EA040019B1 - BISPECIFIC CD33 AND CD3 BINDING PROTEINS - Google Patents

Description

Cross-references to related applications

The present application claims priority under U.S. Provisional Application No.

62/019795 filed July 1, 2014; U.S. Provisional Application No. 62/111470, filed Feb. 3, 2015; and is a continuation of ordinary US application number 14/642497, filed March 9, 2015, the contents of each of which are incorporated into this application by reference.

Sequence listing

The present application contains a sequence listing, which is incorporated by reference in its entirety. The sequence listing document created on June 30, 2015 is named 45375-704.601_SL.txt and has a size of 147965 bytes.

Background Art Information

Acute myeloid leukemia (AML) is an acute leukemia in adults and children. CD33 is expressed on most AML myeloblasts. Some publications report that CD33 expression is usually limited to early multilineage myeloid progenitors and is absent in normal pluripotent hematopoietic stem cells.

Brief description of the invention

Provided herein are binding proteins that specifically bind to human CD33 and bispecific binding proteins that specifically bind to human CD33 and human CD3. Also provided herein are anti-CD33 variable domains and anti-CD3 variable domains to generate a range of bispecific CD33/CD3 binding proteins, such as, for example, tandem diabodies. In addition, the present document also provides bispecific tandem diabodies that bind to CD33 and CD3 and their use in immunotherapy for acute myeloid leukemia (AML) and other hematological malignancies, disorders or conditions.

In particular, binding proteins are provided that show binding to human and cynomolgus CD33. The examples show that these CD33/CD3 tandem diabodies can redirect polyclonal CD3 ⁺ T cells from healthy donors as well as autologous T cells from AML patients to efficiently kill AML CD33 ⁺ cells at low E:T ratios. In this process, which depends on the presence of both CD33 ⁺ target cells and T cells, the redirected T cells are activated, as seen by the induction of CD25 and CD69, and stimulated to proliferate. It has been shown that the anti-AML effect of these tandem diabodies depends on the concentration of antibodies used, as well as the cellular E:T ratio. The tandem diabody is tetravalent and has two CD33 binding sites and two CD3 binding sites. A particular feature of the CD33/CD3 tandem diabodies described here is that they promote strong and efficient apoptosis resulting from bivalent binding that provides avidity for each antigen, namely CD33 and CD3.

Briefly, the proposed CD33/CD3 binding proteins described herein, in particular tandem diabodies, induce strong cytolysis of CD33 ⁺ leukemia cells and primary AML cells in vitro. Examples of bispecific CD33/CD3 binding proteins in tandem diabody antibody format demonstrate cytolytic activity in vivo in cell lines, primary AML cells, as well as in vivo models with patient-derived AML cell lines and primary AML cells. This indicates a high in vivo activity, especially noticeable in a rigorous PDX AML model. In addition, examples of bispecific CD33/CD3 binding proteins in tandem diabody antibody format demonstrate ex vivo cytolytic activity in samples obtained from patients at all stages of AML, including newly diagnosed, relapsed, and refractory patients.

In addition, these CD33/CD3 binding proteins described here are able to achieve significant lysis of CD33 expressing cells within about four hours. Thus, CD33/CD3 binding proteins exhibit high cytotoxicity at low CD33 densities on the cell surface, as well as high cytotoxicity at low effector:target (E:T) ratios. In addition, the CD33/CD3 binding proteins described here not only exhibit strong CD33 and CD3 binding affinities for human proteins, but also show excellent cross-reactivity with the corresponding cynomolgus monkey proteins, for example with human:cynomolgus KD ratios of 5 up to 0.2. In addition, the CD33/CD3 binding proteins described here do not show significant induction of cytokine release in the absence of CD33 ⁺ target cells, which is an important component of the safety profile of these molecules. In addition, the CD33/CD3 tandem diabodies described herein belong to a class of molecules that have half-lives in the range of approximately 8-24 hours, which should allow for convenient dosing.

In one aspect, provided herein are CD33 binding proteins that specifically bind to an epitope of human CD33. In some embodiments, the binding proteins comprise a heavy chain variable domain and a light chain variable domain of human origin.

In some embodiments, the CD33 binding protein has at least one

- 1 040019 a binding site comprising a light chain variable domain and a heavy chain variable domain, wherein the light chain variable domain comprises a CDR1 hypervariable region consisting of a sequence selected from the group consisting of SEQ ID NOs: 21-27, CDR2 consisting of a sequence selected from the group consisting of SEQ ID NOs: 28-34, and a CDR3 consisting of a sequence selected from the group consisting of SEQ ID NOs: 35-41.

In some embodiments, the CD33 binding protein has at least one binding site comprising a light chain variable domain and a heavy chain variable domain, wherein the heavy chain variable domain comprises a CDR1 hypervariable region consisting of a sequence selected from the group consisting of SEQ ID NOs: 42-48, a CDR2 consisting of a sequence selected from the group consisting of SEQ ID NOs: 49-55, and a CDR3 consisting of sequences selected from the group consisting of SEQ ID NOs: 56-63.

In certain cases, the CDR1, CDR2 and CDR3 of the light chain variable domain are selected from the group consisting of SEQ ID NOs: 21, 28 and 35; SEQ ID NOs: 22, 29 and 36; SEQ ID NOs: 23, 30 and 37; SEQ ID NOs: 24, 31 and 38; SEQ ID NOs: 25, 32 and 39; SEQ ID NOs: 26, 33 and 40; and SEQ ID NOs: 27, 34 and 41.

In certain instances, the heavy chain variable domain CDR1, CDR2 and CD3 are selected from the group consisting of SEQ ID NOs: 42, 49 and 56; SEQ ID NOs: 43, 50 and 57; SEQ ID NOs: 43, 50 and 58; SEQ ID NOs: 43, 50 and 59; SEQ ID NOs: 43, 50 and 60; SEQ ID NOs: 44, 51 and 61; SEQ ID NOs: 45, 52 and 62; SEQ ID NOs: 46, 53 and 63; SEQ ID NOs: 47, 54 and 63; and SEQ ID NOs: 48, 55 and 63.

In certain instances, the human CD33 binding site of the heavy chain variable domain and the light chain variable domain is selected from the group consisting of SEQ ID NO: 1 and SEQ ID NO: 11; SEQ ID NO: 2 and SEQ ID NO: 12; SEQ ID NO: 3 and SEQ ID NO: 13; SEQ ID NO: 4 and SEQ ID NO: 14; SEQ ID NO: 5 and SEQ ID NO: 15; SEQ ID NO: 6 and SEQ ID NO: 16; SEQ ID NO: 7 and SEQ ID NO: 17; SEQ ID NO: 8 and SEQ ID NO: 18; SEQ ID NO: 9 and SEQ ID NO: 19; and SEQ ID NO: 10 and SEQ ID NO: 20.

In some embodiments, the CD33 epitope is within ₆₂ DQEVQEETQ ₇₀ (SEQ ID NO: 94) (amino acid residues 62-70 of SEQ ID NO: 93) of human CD33.

In any of the above embodiments, the CD33 binding protein contains at least one additional functional domain. In some cases, the functional domain is an effector domain that binds to an effector cell. In certain instances, the effector domain is a CD3 binding site comprising at least one heavy chain variable domain and at least one light chain variable domain of an antibody, forming an antigen binding site for human CD3.

In certain cases, the CD3 binding site contains a heavy chain variable domain containing the STYAMN CDR1 sequence (SEQ ID NO: 72), the RIRSKYNNYATYYADSVKD CDR2 sequence (SEQ ID NO: 73) and the HGNFGNSYVSWFAY CDR3 sequence (SEQ ID NO: 74). In other cases, the CD3 binding site contains a light chain variable domain containing the CDR1 sequence of RSSTGAVTTSNYAN (SEQ ID NO: 90), the CDR2 sequence of GTNKRAP (SEQ ID NO: 91) and the CDR3 sequence of ALWYSNL (SEQ ID NO: 92).

In certain instances, the CD3 binding site comprises a heavy chain variable domain of SEQ ID NO: 64 and a light chain variable domain of SEQ ID NO: 68; a heavy chain variable domain of SEQ ID NO: 65 and a light chain variable domain of SEQ ID NO: 69; a heavy chain variable domain of SEQ ID NO: 66 and a light chain variable domain of SEQ ID NO: 70; or a heavy chain variable domain of SEQ ID NO: 67 and a light chain variable domain of SEQ ID NO: 71.

In any of the above embodiments, the CD33 binding protein is a dimeric protein. In any of the above embodiments, the CD33 binding protein is multifunctional.

In certain instances, the multifunctional CD33 binding protein has bispecificity for CD33 and CD3, with the binding specificities provided by the heavy chain variable domain and the light chain variable domains for CD33 and CD3 selected from the group consisting of SEQ ID NOS: 2, 12, 65 and 69; SEQ ID NOs: 3, 13, 65 and 69; SEQ ID NOs: 4, 14, 65 and 69; SEQ ID NOs: 5, 15, 65 and 69; SEQ ID NOs: 1, 11, 64 and 68; SEQ ID NOs: 2, 12, 64 and 68; SEQ ID NOs: 2, 12, 66 and 70; SEQ ID NOs: 4, 14, 66 and 70; SEQ ID NOs: 5, 15, 66 and 70; SEQ ID NOs: 3, 13, 64 and 68; SEQ ID NOs: 3, 13, 67 and 71; SEQ ID NOs: 4, 14, 64 and 68; SEQ ID NOs: 5, 15, 64 and 68; SEQ ID NOs: 7, 17, 64 and 68; SEQ ID NOs: 6, 16, 64 and 68; SEQ ID NOs: 6, 16, 67 and 71; SEQ ID NOs: 8, 18, 64 and 68; SEQ ID NOs: 9, 19, 64 and 68; SEQ ID NOs: 9, 19, 67 and 71; and SEQ ID NOs: 10, 20, 64 and 68.

In another aspect, provided herein are bispecific antigen-binding tandem diabodies specific for human CD3 and human CD33. In some embodiments, the tandem diabodies comprise a first polypeptide and a second polypeptide, each polypeptide having at least four sequentially linked chain variable domains, each polypeptide comprising a heavy chain variable domain specific for human CD33; light chain variable domain specific to

- 2 040019 for human CD33; a heavy chain variable domain specific for human CD3 and a light chain variable domain specific for human CD3, and in each polypeptide, four variable chain domains are sequentially linked to each other via peptide linkers L1, L2 and L3 in the following order: VL(CD3)-L1-VH(CD33)-L2VL(CD33)-L3-VH(CD3); VH(CD3)-L1-VL(CD33)-L2-VH(CD33)-L3-VL(CD3); VL(CD33)-L1-VH(CD3)-L2VL(CD3)-L3-VH(CD33); or VH(CD33)-L1-VL(CD3)-L2-VH(CD3)-L3-VL(CD33).

In some embodiments, the implementation of the VL domain specific for human CD33 contains a CDR1 consisting of a sequence selected from the group consisting of SEQ ID NO: 21-27; CDR2 consisting of a sequence selected from the group consisting of SEQ ID NO: 28-34; and a CDR3 consisting of a sequence selected from the group consisting of SEQ ID NOs: 35-41.

