or fungal cells, with mammalian cells (e.g., CHO cells, 293 cells) being used in many embodiments.

[0258] The prodrug compositions of the present invention are produced by culturing host cells containing the expression vector(s), as is well known in the art. Once produced, conventional antibody purification steps, including protein A affinity and / or ion exchange chromatography steps, are performed.

[0259] V. Formulation and Administration of Prodrug Compositions of the Invention 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) of the desired purity with any pharmaceutically acceptable carriers, excipients, or stabilizers (generally as reviewed in Remington's Pharmaceutical Sciences 16th edition, Osol, A. Ed.

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

[0260] 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.

[0261] The prodrug compositions of the present invention are useful in the treatment of cancer. [Example]

[0262] A. Example 1: Construction and Purification of Pro-Constructs Transfection Each protein (e.g., single protein for formats 1, 2, and 4) or pair of constructs (format 3) was expressed from a separate expression vector (pcdna3.4 derivative). Equal amounts of plasmid DNA encoding the hemi-cobra or single-chain construct pair were mixed and transfected into Expi293 cells according to the manufacturer's transfection protocol. Conditioned medium was harvested 5 days posttransfection by centrifugation (6000 rpm x 25') and filtration (0.2 μM filter). Protein expression was confirmed by SDS-PAGE. Once the constructs were purified, the final buffer composition was 25 mM citrate, 75 mM arginine, 75 mM NaCl, 4% sucrose, pH 7. The final preparation was stored at -80°C.

[0263] MMP9 activation Recombinant human (rh) MMP-9 was activated according to the following protocol. Recombinant human MMP-9 (R&D #911-MP-010) was at 0.44 mg / mL (4.7 μM). p-Aminophenylmercuric acetate (APMA) (Sigma) is prepared at a stock concentration of 100 mM in DMSO. The assay buffer is 50 mM Tris, pH 7.5, 10 mM CaCl, 150 mM NaCl, 0.05% Brij-35.

[0264] -Dilute rhMMP9 to approximately 100ug / mL in assay buffer (25uL hMMP9 + 75uL assay buffer)

[0265] -Add p-aminophenylmercuric acetate (APMA) from a 100 mM stock in DMSO to a final concentration of 1 mM (1 uL to 100 uL)

[0266] Incubate at -37°C for 24 hours

[0267] -Dilute MMP9 to 10ng / uL (add 900uL of assay buffer to 100uL of activation solution)

[0268] The concentration of activated rhMMP9 is approximately 100 nM.

[0269] Cleavage of constructs for TDCC assay

[0270] To cleave the construct, 100 μL of 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) was supplemented with activated rhMMP9 supplemented with up to 10 mM CaCl2 to a concentration of 20–35 nM. The sample was incubated overnight (16–20 h) at room temperature. The completeness of cleavage was verified using SDS-PAGE (10–20% TG, TG running buffer, 200 v, 1 h). Samples were typically 98% cleaved.

[0271] B. Example 2: T-cell dependent cytotoxicity (TDCC) assay Firefly luciferase-transduced HT-29 cells were grown to approximately 80% confluency and detached with Versene (0.48 mM EDTA in PBS-Ca-Mg). Cells were centrifuged and resuspended in TDCC medium (5% heat-inactivated FBS in RPMI 1640 with HEPES, GlutaMax, sodium pyruvate, non-essential amino acids, and β-mercaptoethanol). Purified human Pan-T cells were thawed, centrifuged, and resuspended in TDCC medium.

[0272] Co-cultures of HT-29_Luc cells and T cells were added to 384-well cell culture plates. Serially diluted COBRA was 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. The plate was read on a Perkin Elmer Envision with an exposure time of 0.1 seconds per well. Total luminescence was recorded, and the data was analyzed in GraphPad Prism 7 or Version 8.3.1 (depending on timing).

[0273] C. Example 3: General Protocol Design of In Vivo Adoptive T Cell Transfer Efficacy Model These protocols were used for many of the experiments shown. Tumor cells were implanted subcutaneously (SC) into the right flank of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, Cat. No. 005557) and grew to approximately 200 mm 3 The tumors were grown until established tumors with an average volume of 1000 μg / ml were achieved. In parallel, human T cells were cultured for approximately 10 days in T cell medium (X-VIVO 15 [Lonza, catalog number 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M gas-permeable flasks (Wilson Wolf, catalog number 81100S) with MACSiBeads from a T Cell Activation / Expansion Kit (Miltenyi, catalog number 130-091-441), supplemented with recombinant human IL-2 protein. To control tumor growth and human T cell activation / expansion in mice, mice were randomly divided into groups (N=6) based on tumor size on day 0 of the study, and then each group received 2.5 × 10 cells. 6 Mice were given an initial dose of COBRA or a control molecule by intravenous (IV) injection of cultured human T cells. Mice were given 7 doses every 3 days (days 0, 3, 6, 9, 12, 15, and 18), and then tumors were grown to 2000 mm 3 Treatment continued for an additional 2-3 weeks until tumor volumes exceeded 1000 mg / kg or the study was terminated. Tumor volumes were measured every 3 days.