In some embodiments, the implementation of the VH domain specific for human CD33 contains a CDR1 consisting of a sequence selected from the group consisting of SEQ ID NO: 42-48; CDR2 consisting of a sequence selected from the group consisting of SEQ ID NO: 49-55; and a CDR3 consisting of sequences selected from the group consisting of SEQ ID NOs: 56-63.

In some embodiments, the CDR1, CDR2, and CDR3 of the human CD33-specific VL domain are sequences selected from the group consisting of SEQ ID NOs: 21, 28, and 35; SEQ ID NOs: 22, 29 and 36; SEQ ID NOs: 23, 30 and 37; SEQ ID NOs: 24, 31 and 38; SEQ ID NOs: 25, 32 and 39; SEQ ID NOs: 26, 33 and 40; and SEQ ID NOs: 27, 34 and 41.

In some embodiments, the CDR1, CDR2, and CDR3 of the human CD33-specific VH domain are sequences selected from the group consisting of SEQ ID NOs: 42, 49, and 56; SEQ ID NOs: 43, 50 and 57; SEQ ID NOs: 43, 50 and 58; SEQ ID NOs: 43, 50 and 59; SEQ ID NOs: 43, 50 and 60; SEQ ID NOs: 44, 51 and 61; SEQ ID NOs: 45, 52 and 62; SEQ ID NOs: 46, 53 and 63; SEQ ID NOs: 47, 54 and 63; and SEQ ID NOs: 48, 55 and 63.

In some embodiments, the VL and VH domains specific for CD33 are sequences selected from the group consisting of SEQ ID NO: 1 and SEQ ID NO: 11; SEQ ID NO: 2 and SEQ ID NO: 12; SEQ ID NO: 3 and SEQ ID NO: 13; SEQ ID NO: 4 and SEQ ID NO: 14; SEQ ID NO: 5 and SEQ ID NO: 15; SEQ ID NO: 6 and SEQ ID NO: 16; SEQ ID NO: 7 and SEQ ID NO: 17; SEQ ID NO: 8 and SEQ ID NO: 18; SEQ ID NO: 9 and SEQ ID NO: 19; and SEQ ID NO: 10 and SEQ ID NO: 20.

In some embodiments, the human CD3-specific VH domain comprises the STYAMN CDR1 sequence (SEQ ID NO: 72), the RIRSKYNNYATYYADSVKD CDR2 sequence (SEQ ID NO: 73), and the HGNFGNSYVSWFAY CDR3 sequence (SEQ ID NO: 74) or HGNFGNSYVSYFAY (SEQ ID NO: 75).

In some embodiments, the human CD3-specific VL domain comprises the RSSTGAVTTSNYAN CDR1 sequence (SEQ ID NO: 90), the GTNKRAP CDR2 sequence (SEQ ID NO: 91), and the ALWYSNL CDR3 sequence (SEQ ID NO: 92) .

In some embodiments, the CD3-specific VL and VH domains are sequences selected from the group consisting of SEQ ID NO: 64 and SEQ ID NO: 68; SEQ ID NO: 65 and SEQ ID NO: 69; SEQ ID NO: 66 and SEQ ID NO: 70; and SEQ ID NO: 67 and SEQ ID NO: 71.

In some embodiments, each polypeptide contains four chain variable domains selected from the group consisting of SEQ ID NOs: 2, 12, 65, and 69; SEQ ID nO: 3, 13, 65 and 69; SEQ ID NOs: 4, 14, 65 and 69; SEQ ID NOs: 5, 15, 65 and 69; SEQ ID NOs: 1, 11, 64 and 68; SEQ ID NOs: 2, 12, 64 and 68; SEQ ID NOs: 2, 12, 66 and 70; SEQ ID NOs: 4, 14, 66 and 70; SEQ ID NOs: 5, 15, 66 and 70; SEQ ID NOs: 3, 13, 64 and 68; SEQ ID NOs: 3, 13, 67 and 71; SEQ ID NOs: 4, 14, 64 and 68; SEQ ID NOs: 5, 15, 64 and 68; SEQ ID NOs: 7, 17, 64 and 68; SEQ ID NOs: 6, 16, 64 and 68; SEQ ID NOs: 6, 16, 67 and 71; SEQ ID NOs: 8, 18, 64 and 68; SEQ ID NOs: 9, 19, 64 and 68; SEQ ID NOs: 9, 19, 67 and 71; and SEQ ID NOs: 10, 20, 64 and 68.

In some embodiments, L1, L2, and L3 linkers are composed of about 12 or fewer amino acid residues. In certain instances, linkers L1, L2 and L3 are each independently GGSGGS (SEQ ID NO: 95), GGSG (SEQ ID NO: 96) or GGSGG (SEQ ID NO: 97). In other cases, linkers L1 and L3 are GGSGGS (SEQ ID NO: 95) and linker L2 is GGSG (SEQ ID NO: 96) or GGSGG (SEQ ID NO: 97).

In some embodiments, the bispecific tandem diabody has a sequence selected from the group consisting of SEQ ID NOs: 98-121. In other embodiments, the bispecific tandem diabody is tandem diabody 01 (SEQ ID NO: 98), 02 (SEQ ID NO: 99), 03 (SEQ ID NO: 100), 04 (SEQ ID NO: 101), 05 (SEQ ID NO: 102), 06 (SEQ ID NO: 103), 07 (SEQ ID NO: 104), 08 (SEQ ID NO: 105), 09 (SEQ ID NO: 106), 10 (SEQ ID NO: 107) , 11 (SEQ ID NO: 108), 12 (SEQ ID NO: 109), 13 (SEQ ID NO: 110), 14 (SEQ ID NO: 111), 15 (SEQ ID NO: 112), 16 (SEQ ID NO: 113), 17 (SEQ ID NO: 114), 18 (SEQ ID NO: 115), 19 (SEQ ID NO: 116), 20 (SEQ ID NO: 117), 21 (SEQ ID NO: 118), 22 (SEQ ID NO: 119), 23 (SEQ ID NO: 120) or 24 (SEQ ID NO: 121).

In some embodiments, the bispecific antigen-binding tandem diabodies exhibit binding with a dissociation constant K _D of 10 nM or less to CD33 on CD33 ⁺ tumor cells selected from HL-60, KG-1, and U-937.

In some embodiments, the bispecific antigen-binding tandem diabodies

- 3 040019 bind specifically to an epitope of human CD33 that is within

62DQEVQEETQ70 (SEQ ID NO: 94) (amino acid residues 62-70 of SEQ ID NO: 93) of human CD33.

In another aspect, provided herein are polynucleotides encoding a CD33 binding protein or a bispecific tandem diabody according to any of the above embodiments. In another aspect, provided herein are vectors containing the described polynucleotides. In another aspect, provided herein are host cells transformed with the described vectors.

In yet another aspect, provided herein are pharmaceutical compositions comprising a CD33 binding protein or a bispecific tandem diabody according to any of the above embodiments and a pharmaceutically acceptable carrier.

In yet another aspect, provided herein are methods for producing a CD33 binding protein or a bispecific tandem diabody according to any of the above embodiments, comprising introducing into a host cell a polynucleotide encoding a CD33 binding protein or a bispecific tandem diabody according to any of the above embodiments. , or a vector containing the described polynucleotides, culturing the host cell under conditions that express the CD33 binding protein or the bispecific tandem diabody, and purifying the expressed CD33 binding protein or the bispecific tandem diabody.

In addition, provided herein are methods for treating CD33 ⁺ cancer comprising administering a bispecific tandem diabody according to any of the above embodiments to an individual suffering from CD33 ⁺ cancer. In some embodiments, the CD33 ⁺ cancer is acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), progenitor B-cell lymphoblastic leukemia, myeloid sarcoma, multiple myeloma, acute lymphoma, acute lymphoblastic lymphoma, or chronic myelomonocytic leukemia ( CMML). In some embodiments, the CD33 ⁺ cancer is acute myeloid leukemia (AML). In some embodiments, the CD33 ⁺ cancer is multiple myeloma. In some embodiments, the CD33 ⁺ cancer is acute lymphoblastic leukemia (ALL).

In addition, provided herein are methods for treating acute myeloid leukemia (AML) comprising administering a bispecific tandem diabody according to any of the above embodiments to an individual suffering from AML. In some embodiments, the AML is AML with recurrent genetic abnormalities; AML with changes associated with myelodysplasia; myeloid neoplasms associated with therapy; myeloid sarcoma; myeloid proliferations associated with Down syndrome; neoplasm of blast plasmacytoid dendritic cells or AML, not otherwise classified. In some embodiments, the AML is AML-M0, AML-M1, AML-M2, AML-M3, AML-M4, AML-M5, AML-M6, or AML-M7. In other embodiments, the AML is newly diagnosed, relapsed, or refractory.

In addition, provided herein are methods for treating myeloid dysplastic syndrome (MDS) comprising administering a bispecific tandem diabody according to any of the above embodiments to an individual suffering from MDS.

In addition, provided herein are methods for treating myeloproliferative disease (MPD) comprising administering a bispecific tandem diabody according to any of the above embodiments to an individual suffering from MPD.

In addition, provided herein are methods for treating chronic myelomonocytic leukemia (CMML) comprising administering a bispecific tandem diabody according to any of the above embodiments to an individual suffering from CMML.

In addition, provided herein are methods for treating immunosuppression with myeloid-derived suppressor cells (MDSC), comprising administering a bispecific tandem diabody according to any of the above embodiments to an immunosuppressed individual.

The above methods of treatment in certain cases additionally include the administration of cytarabine, azacitidine, decitabine, anthracycline (for example, daunorubicin, idarubicin, doxorubicin, etc.), amsacrine, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast, epipodophyllotoxin (eg, etoposide, teniposide, etc.), anthracenedione (eg, mitoxantrone, pixantrone, losoxantrone, pyroxantrone, amethantron, etc.), an anti-CD20 agent (eg. , rituximab, ocrelizumab, ofatumumab, etc.) or combinations thereof.

Inclusion by reference

All publications, patents and patent applications mentioned in the specification are hereby incorporated by reference to the same extent as if it were specifically and individually indicated that each individual publication, patent or patent application is incorporated by reference.

- 4 040019

Brief description of the drawings

The novel features of the invention are set forth in more detail in the appended claims. A more complete understanding of the features and advantages of the present invention will be achieved by reference to the following detailed description, which sets forth illustrative embodiments where the principles of the invention are used, and in the accompanying drawings of which in FIG. 1 is a schematic representation of the gene organization and domain order of CD3/CD33 tandem diabodies (TandAb®). Tandem diabodies are expressed as a single polypeptide consisting of four variable domains connected via short peptide linkers L1, L2 and L3. Upon expression, the two monomeric polypeptides are non-covalently linked head-to-tail to form a functional homodimeric tandem diabody molecule. L1, L2, L3: linker; VH: heavy chain variable domain; V _L : light chain variable domain;

in fig. 2 - capture of CD3 by a tandem diabody and its mechanism of action. Tandem diabodies are tetravalent bispecific proteins that can capture cytotoxic T cells by binding to CD3. The tandem diabody binds to the CD33 ⁺ tumor cell via two of the four binding domains, and to CD3 to others via the other two binding domains. This T cell/target cell binding (crosslinking) event promotes T cell activation and stimulates subsequent tumor cell destruction via antibody dependent cellular cytotoxicity (ADCC);

in fig. 3 - domain order variants of CD33/CD3 tandem diabodies. Order variants of the variable heavy (VH) and variable light (VL) chain domains within gene sequences encoding tandem diabodies allow the production of antibodies with CD33 and CD3 specificities localized on the inner or outer surface of the molecule. Domain specificities, location of signal sequences (ss) and linkers (L1, L2, L3), and affinity tags (His), as well as 5' and 3' ends are indicated;

in fig. 4 is a comparison of positively enriched T cells from healthy donors with negatively selected T cells from healthy donors. Acute myelogenous leukemia KG-1a cells were incubated with 10 pM (approximately 1 ng/ml) and 25 pM (approximately 2.5 ng/ml) of one of 10 selected tandem diabodies and negatively selected T cells from healthy donors or positively selected T cells from healthy donors at an E:T cell ratio of 1:1 or 3:1 as indicated. After 48 hours, the number of cells was determined and cytotoxicity was assessed using DAPI staining. Results are expressed as mean±SEM for percent dead cells (upper panels) and percent specific cytotoxicity (lower panels) from 3 independent experiments performed in duplicated wells;

in fig. 5 - analysis strategy. Light scatter plots and histograms plotted from analysis of one T cell aliquot from a healthy donor, 1 representative AML cell line (HL60), and a primary AML sample (AMP002), each illustrating a strategy for determining tandem diabody-induced cytotoxicity. FSC, direct light scattering; SSC, side scatter;

in fig. 6 - Screening assay for cytotoxicity in CD33+ AML cell lines. HL-60 (A,B) and KG-1a (C,D) parental cells were incubated with 10 pM (approximately 1 ng/ml) and 25 pM (approximately 2.5 ng/ml) of one of the 22 CD33/CD3- tandem diabodies or non-binding control tandem diabody (00) and healthy donor T cells at an E:T cell ratio of 1:1 (A,C) or 5:1 (B,D) as indicated. After 48 hours, cell numbers were determined and cytotoxicity was assessed by DAPI staining to quantify drug-dependent cytotoxicity. The results are expressed as the mean ±SEM for the percentage of DAPI+ cells from 3 independent experiments performed in duplicate wells. Qualitatively similar results were obtained when cytotoxicity was expressed as a percentage of specific cytotoxicity;

in fig. 7 - sampling of primary AML for research. Frozen aliquots were prepared for analysis from total samples of primary human AML. The percentage of AML blasts after thawing was determined by flow cytometry based on CD45/side scatter properties. The viability of the samples was determined after thawing and also after 48 h in cytokine-containing liquid culture (without the addition of tandem diabody molecules or T cells from healthy donors) by flow cytometry using DAPI as a live/dead cell marker. Viability results after thawing and also after 48 hours are shown for all samples that contained >58% AML blasts. Square: Primary AML samples that showed >50% viability on thaw and >50% viability at 48 h in cytokine-containing liquid culture that were included in the final assays;

in fig. 8 - tandem diabody-induced cytotoxicity in primary AML samples. Primary AML samples were incubated with 2.5 pM (approximately 250 pg/ml), 10 pM (approximately 1 ng/ml) and 25 pM (approximately 2.5 ng/ml) of one of the 9 tandem diabody molecules without the addition of T-5 040019 cells from a healthy donor (A) or with T cells from a healthy donor at an E:T cell ratio

1:3 (B) or 1:1 (C) as indicated. After 48 hours, cell numbers were determined and cytotoxicity was assessed by DAPI staining to quantify drug-dependent cytotoxicity. Results are expressed as mean±SEM for percent specific cytotoxicity from experiments performed in double wells;

in fig. 9A shows the sequence of the extracellular domain of human CD33 (amino acid residues (aa) 18-259) (SEQ ID NO: 93);

in fig. 9B is the complete sequence of tandem diabody 1 (SEQ ID NO: 98);

in fig. 9C is the complete sequence of tandem diabody 2 (SEQ ID NO: 99);

in fig. 9D is the complete sequence of tandem diabody 3 (SEQ ID NO: 100);

in fig. 9E is the complete sequence of tandem diabody 4 (SEQ ID NO: 101);

in fig. 9F is the complete sequence of tandem diabody 5 (SEQ ID NO: 102);

in fig. 9G is the complete sequence of tandem diabody 6 (SEQ ID NO: 103);

in fig. 9H is the complete sequence of tandem diabody 7 (SEQ ID NO: 104);

in fig. 9I is the complete sequence of tandem diabody 8 (SEQ ID NO: 105);

in fig. 9J is the complete sequence of tandem diabody 9 (SEQ ID NO: 106);

in fig. 9K is the complete sequence of tandem diabody 10 (SEQ ID NO: 107);

in fig. 9L - full sequence of tandem diabody 11 (SEQ ID NO: 108);

in fig. 9M - full sequence of tandem diabody 12 (SEQ ID NO: 109);

in fig. 9N - full sequence of tandem diabody 13 (SEQ ID NO:110);

in fig. 9O - full sequence of tandem diabody 14 (SEQ ID NO:111);

in fig. 9P is the complete sequence of tandem diabody 15 (SEQ ID NO: 112);

in fig. 9Q is the complete sequence of tandem diabody 16 (SEQ ID NO: 113);

in fig. 9R is the complete sequence of tandem diabody 17 (SEQ ID NO: 114);

in fig. 9S - full sequence of tandem diabody 18 (SEQ ID NO: 115);

in fig. 9T is the complete sequence of tandem diabody 19 (SEQ ID NO: 116);

in fig. 9U - full sequence of tandem diabody 20 (SEQ ID NO: 117);

in fig. 9V is the complete sequence of tandem diabody 21 (SEQ ID NO: 118);

in fig. 9W - full sequence of tandem diabody 22 (SEQ ID NO: 119);

in fig. 9X is the complete sequence of tandem diabody 23 (SEQ ID NO: 120); and in FIG. 9Y - full sequence of tandem diabody 24 (SEQ ID NO: 121). Underlined sequences are L1, L2 and L3 linkers;

in fig. 10 shows the effect of tandem diabodies 16 and 12 on the growth of HL-60 cells in NOD/scid mice. Eight experimental groups of immunodeficient NOD/scid mice were xenotransplanted by subcutaneous injection of a suspension of 4x106 HL-60 cells on day 0. Prior to injection, HL-60 cells were mixed with 3x106 purified T cells from healthy donors. All animals of the experimental groups undergoing tumor cell and T cell transplantation received intravenous boluses on days 0, 1, 2, 3 and 4 of vehicle (control) or tandem diabody 16 or 12 at three different doses as indicated (0.1 , 1 and 10 μg). One group without treatment with effector cells and vehicle served as an additional negative control;

in fig. 11 Antitumor activity of tandem diabody 16 in an AML xenograft model. NOD/scid mice were sublethally irradiated (2 Gy) and subcutaneously inoculated with 4x106 HL-60 cells. On day 9, animals received a single bolus injection of anti-asialo GM1 rabbit antibody (Ab). When the tumors reached a volume of 50-150 mm ³ (mean value 73 ± 11 mm ³ ) on day 10, the animals were divided into 3 treatment groups. Groups 2 and 3 (n=8) were injected intraperitoneally with 1.5x107 expanded and activated human T cells. Between days 13 and 21 (qdxd9), animals received tandem diabody 16 (group 3) or vehicle in the lateral tail vein (group 1 and group 2);

in fig. 12 - relative number (A) and absolute number (B) of human AML blasts in bone marrow (BM) and spleen of NSG mice on day 38 after treatment with 5 μg (0.25 mg/kg) or 50 μg (2.5 mg/kg). kg) CD33/CD3 tandem antibodies 12 and 16;

in fig. 13 - Kinetics of target cell lysis mediated by CD33/CD3 tandem diabody 16. ^1x104 Calcein labeled HL-60 target cells were incubated with primary human T cells as effector cells at an E:T ratio of 25:1 in the presence of serial dilutions of tandem diabody 16 or no antibody (w/o) for 30 minutes, 1, 2, 3, 4, or 5 hours. At each time point, fluorescent calcein released from lysed target cells was used to calculate specific lysis. The graph was applied to the average values and standard deviation (SD), which is the result of experiments conducted in triplicate;

in fig. 14 - Kinetics of EC ₅₀ values and specific lysis for CD33/CD3 tandem diabody 16. EC ₅₀ values (solid dots) and tandem diabody 16 mediated target cell lysis (open squares) were determined in cytotoxicity assays with calcein release at

- 6 040019 indicated values of the incubation time using a non-linear regression/sigmoid dose-response curve and plotted;

in fig. 15 - cytotoxic activity in samples from a patient with newly diagnosed, relapsed and refractory AML.

Detailed description of the invention

According to a first aspect, the document describes binding proteins having specificity for at least CD33, preferably human CD33. In some embodiments, the CD33 binding proteins are specific for human and cynomolgus CD33, i.e., are cross-reactive. In some embodiments, these cross-reactive binding proteins bind to human and cynomolgus monkey CD33 with similar affinity.

CD33 is expressed on myeloid cells, such as, for example, acute myeloid leukemia (AML) blasts. Antibody libraries can be screened to isolate antibody domains specific for CD33, such as human CD33. For example, IgM phage display libraries can be screened using, for example, a recombinant CD33-Fc fusion protein containing amino acids 1-243 of the human CD33 extracellular domain (FIG. 9A, SEQ ID NO: 93).

In some embodiments, the CD33 binding protein has at least one CD33 binding site comprising a light chain variable domain and a heavy chain variable domain. The light chain variable domain contains light chain CDR1, CDR2 and CDR3, and the heavy chain variable domain contains heavy chain CDR1, CDR2 and CDR3. In some embodiments, these light chain CDRs (CDR1, CDR2, and CDR3) are selected from the human CDR sequences shown in Table 1. 1 (SEQ ID NO: 21-41). In certain instances, the light chain CDR1 is selected from SEQ ID NOs: 21-27. In certain instances, the light chain CDR2 is selected from SEQ ID NOs: 28-34. In certain instances, the light chain CDR3 is selected from SEQ ID NOs: 35-41.

In some embodiments, these heavy chain CDRs (heavy chain CDR1, CDR2, and CDR3) are selected from the human CDR sequences shown in Table 1. 2 (SEQ ID NOS: 42-63). In certain instances, the heavy chain CDR1 is selected from SEQ ID NOs: 42-48. In certain instances, the heavy chain CDR2 is selected from SEQ ID NOs: 49-55. In certain instances, the heavy chain CDR3 is selected from SEQ ID NOs: 56-63.

In some embodiments, the light and heavy chain CDRs are selected without surrounding framework sequences of the respective variable domains, which include framework sequences from other immunoglobulins or consensus framework regions, optionally further mutated and/or substituted with other suitable framework sequences. Thus, provided herein in some embodiments is a CD33 binding protein comprising a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 21; The light chain CDR2 is SEQ ID NO: 28; and the light chain CDR3 is SEQ ID NO: 35. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 22; The light chain CDR2 is SEQ ID NO: 29; and the light chain CDR3 is SEQ ID NO: 36. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 23; The light chain CDR2 is SEQ ID NO: 30; and the light chain CDR3 is SEQ ID NO: 37. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 24; The light chain CDR2 is SEQ ID NO: 31; and the light chain CDR3 is SEQ ID NO: 38. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 25; The light chain CDR2 is SEQ ID NO: 32; and the light chain CDR3 is SEQ ID NO: 39. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 26; The light chain CDR2 is SEQ ID NO: 33; and the light chain CDR3 is SEQ ID NO: 40. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the light chain CDR1 is SEQ ID NO: 27; The light chain CDR2 is SEQ ID NO: 34 and the light chain CDR3 is SEQ ID NO: 41.