[0274] D. Example 4: In vivo activity with EGFR / MMP9 hemi-COBRA versus Pro77 and Pro53. 5×10 6 or 5 x 10 LoVo cells 6HT29 cells were implanted subcutaneously into the right flank of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog no. 005557) and allowed to grow until tumors were established. In parallel, human T cells were cultured for 10 days in T cell medium (X-VIVO 15 [Lonza, catalog no. 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M gas-permeable flasks (Wilson Wolf, catalog no. 81100S) with MACSiBeads from the T Cell Activation / Expansion Kit (Miltenyi, catalog no. 130-091-441) and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion in mice was controlled, whereby on day 0 of the study, mice were randomly divided into groups (N=6) based on tumor size, and then each group received 2.5 x 10 6 Mice were given an initial dose of COBRA or a control molecule by intravenous (IV) injection of cultured human T cells. Mice were given 7 doses every 3 days (days 0, 3, 6, 9, 12, 15, and 18), and then tumors were grown to 2000 mm 3 Tumor volumes were administered until tumors reached a volume exceeding 100 μg / mL or the study was terminated. Groups received 0.2 mg / kg (mpk) of anti-EGFR x CD3 positive control, Pro51 bispecific antibody (bsAb), 0.5 mpk of negative control, anti-hen egg white lysozyme (HEL) x CD3 bsAb Pro98, 0.5 mpk of each of anti-EGFR hemi-COBRA with an MMP9-cleavable linker containing Pro77 and Pro53, or 0.5 mpk of each of anti-EGFR hemi-COBRA with a non-cleavable (NCL) linker containing Pro74 and Pro72. Tumor volumes were measured every 3 days.

[0275] E. Example 5: EGFR / MMP9 In Vivo Activity with COBRA Pro140. 5×10 6 or 5 x 10 LoVo cells 6HT29 cells were implanted subcutaneously into the right flank of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog no. 005557) and allowed to grow until tumors were established. In parallel, human T cells were cultured for 10 days in T cell medium (X-VIVO 15 [Lonza, catalog no. 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M gas-permeable flasks (Wilson Wolf, catalog no. 81100S) with MACSiBeads from the T Cell Activation / Expansion Kit (Miltenyi, catalog no. 130-091-441) and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion in mice was controlled, whereby on day 0 of the study, mice were randomly divided into groups (N=6) based on tumor size, and then each group received 2.5 x 10 6 Mice were given an initial dose of COBRA or a control molecule by intravenous (IV) injection of cultured human T cells. Mice were given 7 doses every 3 days (days 0, 3, 6, 9, 12, 15, and 18), and then tumors were grown to 2000 mm 3 Tumors were administered until a volume exceeding 100 μg / ml was reached or the study was terminated. Groups received 0.2 mpk of anti-EGFR x CD3 positive control, Pro51 bispecific antibody (bsAb), 0.5 mpk of negative control, anti-hen egg white lysozyme (HEL) x CD3 bsAb Pro98, or 0.5 mpk of anti-EGFR COBRA Pro140 containing an MMP9-cleavable linker. Tumor volumes were measured every 3 days.