Also provided herein in some embodiments is a CD33 binding protein comprising a heavy chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 42; The heavy chain CDR2 is SEQ ID NO: 49; and the heavy chain CDR3 is SEQ ID NO: 56. In some embodiments, the CD33 binding protein comprises a heavy chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 43; The heavy chain CDR2 is SEQ ID NO: 50 and the heavy chain CDR3 is SEQ ID NO: 57. In some embodiments, the CD33 binding protein

- 7 040019 contains the variable domain of the heavy chain, while the CDR1 of the heavy chain is SEQ ID NO: 43; The heavy chain CDR2 is SEQ ID NO: 50; and the heavy chain CDR3 is SEQ ID NO: 58. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 43; The heavy chain CDR2 is SEQ ID NO: 50; and the heavy chain CDR3 is SEQ ID NO: 59. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 43; The heavy chain CDR2 is SEQ ID NO: 50; and the heavy chain CDR3 is SEQ ID NO: 60. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 44; The heavy chain CDR2 is SEQ ID NO: 51; and the heavy chain CDR3 is SEQ ID NO: 61. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 45; The heavy chain CDR2 is SEQ ID NO: 52; and the heavy chain CDR3 is SEQ ID NO: 62. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 46; The heavy chain CDR2 is SEQ ID NO: 53; and the heavy chain CDR3 is SEQ ID NO: 63. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 47; The heavy chain CDR2 is SEQ ID NO: 54; and the heavy chain CDR3 is SEQ ID NO: 63. In some embodiments, the CD33 binding protein comprises a light chain variable domain, wherein the heavy chain CDR1 is SEQ ID NO: 48; The heavy chain CDR2 is SEQ ID NO: 55 and the heavy chain CDR3 is SEQ ID NO: 63.

In other embodiments, the CD33 binding protein comprises a light chain variable domain selected from the amino acid sequences of SEQ ID NOs: 1-10 shown in Table 1. 3. In other embodiments, the implementation of the CD33-binding protein contains a heavy chain variable domain selected from the amino acid sequences of SEQ ID NO: 11-20 shown in table. 4. In yet other embodiments, the CD33 binding protein comprises a light chain variable domain selected from the amino acid sequences of SEQ ID NOs: 1-10 shown in Table 1. 3 and a heavy chain variable domain selected from the amino acid sequences of SEQ ID NOs: 11-20 shown in Table 3. 4.

The term binding protein refers to an immunoglobulin derivative with antigen-binding properties, ie immunoglobulin polypeptides or fragments thereof, which contain an antigen-binding site. The binding protein contains the variable domains of the antibody or its fragments. Each antigen-binding domain is formed by a heavy chain variable domain (VH) of an antibody, i.e. an immunoglobulin, and a light chain variable domain (VL) of an antibody linked to the same epitope, with the heavy chain variable domain (VH) containing three complementarity-determining regions ( CDR) heavy chain: CDR1, CDR2 and CDR3; and the light chain variable region (VL) contains three light chain complementarity determining regions (CDRs): CDR1, CDR2 and CDR3. In some cases, the binding protein according to some embodiments herein does not contain immunoglobulin constant domains. In some instances, the light and heavy chain variable domains forming the antigen binding site are covalently linked to each other, such as via a peptide linker, or, in other instances, the light and heavy chain variable domains are non-covalently linked to each other to form the antigen binding site. The term binding protein also refers to antibody fragments or antibody derivatives, including, for example, Fab fragments, Fab', F(ab') ₂ , Fv, single chain Fv, tandem single chain Fv ((scFv) ₂ , bispecific T-cell antibodies (Bi -specific T-cell engagers, BiTE®), dual affinity retargeting antibodies (DART™), diabody and tandem diabody (TandAb®). , three or more binding sites for CD33.

- 8 040019

Table 1. Amino acid sequences of CDR1, CDR2, and CDR3 of the anti-CD33 light chain variable region

CDR Sequence ID Light chain CDR sequence CDR1 SEQIDNO:21 GGNNGSTTVH SEQ ID NO:22 SGSRSNIGSNTVN SEQ ID NO:23 SGSSSNIGSNTVN SEQ ID NO:24 TGSSSNIGAGYDVH SEQ ID NO:25 SGSSSNIGSNIVN SEQ ID NO:26 SGSSSNIGSNTVK SEQ ID NO:27 SGSSSNIGDNWN CDR2 SEQ ID NO:28 DDNERPS SEQ ID NO:29 GNNQRPS SEQ ID NO:30 SDNQRPS SEQIDNO:31 GNSNRPS SEQ ID NO:32 SNNQRPS SEQ ID NO:33 SNNQRSS SEQ ID NO:34 STNKRPS CDR3 SEQ ID NO:35 QVWDSGSDH SEQ ID NO:36 ATWDDSLIG SEQ ID NO:37 ATWDDSLNG SEQ ID NO:38 QSYDSSLSD SEQ ID NO:39 AAWDDSLKG SEQ ID NO:40 AAWDDSLNG SEQ IDNO:41 AAWDDSLSA

- 9 040019

Table 2. Amino acid sequences of CDR1, CDR2, and CDR3 of the anti-CD33 heavy chain variable region

CDR Sequence identifier Heavy chain CDR sequence CDR1 SEQ ID NO:42 SNYGIH SEQ ID NO:43 TSYDIN SEQ ID NO:44 TSYYMH SEQ ID NO:45 TSYWIG SEQ ID NO:46 SSYAIS SEQ ID NO:47 SSYGIS SEQ ID NO:48 DSYAIS CDR2 SEQ ID NO:49 LISYDGNKKFYADSVKG SEQ ID NO:50 WMNPNSGNTGFAQKFQG SEQIDNO:51 GIINPSGGSTSYAQKFQG SEQ ID NO:52 HYPGDSDTRYSPSFQG SEQ ID NO:53 GIYPIFGSANYAQKFQG SEQ ID NO:54 GIIPIFGSAHYAQKFQG SEQ ID NO:55 GIIPIFGSAHYSQKFQG CDR3 SEQ ID NO:56 DRLESAAFDY SEQ ID NO:57 SEQ ID NO:58 SEQ ID NO:59 SEQ ID NO:60 DRANTDFSYGMDV DRAVTDYYYGMDV DRANTDYSFGMDV DRANTDYSLGMDV SEQIDNO:61 DVVPAAIDYYGMDV SEQ ID NO:62 HKRGSDAFDI SEQ ID NO:63 EYYYDSSEWAFDI

- 10 040019

Table 3. Amino acid sequences of all aHTu-CD33 light chain variable domains (amino acid sequences of CDR1, CDR2, and CDR3 of the light chain variable region are in bold and underlined)

clone antiCD33 Sequence identifier Light chain variable domain (VL) sequence 01 SEQIDNO:1 SYELTOPPSVSVAPGOTAMITCGGNNIGSTTVHWYOQKPGOAPVLVV YDDNERPSGIPERFSGSNSGSTATLTINRVEAGDEADYYCOVWDSGSD HVVFGGGTKLTVL 02 SEQ ID NO:2 OSVLTOPPSASGTPGORVTISCSGSRSNIGSNTVNWYOOLPGTAPKLLI YGNNQRPSGVPDRFSGSKSGSSASLAISGLOSEDEADYYCATWDDSLI GWVFGGGTKLTVL 03 SEQ ID NO:3 OSVLTOPPSASGTPGORVTISCSGSRSNIGSNTVNWYOQLPGTAPKLLI YGNNQRPSGVPDRFSSGSKSGTSASLAISGLQSEDEADYYCATWDDSLI GWVFGGGTKLTVL 04 SEQ ID NO:4 OSVLTOPPSASGTPGORVTISCSGSRSNIGSNTVNWYOQLPGTAPKLLI YGNNQRPSGVPDRFSSGSKSGTSASLAISGLQSEDEADYYCATWDDSLI GWVFGGGTKLTVL 05 SEQ ID NO:5 OSVLTOPPSASGTPGORVTISCSGSRSNIGSNTVNWYOQLPGTAPKLLI YGNNORPSGVPDRFSGSKSGTSASLAISGLQSEDEADYYCATWDDSLI GWVFGGGTKLTVL 06 SEQ ID NO:6 OSVLTOPPSASGTPGORVTISCSGSSSNIGSNTVNWYOQLPGTAPKLLI YSDNQRPSGVPDRFSGSKSGSSASLAISGLOSDDEADYYCATWDDSLN GAVFGGGTKLTVL 07 SEQ ID NO:7 OSVLTOPPSVSGAPGORVTISCTGSSSNIGAGYDVHWYOQLPGTAPKL LIYGNSNRPSGVPDRFSGSKSGTSASLAITGLOAEDEADYYCOSYDSSL SDVVFGGGTKLTVL 08 SEQ ID NO:8 OSVLTOPPSASGTPGORVTISCSGSSSNIGSNIVNWYOQLPGTAPKLLIY SNNORPSGVPDRFSGSKSGTSASLAISGLQSEDEADYYCAAWDDSLKG YVFGGGTKLTVL 09 SEQ ID NO:9 OSVLTOPPSASGTPGORVTISCSGSSSNIGSNTVKWYOQLPGTAPKLLI YSNNQRSSGVPDRFSGSKSGSSASLAISGLOSEDEADYYCAAWDDSLN GYVFGGGTKLTVL 10 SEQ ID NO: 10 OSVLTOPPSASGTPGORVTISCSGSSSNIGDNWNWYOOLPGTAPKLLI YSTNKRPSGVPDRFSGSKSGSSASLAISGLOSEDEADYYCAAWDDSLS AYVFGGGTKLTVL

- 11 040019

Table 4. Amino acid sequence of the heavy chain variable domain of aHTu-CD33 (amino acid sequences of CDR1, CDR2 and CDR3 of the heavy chain variable region are in bold and underlined)

AntiCD33 clone Sequence ID Heavy chain variable domain (VH) sequence 01 SEQIDNO:11 OVOLOESGGGWOPGRSLRLSCAASGFSFSNYGIHWVROAPGKGLEWVA LISYDGNKKFYADSVKGRFAISRDTSKNTVDLQMTSLRPEDTAVYYCAK DRLESAAFDYWGQGTLVTVSS 02 SEQ ID NO: 12 QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYDINWVRQAPGQGLEWM GWMNPNSGNTGFAOKFOGRVTMTRDTSTSTVYMELSSLRSEDTAVYYC ARDRANTDFSYGMDVWGQGTLVTVSS 03 SEQ ID NO: 13 VT VS S 04 SEQ ID NO: 14 QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYDINWVRQAPGQGLEWM GWMNPNSGNTGFAOKFOGRVTMTRDTSTSTVYMELSSLRSEDTAVYYC ARDRANTDYSFGMDVWGQGTLVTVSS 05 SEQ ID NO: 15 OVOLVOSGAEVKKPGASVKVSCKASGYTFTSYDINWVRQAPGOGLEWM GWMNPNSGNTGFAOKFOGRVTMTRDTSTSTVYMELSSLRSEDTAVYYC ARDRANTDYSLGMDVWGQGTLVTVSS 06 SEQ ID NO: 16 OVOLVQSGAEVKKPGASVKVSCKASGYTFTSYYMHWVROAPGOGLEW MGIINPSGGSTSYAOKFOGRVTMTRDTSTSTVYMELSSLRSEDTAVYYC ARDWPAAIDYYGMDVWGQGTTVTVSS 07 SEQ ID NO: 17 OVOLVQSGAEVKKPGESLKISCKGSGYSFTSYWIGWVROMPGKGLEWM GIIYPGDSDTRYSPSFOGOVTISADKSISTAYLQWSSLKASDTAMYYCAR HKRGSDAFDIWGQGTTVTVSS 08 SEQ ID NO: 18 OVOLVOSGAEVKKPGSSVKVSCKASGGTFSSYAISWVROAPGOGLEWMG GIYPIFGSANYAQKFQGRVTITADESTSTAYMELSSLRSEDTAVYYCARE YYYDSSEWAFDIWGQGTLVTVSS 09 SEQ ID NO: 19 OVOLVOSGAEVKKPGSSVKVSCKASGGTFSSYGISWVROAPGOGLEWM GGIIPIFGSAHYAOKFOGRVTITADESTSTAYMELSSLRSEDTAVYYCAR E YYYDSSEWAFDIWGQGTLVTVSS 10 SEQ ID NO:20 OVOLVQSGAEVKKPGSSVKVSCKASGGTFDSYAISWVROAPGOGLEWM GGIIPIFGSAHYSOKFOGRVTITADESTSTAYMELSSLRSEDTAVYYCARE YYYDSSEWAFDIWGQGTLVTVSS

In some embodiments, a binding protein conferring specificity for CD33 is selected from one of the following combinations of a heavy chain variable domain and a light chain variable domain forming a binding site for human CD33, as shown in Table 1. 3 and table. 4. Non-limiting examples include: (i) SEQ ID NO: 1 and SEQ ID NO: 11; (ii) SEQ ID NO: 2 and SEQ ID NO: 12; (iii) SEQ ID NO: 3 and SEQ ID NO: 13; (iv) SEQ ID NO: 4 and SEQ ID NO: 14; (v) SEQ ID NO: 5 and SEQ ID NO: 15; (vi) SEQ ID NO: 6 and SEQ ID NO: 16; (vii) SEQ ID NO: 7 and SEQ ID NO: 17; (viii) SEQ ID NO: 8 and SEQ ID NO: 18; (ix) SEQ ID NO: 9 and SEQ ID NO: 19; and (x) SEQ ID NO: 10 and SEQ ID NO: 20.

Also described herein are binding proteins that not only have specificity for CD33, but that also have at least one additional functional domain. In a further embodiment, at least one additional functional domain is an effector domain. The effector domain contains the binding site of an antibody specific for the effector cell, which can stimulate or initiate cytotoxicity, phagocytosis, antigen presentation, cytokine release. Such effector cells are, for example, but not limited to, T cells. In particular, the effector domain comprises at least one antibody heavy chain variable domain and at least one antibody light chain variable domain forming an antigen-binding site for an antigen on T cells, such as, for example, human CD3.

Thus, in some embodiments, the CD33 binding protein is multifunctional. As used herein, the term multifunctional means that the binding protein exhibits two or more different biological functions. For example, different biological functions represent different specificities for different antigens. In certain instances, a multifunctional CD33 binding protein is multispecific, that is, it has binding specificity for CD33 and one or more additional antigens. In certain instances, the binding protein is bispecific with specificities for CD33 and CD3. Such bispecific binding proteins include, for example, bispecific monoclonal antibodies of the classes IgA, IgD, IgE, IgG or IgM, diabodies, single chain diabodies (scDb), tandem single chain Fv (scFv) ₂ , for example bispecific T-cell antibodies (Bispecific T-cell engagers , BiTE®), dual affinity retargeting antibodies (DART™), tandem diabodies (TandAb®), and Flexibodies.

In certain embodiments, the CD3-binding region of the bispecific CD33- and

- 12 040019

The CD3-binding protein has specificity for human CD3 and, in some cases, cynomolgus CD3. Examples of such a binding site are polypeptides containing CDR1, CDR2 and CDR3 of the VH domain from the sequences shown in table. 5 (SEQ ID NO: 64-67), and CDR1, CDR2 and CDR3 of the VL domain from the sequence shown in table. 6 (SEQ ID NOS: 68-71). In certain instances, the CD3 binding site is a combination of a heavy chain variable domain of the sequence of SEQ ID NO: 64 and a light chain variable domain of the sequence of SEQ ID NO: 68. In certain instances, the CD3 binding site is a combination of a heavy chain variable domain of the sequence of SEQ ID NO: 65 and a light chain variable domain of SEQ ID NO: 69. In certain instances, the CD3 binding site is a combination of a heavy chain variable domain of SEQ ID NO: 66 and a light chain variable domain of SEQ ID NO: 70. In certain instances, the CD3 binding site is a combination of a heavy chain variable domain of SEQ ID NO: 67 and a light chain variable domain of SEQ ID NO: 71.

Table 5. Amino acid sequence of the anti-CD3 heavy chain variable domain (amino acid sequences of CDR1, CDR2, and CDR3 of the heavy chain variable domains are in bold and underlined)

anti-SVZ VH domain sequence SEQ ID NO:64CD3-01 EVOLVESGGGLVQPGGSLRLSCAASGFTFSTYAMNWVROAPGKGLEWVGRIRSKYNNY ATYYADSVKDRFTISRDDSKNSLYLQMNSLKTEDTAVYYCARHGNFGNSYVSYFAYWG QGTLVTVSS SEQ ID NO:65 CD3-02 EVOLVESGGGLVQPGGSLRLSCAASGFTFSTYAMNWVROAPGKGLEWVGRIRSKYNNY ATYYADSVKDRFTISRDDSKNSLYLQMNSLKTEDTAVYYCARHGNFGNSYVSWFAYWG QGTLVTVSS SEQ ID NO:66CD3-03 EVOLVESGGGLVQPGGSLRLSCAASGFTFSTYAMNWVROAPGKGLEWVGRIRSKYNNY ATYYADSVKDRFTISRDDSKNSLYLQMNSLKTEDTAVYYCARHGNFGNSYVSWFAYWG QGTLVTVSS SEQ ID NO:67CD3-04 EVOLVESGGGLVQPGGSLRLSCAASGFTFSTYAMNWVROAPGKGLEWVGRIRSKYNNY ATYYADSVKDRFTISRDDSKNSLYLQMNSLKTEDTAVYYCARHGNFGNSYVSWFAYWG QGTLVTVSS

Table 6. Amino acid sequence of the anti-CD3 light chain variable domain (amino acid sequences of CDR1, CDR2 and CDR3 of the light chain variable domains are in bold and underlined)

anti-POPs VL domain sequence SEQ ID NO:68CD3-01 DIQMTOSPSLSLSASVGDRVTITCRSSTGAVTSNYANWVOQKPGKAPKALIGGTNKRAP GVPSRFSGSLIGDKATLTISSLQPEDFATYYCALWYSNLWVFGQGTKVEIK SEQ ID NO:69 CD3-02 DIQMTOSPSSLASVGDRVTITCRSSTGAVTSNYANWVOQKPGKAPKGLIGGTNKRAP GVPARFSGSGSGTDFTLTISSLOPEDFATYYCALWYSNLWVFGOGTKVEIK SEQ ID NO:70CD3-03 DIOMTQSPSSLSASVGDRVTITCRSSTGAVTTSNYANWVQQKPGKAPKGLIGGTNKRAP GVPSRFSGSLIGDKATLTISSLQPEDFATYYCALWYSNLWVFGQGTKVEIK SEQIDNO:71CD3-04 DIQMTOSPSLSLSASVGDRVTITCRSSTGAVTSNYANWVOQKPGKAPKGLIGGTNKRAP GVPSRFSGSLIGTDFTLTISSLOPEDFATYYCALWYSNLWVFGOGTKVEIK

In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain containing the STYAMN CDR1 sequence (SEQ ID NO: 72). In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain containing the CD112 sequence RIRSKYNNYATYYADSVKD (SEQ ID NO: 73). In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain containing the CDR3 sequence HGNFGNSYVSWFAY (SEQ ID NO: 74). In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain containing the CDR3 sequence HGNFGNSYVSYFAY (SEQ ID NO: 75). In yet other embodiments, the CD3 binding site has a heavy chain variable domain comprising the sequence of CDR1, CDR2, and CDR3 shown in SEQ ID NOs: 72-74, respectively. In still other embodiments, the CD3 binding site has a heavy chain variable domain comprising the sequence of CDR1, CDR2, and CDR3 shown in SEQ ID NOs: 72, 73, and 75, respectively.

In other embodiments, the CD3 binding region of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain containing a CDR1 sequence selected from the group consisting of NTYAMN (SEQ ID NO: 76), NTYAMH (SEQ ID NO: 77), and NKYAMN ( SEQ ID NO: 78). In other embodiments, the CD3 binding region of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain containing a CDR2 sequence selected from the group consisting of RIRNKYNNYATYYADSVKD (SEQ ID

- 13 040019

NO: 79), RIRNKYNNYATEYADSVKD (SEQ ID NO: 80), RIRSKYNNYATEYAASVKD (SEQ ID NO: 81), RIRNKYNNYATEYAASVKD (SEQ ID NO: 82), RIRSKYNNYATYYADSVKG (SEQ ID NO: 83), and RIRSKYNNYATEYADSVKS (SEQ ID NO: 84). In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a heavy chain variable domain comprising a CDR3 sequence selected from the group consisting of HGNFGDSYVSWFAY (SEQ ID NO: 85), HGNFGNTYVSWFAY (SEQ ID NO: 86), HGNFGCSYVSWFAY (SEQ ID NO: 87), HGNFGNSYISYWAY (SEQ ID NO: 88) and HGNFGNSYVSFFAY (SEQ ID NO: 89).

In still other embodiments, the CD3 binding site has a heavy chain variable domain comprising the sequence of CDR1, CDR2, and CDR3 as set forth in SEQ ID NOs: 76, 73, and 74, respectively, SEQ ID NOs: 76, 79, and 74, respectively, SEQ ID NOs. : 76, 80 and 74, respectively, SEQ ID NOs: 76, 81, and 74, respectively, SEQ ID NOs: 76, 82, and 74, respectively, SEQ ID NOs: 76, 83, and 74, respectively, SEQ ID NOs: 72, 83, and 74 respectively, SEQ ID NOs: 72, 83, and 85, respectively, SEQ ID NOs: 76, 83, and 86, respectively, SEQ ID NOs: 77, 83, and 74, respectively, SEQ ID NOs: 72, 83, and 87, respectively, SEQ ID NOs: 78, 73 and 88 respectively or SEQ ID NOs: 78, 84 and 89 respectively.