[0276] F. Example 6: EGFR / MMP9 In Vivo Activity with COBRA Pro186. 5×10 6HT29 cells were subcutaneously implanted into the right flank of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl / SzJ) mice (The Jackson Laboratory, catalog no. 005557) and allowed to grow until tumors were established. In parallel, human T cells were cultured for 10 days in T cell medium (X-VIVO 15 [Lonza, catalog no. 04-418Q], 5% human serum, 1% penicillin / streptomycin, 0.01 mM 2-mercaptoethanol) in G-Rex100M gas-permeable flasks (Wilson Wolf, catalog no. 81100S) with MACSiBeads from the T Cell Activation / Expansion Kit (Miltenyi, catalog no. 130-091-441) and supplemented with recombinant human IL-2 protein. Tumor growth and human T cell activation / expansion in mice was controlled, whereby on day 0 of the study, mice were randomly divided into groups (N=6) based on tumor size, and then each group received 2.5 x 10 6 Mice were given an initial dose of COBRA or a control molecule by intravenous (IV) injection of cultured human T cells. Mice were given 7 doses every 3 days (days 0, 3, 6, 9, 12, 15, and 18), and then tumors were grown to 2000 mm 3 Tumor volumes were measured every 3 days. Tumor volumes were measured every 3 days. Groups received 0.1 mg / kg (mpk) of anti-EGFRxCD3 positive control, Pro51 bispecific antibody (bsAb), 0.3 mpk of non-cleavable (NCL) control linker containing anti-EGFR COBRA Pro214, 0.1 or 0.3 mpk of MMP9-cleavable linker containing anti-EGFR COBRA Pro140, or 0.1 or 0.3 mpk of MMP9-cleavable linker containing anti-EGFR COBRA Pro186.

[0277] G. Example 7: Successful humanization of anti-EGFR sequences The results are shown below. [Table 8]

[0278] These results demonstrate both that the EGFR binding domain was successfully humanized and that there is strong binding to the target EGFR when two binding sites are present on the molecule.

[0279] Example: Successful humanization of EpCAM sdABD The results are shown below. [Table 9]

[0280] These results demonstrate both the successful humanization of the EpCAM-binding domain.

[0281] H. Example 8: COBRA™: A Novel Conditionally Active Bispecific Antibody That Regresses Established Solid Tumors in Mice Despite clinical success with bispecific antibodies (bsAbs) targeting hematologic malignancies (e.g., blinatumomab, a CD19 × CD3 bsAb), efficacy in solid tumor indications remains a significant challenge. Because T cell-redirecting bsAbs are so potent, even very low levels of cell surface target antigen expression on normal tissues can quickly become a safety disadvantage and severely limit the dose levels that can be achieved in patients. This limits the ability to reach effective concentrations and reduces the therapeutic potential of these highly active molecules. Additionally, identifying "clean" target antigens that are uniquely expressed on tumors but not on normal tissues has been extremely challenging at best.

[0282] To overcome these challenges, the inventors developed a novel recombinant bsAb platform called COBRA™ (Conditional Bispecific Redirected Activation). COBRA is engineered to enable targeting of more widely expressed and validated tumor cell surface antigens by focusing T cell engagement in the tumor microenvironment. The COBRA molecule is designed to bind to the target antigen, which may be expressed on both tumor and normal cells, but does not engage T cells unless exposed to the proteolytic microenvironment that is common in tumors but not in normal healthy tissue. Upon binding to the tumor target antigen, protease-dependent linker cleavage enables COBRA to convert the inactive anti-CD3 scFv into an active anti-CD3 scFv binding domain. Subsequently, upon conversion, COBRA can simultaneously co-engage T cells with the target antigen, resulting in a potent cytolytic T cell response against tumor cells. Additionally, COBRA is designed with a half-life extending moiety that is removed from the active molecule upon proteolytic cleavage, allowing for the sustained presence of inactive COBRA in the circulation prior to binding to the tumor target, and allowing for more rapid clearance of unbound active COBRA molecules, thereby reducing the potential for cytotoxic activity in normal tissues.

[0283] Here, we demonstrate the novel design of the COBRA molecule and demonstrate its ability to engage CD3 and epidermal growth factor receptor (EGFR) and induce potent cytotoxic activity in T cell cultures and tumor-bearing mice implanted with human T cells. We report low- to 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.

[0284] Figures 64A-64C show the design and predicted folding mechanism of COBRA. Figure 64A shows a schematic of PRO186 COBRA. Figure 64B shows the predicted COBRA fold. COBRA contains an inactive VH and VL domain paired with an anti-CD3 VH and VL domain. Uncleaved PRO186 COBRA binds to EGFR, impairs CD3 binding, and binds serum albumin. Figure 64C shows an analytical size exclusion chromatogram of PRO186. The data indicate that uncleaved PRO186 folds into a single structure.

[0285] Figures 65A-65C show exemplary embodiments of constructs described herein that include PRO186 (pre-cleaved Pro186), PRO186 cleavage products, and PRO186 active dimers. One cleavage product contains anti-CD3 VH and VL domains, binds to EGFR, and impairs CD3 binding. The other cleavage product contains anti-CD3 inactive VH and VL domains and binds serum albumin. The active PRO186 dimer contains an active anti-CD3 agonist (anti-CD3 VH and VL dimer) and binds to CD3 and EGFR.