In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a light chain variable domain containing the CDR1 sequence RSSTGAVTTSNYAN (SEQ ID NO: 90). In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a light chain variable domain containing the GTNKRAP CDR2 sequence (SEQ ID NO: 91). In other embodiments, the CD3 binding site of the bispecific CD33 and CD3 binding protein has a light chain variable domain containing the CDR3 sequence of ALWYSNL (SEQ ID NO: 92). In yet other embodiments, the CD3 binding site has a light chain variable domain comprising the sequence of CDR1, CD2, and CD3 shown in SEQ ID NOs: 90-92, respectively.

In certain instances, the CD3 binding site has a high affinity for CD3. Alternatively, in other cases, the CDR1, CDR2, CDR3 from the heavy chain domain as well as the light chain domain or optionally the light chain variable domains and the heavy chain variable domains are derived from other anti-CD3 antibodies such as, for example, UCHT1, muromonab-CD3 (OKT3), teloxizumab (TRX4), teplizumab (MGA031), visilizumab (Nuvion), etc.

In another aspect, this document describes CD33 binding proteins, as well as bispecific CD33 and CD3 binding proteins, that are humanized or fully human, that is, of human origin.

In some embodiments, the bispecific CD33 and CD3 binding protein has one of the following combinations providing specificity for CD33 and CD3 via the light and heavy chain variable domains for CD33 and CD3, including but not limited to: (i) SEQ ID NO : 2, 12, 65 and 69, (ii) SEQ ID NOs: 3, 13, 65 and 69, (iii) SEQ ID NOs: 4, 14, 65 and 69, (iv) SEQ ID NOs: 5, 15, 65 and 69, (v) SEQ ID NOs: 1, 11, 64 and 68, (vi) SEQ ID NOs: 2, 12, 64 and 68, (vii) SEQ ID NOs: 2, 12, 66 and 70, ( viii) SEQ ID NOs: 4, 14, 66 and 70, (ix) SEQ ID NOs: 5, 15, 66 and 70, and (x) SEQ ID NOs: 3, 13, 64 and 68, (xi) SEQ ID NOs: 3, 13, 67 and 71, (xii) SEQ ID NOs: 4, 14, 64 and 68, (xiii) SEQ ID NOs: 5, 15, 64 and 68, (xiv) SEQ ID NOs: 7, 17 , 64 and 68, (xv) SEQ ID NOs: 6, 16, 64 and 68, (xvi) SEQ ID NOs: 6, 16, 67 and 71, (xvii) SEQ ID NOs: 8, 18, 64 and 68, (xviii) SEQ ID NOs: 9, 19, 64 and 68; (xix) SEQ ID NOs: 9, 19, 67 and 71, and (xx) SEQ ID NOs: 10, 20, 64 and 68.

Conservative variants of CDR sequences and heavy and light chain domains.

In alternative embodiments, the implementation of the heavy and light chain domains include immunologically active homologues or variants of the CDR sequences described here. Thus, in some embodiments, the sequence of a CDR in a heavy or light chain domain that binds to CD33 or CD3 is similar, but not identical, to the amino acid sequence shown in SEQ ID NOs: 21-63 or 72-92. In certain instances, a variant CDR sequence has 99, 98, 97, 96, 95, 94, 93, 92, 91, 90, 89, 88, 87, 86, 85, 84, 83, 82, 81, or 80% sequence identity compared with the sequence presented in SEQ ID NO: 21-63 or 72-90, and which is immunologically active.

In other cases, the variant CDR sequence includes 1, 2, 3, 4, or 5 conservative amino acid substitutions. Conservative substitutions include amino acid substitutions that replace a given amino acid with another amino acid with similar properties and further include, among aliphatic amino acids, substitution of alanine, valine, leucine and isoleucine; substitution of hydroxyl residues of serine and threonine, substitution of acidic residues of aspartate and glutamate, substitution between amide residues of asparagine and glutamine, substitution of basic residues of lysine and arginine, and substitutions from among the aromatic residues of phenylalanine and tyrosine.

In still other instances, the variant CDR sequence includes substitutions that enhance properties of the CDR, such as increased stability, protease resistance, and/or CD33 or CD3 binding affinities.

In other cases, the variant CDR sequence is modified to replace non-critical residues or residues in non-critical regions. Amino acids that are non-critical can be identified using known methods such as affinity maturation,

- 14 040019 CDR walking, site-directed mutagenesis, crystallization, nuclear magnetic resonance, photoaffinity labeling or alanine-scanning mutagenesis.

In other alternative embodiments, the CD33 and CD3 binding proteins comprise heavy and light chain domains that are immunologically active homologues or sequence variants of the heavy and light chain domains presented herein. Thus, in some embodiments, the CD33 and CD3 binding protein comprises a heavy or light chain domain sequence that is similar, but not identical, to the amino acid sequence shown in SEQ ID NOs: 1-20 or 64-71. In certain instances, a variant heavy or light chain sequence has 99%, 98%, 97%, 96%, 95%, 94%, 93%, 92%, 91%, 90%, 89%, 88%, 87%, 86%, 85%, 84%, 83%, 82%, 81%, or 80% sequence identity compared to the sequence presented in SEQ ID NO: 1-20 or 64-71, and which is immunologically active.

In other instances, the variant heavy or light chain domain sequence includes 1, 2, 3, 4, or 5 conservative amino acid substitutions. Conservative substitutions include amino acid substitutions that replace a given amino acid with another amino acid with similar properties and further include, among the aliphatic amino acids, substitution of alanine, valine, leucine and isoleucine; substitution of hydroxyl residues of serine and threonine, substitution of acidic residues of aspartate and glutamine, substitution between amide residues of asparagine and glutamine, substitution of basic residues of lysine and arginine, and substitutions from among the aromatic residues of phenylalanine and tyrosine.

In still other instances, the variant light or heavy chain domain sequence includes substitutions that enhance properties of the CDR, such as increased stability, protease resistance, and/or CD33 or CD3 binding affinities.

In other cases, the variant light chain domain sequence is modified to change non-critical residues or residues in non-critical regions. Amino acids that are non-critical can be identified using known methods such as affinity maturation, CDR walking, site-directed mutagenesis, crystallization, nuclear magnetic resonance, photoaffinity labeling, or alanine scanning mutagenesis.

CD33- and CD3-bispecific and tandem diabodies.

In another aspect, the CD33 binding protein or bispecific CD33 and CD3 binding protein is a dimer, ie it contains two polypeptides with antigen binding sites for CD33 and CD3.

Also, in another aspect, the document provides a dimeric and bispecific CD33 and CD3 binding protein in a tandem diabody (TandAb®) format. Such tandem diabodies are constructed by linking four antibody variable binding domains (two heavy chain variable domains (VH) and two light chain variable domains (VL)) into one gene construct (FIG. 1), allowing homodimerization. In such tandem diabodies, the linker is of such a length that it prevents intramolecular pairing of the variable domains, so that the molecule cannot reverse fold to form a single stranded diabody, but rather has to pair with the complementary domains of the other strand. The domains are also organized such that the respective VH and VL domains pair during this dimerization. Upon expression from a single gene construct, two identical polypeptide chains fold head-to-tail to form a functional non-covalent homodimer of approximately 105 kDa (FIG. 1). Despite the absence of intermolecular covalent bonds, the homodimer is highly stable after formation, remains intact and does not convert back to the monomeric form.

Tandem diabodies have a number of properties that provide advantages over traditional monoclonal antibodies and other smaller bispecific molecules. The tandem diabodies contain only the variable domains of the antibody and therefore are expected to have no side effects or non-specific interactions that may be associated with the Fc fragment. For example, Fc receptors that can bind to Fc domains are found on numerous cell types, such as white blood cells (eg, basophils, B cells, eosinophils, natural killer cells, neutrophils, etc.) or cells Kupfer. Since tandem diabodies allow bivalent binding to each of CD33 and CD3, the avidity matches that of IgG. The size of the tandem diabody at approximately 105 kDa is smaller than the size of the IgG, which may allow enhanced penetration into the tumor. However, this size is well above the renal clearance threshold for first pass clearance, providing a pharmacokinetic advantage over smaller bispecific formats based on non-antibody antibody binding domains or scaffolds. In addition, tandem diabodies are preferred over other bispecific binding proteins such as BiTE or DART molecules based on these pharmacokinetic and avid properties, resulting in longer intrinsic half-lives and rapid cytotoxicity. Tandem diabodies are well expressed in host cells, such as mammalian CHO cells. It is assumed that a reliable upstream and downstream manufacturing process is available for tandem diabodies.

- 15 040019

The CD33 and CD3 bispecific tandem diabodies described herein are designed to allow specific targeting of CD33 ⁺ tumor cells by recruiting cytotoxic T cells. This improves ADCC (antibody dependent cellular cytotoxicity) compared to full-length antibodies directed to a single antigen and not capable of direct recruitment of cytotoxic T cells. In contrast, by capturing CD3 molecules expressed specifically on these cells, the tandem diabody can cross-link cytotoxic T cells with CD33 ⁺ tumor cells in a highly specific manner, thereby greatly increasing the cytotoxic potential of such molecules. This mechanism is shown in Fig. 2. Tandem diabodies exhibit strong, specific and effective ADCC. It is reported that T cells may be involved in the control of tumor growth. For example, the presence of cytotoxic T cells in colorectal tumors as well as lymph nodes from patients with non-Hodgkin's lymphoma (NHL) has been shown to correlate with better clinical outcome. In addition, the potential of therapies designed to induce T cell responses has been demonstrated for melanoma vaccines as well as antibodies directed against CTLA-4, a negative regulator of T cell activation. The tandem diabodies described here capture cytotoxic T cells by binding to surface-expressed CD3, which forms part of the T cell receptor. Co-binding of this tandem diabody to CD3 and CD33 expressed on the surface of specific tumor cells induces T cell activation and mediates subsequent tumor cell lysis (FIG. 2).

Thus, in another aspect, a multispecific tandem diabody is provided. In some embodiments, the multispecific tandem diabody has specificities for two, three or more different epitopes, whereby the two or more epitopes can be of the same target antigen or different target antigens. In certain embodiments, the multispecific tandem diabody is bispecific and tetravalent, i.e. it contains four antigen binding sites. Such a bispecific tandem diabody binds via at least one antigen binding site to human CD3 and human CD33, wherein in certain cases the tandem diabody binds via two antigen binding sites to human CD3 and via two other antigen binding sites to human CD33, i.e. tandem The diabody binds bivalently to each antigen.

In some embodiments, the bispecific antigen-binding tandem diabody is specific for human CD33 and human CD3, wherein said tandem diabody comprises a first polypeptide and a second polypeptide, each polypeptide having at least four sequentially linked chain variable domains, wherein each polypeptide contains:

(i) a heavy chain variable domain (VH) specific for human CD33;

(ii) a light chain variable domain (VL) specific for human CD33;

(iii) a VH domain specific for CD3; and (iv) a VL domain specific for CD3.

In specific embodiments, the bispecific tandem diabody specifically binds to an epitope of human CD33 that is within ₆₂ DQEVQEETQ ₇₀ (SEQ ID NO: 94) (amino acid residues 62-70 of SEQ ID NO: 93) of human CD33. In specific instances, such a tandem diabody comprises a first polypeptide and a second polypeptide, each polypeptide having at least four sequentially linked chain variable domains, each polypeptide comprising:

(i) a heavy chain variable domain specific for an epitope of human CD33 that is within ₆₂ DQEVQEETQ ₇₀ (SEQ ID NO: 94) (amino acid residues 62-70 of SEQ ID NO: 93) of human CD33;

(ii) a light chain variable domain specific for a human CD33 epitope that is within ₆₂ DQEVQEETQ ₇₀ (SEQ ID NO: 94) (amino acid residues 62-70 of SEQ ID NO: 93) of human CD33;

(iii) a heavy chain variable domain specific for human CD3; and (iv) a light chain variable domain specific for human CD3.

In other embodiments, CD33/CD3 tandem diabodies are described herein that have an affinity for CD33 on CD33 ⁺ cells with a K _D dissociation constant of 10 nM or less, 5 nM or less, 1 nM or less, or 0, 5 nM or less. CD33 ⁺ cells can be selected from tumor cells such as, for example, HL-60 or KG-1.

In other embodiments, the CD33/CD3 tandem diabody described herein binds to CD3 and, in certain cases, the CD3 epsilon chain on CD3 ⁺ cells, in particular T cells, with a KD dissociation constant of 10 nM or less, 5 nM, or less, or 2 nM or less.

In some embodiments, each bispecific tandem diabody polypeptide comprises one of the following combinations of the four variable chain domains: (i) SEQ ID NOs: 2, 12, 65 and 69, (ii) SEQ ID NOs: 3, 13, 65 and 69, ( iii) SEQ ID NOs: 4, 14, 65 and 69, (iv) SEQ ID NOs: 5, 15, 65 and 69, (v) SEQ

- 16 040019

ID NOs: 1, 11, 64 and 68, (vi) SEQ ID NOs: 2, 12, 64 and 68, (vii) SEQ ID NOs: 2, 12, 66 and 70, (viii) SEQ ID NOs: 4,

14, 66 and 70, (ix) SEQ ID NOs: 5, 15, 66 and 70, and (x) SEQ ID NOs: 3, 13, 64 and 68, (xi) SEQ ID NOs: 3, 13, 67 and 71, (xii) SEQ ID NOs: 4, 14, 64 and 68, (xiii) SEQ ID NOs: 5, 15, 64 and 68, (xiv) SEQ ID NOs: 7, 17, 64 and 68, (xv) SEQ

ID NOs: 6, 16, 64 and 68, (xvi) SEQ ID NOs: 6, 16, 67 and 71, (xvii) SEQ ID NOs: 8, 18, 64 and 68, (xviii) SEQ ID NOs:

9, 19, 64 and 68; (xix) SEQ ID NOs: 9, 19, 67 and 71, and (xx) SEQ ID NOs: 10, 20, 64 and 68.

The term dimer as used herein refers to a complex of two polypeptides. In certain embodiments, the implementation of the two polypeptides are non-covalently linked to each other, in particular, provided that there is no covalent bond between the two polypeptides. In certain instances, the two polypeptides are linked by covalent bonds, such as disulfide bonds, which are formed to help stabilize the dimer. In certain embodiments, the dimer is homodimeric, that is, it contains two identical polypeptides. The term polypeptide refers to a polymer of amino acid residues linked by amide bonds. In certain instances, the polypeptide is a single-stranded fusion protein that is unbranched. In a polypeptide, the variable domains of an antibody are sequentially linked to each other. In other cases, the polypeptide may have consecutive amino acid residues in addition to the N-terminal and/or C-terminal residues of the variable domain. For example, such contiguous amino acid residues may contain a Tag sequence, in some cases at the C terminus, which may be useful for purification and detection of the polypeptide.

In one aspect, each bispecific tandem diabody polypeptide comprises four variable domains, a CD3 binding protein variable light chain (VL) and a CD3 binding protein variable heavy chain (VH), and a CD33 binding protein variable light chain (VL) and a variable heavy chain (VH). In certain embodiments, the four variable domains are linked via peptidic linkers L1, L2, and L3, and in some cases are organized in an N-terminal to C-terminal direction as follows:

domain order:

(1) VL(CD3)-L1-VH(CD33)-L2-VL(CD33)-L3-VH(CD3); or (2) VH(CD3)-L1-VL(CD33)-L2-VH(CD33)-L3-VL(CD3); or (3) VL(CD33)-L1-VH(CD3)-L2-VL(CD3)-L3-VH(CD33); or (4) VH(CD33)-L1-VL(CD3)-L2-VH(CD3)-L3-VL(CD33).

According to published studies, the length of the linkers affects the flexibility of the antigen-binding tandem diabody. Accordingly, in some embodiments, the length of the peptide linkers L1, L2 and L3 is such that the domains of one polypeptide can form intermolecular bonds with the domains of another polypeptide to form a dimeric antigen-binding tandem diabody. In certain embodiments, such linkers are short, ie, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 amino acid residues long. Thus, in certain instances, linkers are composed of about 12 or fewer amino acid residues. In the case of 0 amino acid residues, the linker is a peptide bond. Such short linkers promote intermolecular dimerization of the two polypeptides by binding and forming the correct antigen-binding sites between the antibody light chain variable domains and the antibody heavy chain variable domains of the different polypeptides. Typically, shortening the linker to about 12 amino acid residues or less prevents intermolecular interaction between adjacent domains of the same polypeptide chain with each other. In some embodiments, these linkers consist of about 3-10, eg 4, 5, or 6 consecutive amino acid residues.

With regard to the amino acid composition of the linkers, choose peptides that do not interfere with the dimerization of the two polypeptides. For example, linkers containing glycine and serine residues typically provide protease resistance. The amino acid sequence of the linkers can be optimized, for example, by phage display techniques to improve antigen binding and yield of the antigen-binding polypeptide dimer. In some embodiments, examples of peptide linkers suitable for a tandem diabody are GGSGGS (SEQ ID NO: 95), GGSG (SEQ ID NO: 96), or GGSGG (SEQ ID NO: 97).

Non-limiting examples of tandem diabodies described herein are tandem diabodies having an anti-CD33 VL and VH domain, an anti-CD3 VL and VH domain, an order of domains, and a linker according to Table. 7.

- 17 040019

Table 7 Exemplary CD33/CD3 tandem diabodies (TandAbs)

Tand. diabody Anti-SOPZ domain Anti-SVZ domain Domain order Linker VL vh vh VL L1/L3 L2 01 SEQID NO:2 SEQID NO: 12 SEQID NO:65 SEQID NO:69 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 02 SEQID NO:3 SEQID NO: 13 SEQID NO:65 SEQID NO:69 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 03 SEQID NO:4 SEQID NO: 14 SEQID NO:65 SEQID NO:69 1 GGSGGS (SEO ID NO:95) GGSG (SEO ID NO: 96) 04 SEQID NO:5 SEQID NO: 15 SEQID NO:65 SEQID NO:69 1 GGSGGS (SEO ID NO:95) GGSG (SEO ID NO: 96) 05 SEQID NO:4 SEQID NO: 14 SEQID NO:65 SEQID NO:69 1 GGSGGS (SEOID NO:95) GGSGG (SEO ID NO: 97) 06 SEQID NO:5 SEQID NO: 15 SEQID NO:65 SEQID NO:69 1 GGSGGS (SEOID NO:95) GGSGG (SEO ID NO: 97) 07 SEQID NO:1 SEQID NO: 11 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSGGS (SEO ID NO: 95) 08 SEQID NO:2 SEQID NO: 12 SEQID NO:64 SEQID NO:68 3 GGSGGS (SEOID NO:95) GGSGGS (SEO ID NO: 95) 09 SEQID NO:2 SEQID NO: 12 SEQID NO:66 SEQID NO:70 1 GGSGGS (SEQ ID NO:95) GGSG (SEO ID NO: 96) 10 SEQID NO:4 SEQID NO: 14 SEQID NO:66 SEQID NO:70 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) eleven SEQID NO:5 SEQID NO: 15 SEQID NO:66 SEQID NO:70 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 12 SEQID NO:3 SEQID NO: 13 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 13 SEQID NO:3 SEQID NO: 13 SEQID NO:67 SEQID NO:71 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 14 SEQID NO:2 SEQID NO: 12 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 15 SEQID NO:4 SEQID NO: 14 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 16 SEQID NO:5 SEQID NO: 15 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 17 SEQID NO:7 SEQID NO: 17 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 18 SEQID NO:7 SEQID NO: 17 SEQID NO:64 SEQID NO:68 2 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 19 SEQID NO:6 SEQID NO: 16 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 20 SEQID NO:6 SEQID NO: 16 SEQID NO:67 SEQID NO:71 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 21 SEQID NO:8 SEQID NO: 18 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID GGSG (SEO ID NO: 96)

- 18 040019

NO:95) 22 SEQID NO:9 SEQID NO: 19 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 23 SEQID NO:9 SEQID NO: 19 SEQID NO:67 SEQID NO:71 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96) 24 SEQID NO: 10 SEQID NO:20 SEQID NO:64 SEQID NO:68 1 GGSGGS (SEOID NO:95) GGSG (SEO ID NO: 96)

In some embodiments, the tandem diabody is tandem diabody 01 (SEQ ID NO: 98), 02 (SEQ ID NO: 99), 03 (SEQ ID NO: 100), 04 (SEQ ID NO: 101), 05 (SEQ ID NO: 101). NO: 102), 06 (SEQ ID NO: 103), 07 (SEQ ID NO: 104), 08 (SEQ ID NO: 105), 09 (SEQ ID NO: 106), 10 (SEQ ID NO: 107), 11 (SEQ ID NO: 108), 12 (SEQ ID NO: 109), 13 (SEQ ID NO: 110), 14 (SEQ ID NO: 111), 15 (SEQ ID NO: 112), 16 (SEQ ID NO : 113), 17 (SEQ ID NO: 114), 18 (SEQ ID NO: 115), 19 (SEQ ID NO: 116), 20 (SEQ ID NO: 117), 21 (SEQ ID NO: 118), 22 (SEQ ID NO: 119), 23 (SEQ ID NO: 120), or 24 (SEQ ID NO: 121) as shown in FIG. 9B-9Y.

In some embodiments, a CD33 binding protein and a CD33/CD3 bispecific binding protein (e.g., a CD33/CD3 bispecific tandem diabody) described herein are produced by expression of polynucleotides encoding a tandem diabody polypeptide that is linked to another identical polypeptide to form an antigen-binding tandem diabody. . Thus, another aspect is a polynucleotide, eg DNA or RNA, encoding an antigen-binding tandem diabody polypeptide as described herein.

The polynucleotide is constructed using known methods, for example, by combining genes encoding at least four antibody variable domains separated by peptide linkers or, in other embodiments, directly linked by a peptide bond, into a single genetic construct, operably linked to a suitable promoter and optionally , a suitable transcription terminator, and its expression in a bacterium or other suitable expression system such as, for example, CHO cells. Depending on the vector system and host used, any number of suitable transcription and translation elements can be used, including constitutive and inducible promoters. The promoter is chosen to direct the expression of the polynucleotide in the appropriate host cell.

In some embodiments, the polynucleotide is inserted into a vector, preferably an expression vector, which is the following embodiment. This recombinant vector can be constructed according to known methods.

A variety of vector/host expression systems can be used to incorporate and express a polynucleotide encoding a polypeptide of the disclosed antigen-binding tandem diabody. Examples of expression vectors for expression in E. coli are pSKK (Le Gall et al., J Immunol Methods. (2004) 285(1):111-27) or pcDNA5 (Invitrogen) for expression in mammalian cells.