[0286] Figure 66 provides an illustration of COBRA conversion to an active dimer upon protease cleavage.

[0287] Figures 67A-67B provide characterization of COBRA binding. Figure 67A shows binding activity to human, cynomolgus monkey, and mouse antigens. Figure 67B shows PRO186 binding to human CD3 epsilon, active PRO186 binding of human CD3 epsilon, and active PRO186 binding of human EGFR. Binding kinetics were assessed by Octet (Forte Bi) using EGFR (Acro Biosystems), serum albumin (Athens Research Technology), and CD3ε (Creative Biomart).

[0288] Figures 68A-68B show the cleavage of the PRO186 linker by MMP2 and MMP9. Figure 68A shows a Western blot of active binding product molecules upon cleavage. Figure 68B shows the accumulation of active binding product molecules versus cleavage time.

[0289] Figure 69 shows the in vitro activity of conditional PRO186 constructs. Figure 69 - left panel shows the results of a T cell killing assay. Figure 69 - right panel shows the level of IFN-gamma release in relation to the concentration of the test article.

[0290] Figure 70 shows EGFR expression relative to activity in three tumor cell lines: LoVo (a colorectal cancer (CRC) cell line), HT-29 (a colorectal cancer (CRC) cell line), and SCC25 (a head and neck cancer cell line). For in vitro EGFR expression, antibody binding / cell was measured using 1:1 PE-labeled anti-EGFR mAb #EGFR.1 and BD QuantiBrite Beads. 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 10:1 E:T for 48 hours and measured by Steady-Glo (Promega). For IFNγ release assays, IFNγ was measured at 24 hours using a Meso Scale Discovery V-Pex at 10:1 E:T.

[0291] Figures 71A and 71B show the expression of EGFR, MMP2, and MMP9 on tumor cells and tumor xenografts. Figure 71A shows the EGFR cell surface density on three cancer cell lines: LoVo, HT-29, and SCC25. Figure 71B shows immunohistochemical staining of EGFR, MMP2, and MMP9 on tumor xenografts.

[0292] Figure 72 provides a schematic diagram of the experimental procedure for the adoptive human T cell transfer model in tumor-bearing mice. The experiment was used to measure in vivo antitumor efficacy and pharmacokinetics (PK). The procedure included: (1) subcutaneously implanting tumors into the right flank of NSG mice; (2) allowing established tumors to grow, such as tumors of approximately 200 mm; (3) dosing mice q3dx7 starting on day 0; (4) administering the final dose on day 18; and (5) terminating the study. The procedure also included: (a) activating and expanding human T cells in culture for 10 days, so that expansion began at the same time as tumor implantation; and (b) harvesting T cells on day 0, which were performed in parallel with the in vivo experiments.

[0293] Figure 73 shows regression of established solid tumors in mice by PRO186. Figure 73 - left panel shows regression of LoVo-derived tumors. Figure 73 - center panel shows regression of HT-29-derived tumors. Figure 73 - right panel shows regression of SCC25-derived tumors.

[0294] Figures 74A-B show that cleaved PRO186 is cleared more rapidly than intact (uncleaved) PRO186. Figure 74A shows the pharmacokinetics of the test article in the plasma of non-tumor-bearing mice. Figure 74B shows the tumor volume of LoVo-derived tumors in mice administered the test article.

[0295] Conclusions: The inventors designed a multivalent sdAb-diabody fusion that, upon proteolytic action, converts into a highly potent bispecific redirected T cell therapeutic. In vitro assays demonstrated that protease-dependent linker cleavage increased the potency of T cell-mediated killing by 200-fold, resulting in a therapeutic with sub-picomolar potency. Administration of PRO186 (Pro186) in mice bearing established xenografts resulted in protease cleavage-dependent T cell-mediated tumor regression in multiple tumor models. PRO186 exhibited (1) an extended in vivo half-life upon administration and (2) rapid clearance after proteolytic activation, thereby demonstrating that PRO186 is a therapeutic with an improved safety profile over conventional T cell-redirecting bispecifics.

Claims

1. A fusion protein comprising, from the N-terminus to the C-terminus: a) a first sdABD-TTA; and b) a first domain linker; and c) a constrained Fv domain, 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 a vlCDR1, a vlCDR2, and a vlCDR3; and d) a second domain linker; and e) a second sdABD-TTA; and f) a cleavable linker (CL); and g) a constrained pseudo-Fv domain, i) a first pseudo-light variable domain; ii) a non-cleavable linker (NCL), and iii) a first pseudo-heavy variable domain; and h) a third domain linker; and i) a third sdABD that binds to human serum albumin; 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; the first variable light domain and the first pseudo-variable heavy domain associate intramolecularly to form an inactive Fv; The fusion protein, wherein at least one of the sdABD-TTAs is sdABD-B7H3 having a sequence selected from SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53, and SEQ ID NO:

57.

2. 2. The fusion protein of claim 1, wherein the fusion protein is Pro664 and has sequence number 282.

3. A fusion protein comprising, from the N-terminus to the C-terminus: a) a first single domain antigen binding domain (sdABD) (sdABD-TTA) that binds to a human tumor target antigen (TTA); b) a first domain linker; and c) a constrained Fv domain, 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 a vlCDR1, a vlCDR2, and a vlCDR3; and d) a second domain linker; and e) a second sdABD-TTA; and f) a cleavable linker (CL); and g) a constrained pseudo-Fv domain, i) a first pseudo-light variable domain; ii) a non-cleavable linker (NCL), and iii) a first pseudo-heavy variable domain; and h) a third domain linker; and i) a third sdABD that binds to human serum albumin; 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; the first variable light domain and the first pseudo-variable heavy domain associate intramolecularly to form an inactive Fv; The fusion protein, wherein at least one of the sdABD-TTAs is sdABD-EpCAM having a sequence selected from SEQ ID NO:69 and SEQ ID NO:

73.

4. A fusion protein comprising, from the N-terminus to the C-terminus: a) a first sdABD-TTA; and b) a first domain linker; and c) a constrained Fv domain, 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 a vlCDR1, a vlCDR2, and a vlCDR3; and d) a second domain linker; and e) a second sdABD-TTA; and f) a cleavable linker (CL); and g) a constrained pseudo-Fv domain, i) a first pseudo-light variable domain; ii) a non-cleavable linker (NCL), and iii) a first pseudo-heavy variable domain; and h) a third domain linker; and i) a third sdABD that binds to human serum albumin; 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; the first variable light domain and the first pseudo-variable heavy domain associate intramolecularly to form an inactive Fv; The fusion protein, wherein at least one of the sdABD-TTAs is sdABD-Trop2 having a sequence selected from 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 the N-terminus to the C-terminus: a) a first sdABD-TTA; and b) a first domain linker; and c) a constrained Fv domain, 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 a vlCDR1, a vlCDR2, and a vlCDR3; and d) a second domain linker; and e) a second sdABD-TTA; and f) a cleavable linker (CL); and g) a constrained pseudo-Fv domain, i) a first pseudo-light variable domain; ii) a non-cleavable linker (NCL), and iii) a first pseudo-heavy variable domain; and h) a third domain linker; and i) a third sdABD that binds to human serum albumin; 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; the first variable light domain and the first pseudo-variable heavy domain associate intramolecularly to form an inactive Fv; The fusion protein, wherein at least one of the sdABD-TTAs is sdABD-CA9 having a sequence selected from SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109, and SEQ ID NO:

113.

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

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

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

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

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

12. The fusion protein of any one of claims 1 and 3 to 11, wherein the half-life prolonging domain has SEQ ID NO:

117.

13. 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, P 13. The fusion protein of any one of claims 1 and 3 to 12, having a sequence selected from the group consisting of Pro476, Pro706, Pro709, Pro470, Pro471, Pro551, Pro552, Pro623, Pro624, Pro698, Pro655, Pro656, Pro657, Pro658, Pro516, Pro517, Pro518, and Pro519.

14. A nucleic acid encoding the fusion protein according to any one of claims 1 to 13.

15. An expression vector comprising the nucleic acid of claim 14.

16. A host cell comprising the expression vector of claim 15.

17. 17. A method of producing a fusion protein, comprising culturing the host cell of claim 16 under conditions in which the protein is expressed, and recovering the fusion protein.

18. A method for treating cancer, comprising administering to a patient the protein of any one of claims 1 to 13.

19. A single domain antigen binding domain that binds to human Trop2 having a sequence selected from SEQ ID NO:77, SEQ ID NO:81, SEQ ID NO:85, SEQ ID NO:89, and SEQ ID NO:

93.

20. A single domain antigen-binding domain that binds to human B7H3 having a sequence selected from SEQ ID NO:41, SEQ ID NO:45, SEQ ID NO:49, SEQ ID NO:53, and SEQ ID NO:

57.

21. A single domain antigen-binding domain that binds to human CA9 having a sequence selected from SEQ ID NO:101, SEQ ID NO:105, SEQ ID NO:109, and SEQ ID NO:

113.

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.

Images