Thus, in some embodiments, the antigen-binding tandem diabody described herein is prepared by introducing a vector encoding the polypeptide described above into a host cell and culturing said host cell under conditions under which the polypeptide chains are expressed, can be isolated and optionally in subsequently cleared.

In other aspects, the CD33 binding protein or CD33/CD3 bispecific binding protein (eg, CD33/CD3 tandem diabody) described herein has a modification. Representative modifications include, but are not limited to, acetylation, acylation, ADP-ribosylation, amidation, flavin covalent attachment, heme group covalent attachment, nucleotide or nucleotide derivative covalent attachment, lipid or lipid derivative covalent attachment, phosphatidylinositol covalent attachment, drug conjugation, transverse crosslinking, cyclization, disulfide bond formation, demethylation, covalent crosslinking, cystine formation, pyroglutamate formation, formylation, γ-carboxylation, glycosylation, GPI anchoring, hydroxylation, iodination, methylation, myristoylation, oxidation, proteolytic processing, phosphorylation, prenylation, racemization, selenoylation, sulfation, transfer RNA mediated addition of amino acids to proteins such as arginylation, and ubiquitination. In other embodiments, the CD33 binding protein or CD33/CD3 binding protein is modified with additional amino acids, such as a leader or secretory sequence, or a polypeptide purification sequence.

In other aspects, provided herein are pharmaceutical compositions comprising a CD33 binding protein, an antigen-binding tandem diabody, a vector containing a polynucleotide encoding an antigen-binding tandem diabody polypeptide, or a host cell trans

- 19 040019 formed by this vector, and at least one pharmaceutically acceptable carrier. The term pharmaceutically acceptable carrier includes, but is not limited to, any carrier that does not interfere with the effectiveness of the biological activity of the ingredients and is non-toxic to the patient to whom it is administered. Examples of suitable pharmaceutical carriers are well known in the art and include phosphate buffered saline solutions, water, emulsions such as oil/water emulsions, various types of wetting agents, sterile solutions, and the like. Such carriers may be formulated by conventional means and may be administered to the subject at a suitable dose. Preferably the compositions are sterile. These compositions may also contain adjuvants such as a preservative, emulsifying agents and dispersing agents. Prevention of the action of microorganisms can be ensured by the inclusion of various antibacterial and antifungal agents.

Bispecific CD33/CD3-binding proteins with high binding affinity for CD33 and CD3 are highly active in a large number of primary AML samples, suggesting that these molecules may exhibit activity against human AML across the entire cytogenetic/molecular spectrum of diseases, even in cases minimal expression of CD33. It should be noted that drug-specific cytotoxicity is also observed in the presence of residual autologous T cells and increases significantly with the addition of controlled amounts of T cells from healthy donors (see example 6).

Bispecific CD33/CD3 binding proteins, in particular tandem diabodies, can induce strong cytolysis of CD33 ⁺ leukemic cells in vitro. The data indicate that high affinity binding to CD33 and CD3 maximizes bispecific protein-induced T cell activation and anti-AML efficacy. High affinity CD33/CD3-targeted bispecific binding proteins, such as the tandem diabodies described here, exhibit cytolytic activity in primary AML in vitro. Thus, these bispecific binding proteins and tandem diabodies are suitable for a therapeutic approach in the treatment of acute myeloid leukemia (AML) or other hematological malignancies, for example, myeloid dysplastic syndrome (MDS) or myeloproliferative disease (MPD).

Thus, provided herein are methods in which the antigen-binding tandem diabody described herein is administered at an effective dose to a subject, e.g., a patient, for the treatment of CD33 ⁺ cancer (e.g., acute myeloid leukemia (AML)), disease, or condition. CD33+ cancers include, but are not limited to, acute leukemia such as acute myeloid leukemia, acute lymphoblastic leukemia (ALL), including progenitor B cell lymphoblastic leukemia, myeloid sarcoma, multiple myeloma, acute lymphoma such as acute lymphoblastic lymphoma, chronic myelomonocytic leukemia, and the like. CD33 ⁺ diseases and conditions include immunosuppressive conditions or environments mediated by myeloid-derived suppressor cells (MDSC) in certain types of cancer and chronic inflammation.

In some embodiments, the antigen-binding tandem diabody described herein is administered for the treatment of acute myeloid leukemia (AML). In certain embodiments, the antigen-binding tandem diabody described herein is administered to treat a subtype of acute myeloid leukemia.

The Franco-American-British classification system subdivides AML into eight subtypes: AML-M0 (minimal differentiation), AML-M1 (no cell maturation), AML-M2 (with granulocyte maturation), AML-M3 (promyelocytic or acute promyelocytic leukemia), AML-4 (acute myelomonocytic leukemia), AML-M5 (acute monoblastic or monocytic leukemia), AML-M6 (acute erythroid leukemia), and AML-M7 (acute megakaryoblastic leukemia). In certain instances, the antigen-binding tandem diabody described herein is administered to treat AML-M0, AML-M1, AML-M2, AML-M3, AML-M4, AML-M5, AML-M6, or AML-M7.

The classification system for AML according to the World Health Organization (WHO) divides AML into the following subtypes: AML with recurrent genetic abnormalities; AML with changes associated with myelodysplasia; myeloid neoplasms associated with therapy; myeloid sarcoma, myeloid proliferations associated with Down's syndrome, neoplasm of blast plasmacytoid dendritic cells; and AML not otherwise classified. In certain other cases, the antigen-binding tandem diabody described herein is administered to treat AML with recurrent genetic abnormalities; AML with changes associated with myelodysplasia; myeloid neoplasms associated with therapy; myeloid sarcoma; myeloid proliferation associated with Down syndrome; neoplasms from blast plasmacytoid dendritic cells; and AML not otherwise classified.

In some other embodiments, the antigen-binding tandem diabody described herein is administered to treat newly diagnosed, relapsed, or refractory AML.

In other embodiments, the antigen-binding tandem diabody described herein is administered to treat pre-leukemic blood disorders such as myeloid dysplastic syndrome (MDS) or myeloproliferative disease (MPD). In certain cases, antigen-binding

- 20 040019 the tandem diabody described here is administered for the treatment of MDS. In certain instances, the antigen-binding tandem diabody described herein is administered to treat MPD.

In other embodiments, the antigen-binding tandem diabody described herein is administered to treat multiple myeloma. In other embodiments, the antigen-binding tandem diabody described herein is administered for the treatment of chronic myelomonocytic leukemia (CMML).

In other embodiments, the antigen-binding tandem diabody described herein is administered to inhibit or kill myeloid-derived suppressor cells (MDSC). Myeloid-derived suppressor cells (MDSC) highly express CD33 in certain diseased isolated tissues and have strong immunosuppressive activities. In certain types of human cancer (CD33+ as well as non-CD33 ⁺ ), MDSCs proliferate and become activated to suppress tumor-specific CD4 ⁺ T cell responses and induce Treg cells, allowing the tumor or cancer to grow rapidly in the microenvironment. In chronic inflammation, MDSCs are reported to proliferate and be found at sites of inflammation to suppress T cell immune function. In other embodiments, the antigen-binding tandem diabody described herein is administered to treat a condition associated with MDSCs. In yet other embodiments, the antigen-binding tandem diabody described herein is administered to treat immunosuppression. In still other embodiments, the antigen-binding tandem diabody described herein is administered to treat inflammation suppressed by MDSC. In yet other embodiments, the antigen-binding tandem diabody described herein is administered to treat a reduced immune response caused by MDSC. In yet other embodiments, the antigen-binding tandem diabody described herein is administered to treat angiogenesis, tumor invasion, or cancer metastases stimulated by MDSC. In yet other embodiments, the antigen-binding tandem diabody described herein is administered to treat a cancer or tumor that is enhanced, dilated, aggravated, or enlarged by MDSCs.

The antigen-binding tandem diabody described herein is contemplated for use as a medicament. Administration is achieved in various ways, for example by intravenous, intraperitoneal, subcutaneous, intramuscular, topical or intradermal administration. In some embodiments, the route of administration depends on the type of therapy and the type of compound contained in the pharmaceutical composition. The dosage regimen will be determined by the attending physician and other clinical factors. Doses for any patient will depend on many factors, including patient size, body surface area, age, sex, particular compound to be administered, time and route of administration, type of therapy, general health, and other concurrently administered drugs. As used herein, an effective dose refers to the amount of active ingredient that is sufficient to affect the course and severity of a disease, causing a reduction or remission of such a disease. An effective dose suitable for the treatment and/or prevention of AML can be determined using known methods.

In other embodiments, the antigen-binding tandem diabody described herein is administered in combination with standard therapy for CD33 ⁺ cancers, diseases, or conditions. Standard therapies include chemotherapy, immunotherapy, hormone therapy, radiation, surgery, gene therapy, and the like. In certain instances, the antigen-binding tandem diabody described herein is administered in combination with standard AML therapy. In certain instances, the antigen-binding tandem diabody described herein is administered in combination with cytarabine, azacitidine, decitabine, anthracycline (e.g., daunorubicin, idarubicin, doxorubicin, and the like), amsacrine, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast or epipodophyllotoxin (eg, etoposide, teniposide, etc.), anthracenedione (eg, mitoxantrone, pixantrone, losoxantrone, pyroxantrone, amethantron, etc.), an anti -CD20 (eg, rituximab, ocrelizumab, ofatumumab) or a combination thereof. In certain instances, the antigen-binding tandem diabody described herein is administered in combination with cytarabine (apa-C). In certain instances, the antigen-binding tandem diabody described herein is administered in combination with azacitidine. In certain instances, the antigen-binding tandem diabody described herein is administered in combination with decitabine. In other instances, the antigen-binding tandem diabody described herein is administered in combination with an anthracycline (eg, daunorubicin, idarubicin, doxorubicin, and the like). In other instances, the antigen-binding tandem diabody described herein is administered in combination with checkpoint inhibitors (eg, PD-1 inhibitor, CTLA-4 inhibitor, etc.). In yet other instances, the antigen-binding tandem diabody described herein is administered in combination with an epipodophyllotoxin (eg, etoposide, teniposide, etc.). In other instances, the antigen-binding tandem diabody described herein is administered in combination with an anthracenedione (eg, mitoxantrone, pixantrone, losoxantrone, pyroxantrone, amethantron, etc.).

The examples below further illustrate the described embodiments without limiting the scope of the invention.

- 21 040019

Example 1

Cloning of expression DNA constructs encoding single-stranded Fv antibodies.

For bacterial expression of anti-CD33 single chain Fv (scFv) antibodies in E. coli, the DNA coding sequences of all molecules were cloned into a bacterial expression vector. Expression constructs were designed, all of which contained coding sequences for an N-terminal signal peptide and a C-terminal hexahistidine (6χHis) tag (SEQ ID NO: 122) to promote periplasmic antibody secretion and purification, respectively. The amino acid sequences of the VL and VH domains from all anti-CD33 scFv clones are shown in Table 1. 3 and 4.

Expression of recombinant anti-CD33 scFv antibodies in E. coli.

Recombinant scFv antibodies were expressed as soluble secreted proteins in the periplasm of E. coli. In the first step, small culture medium supplemented with ampicillin was inoculated with transformed bacteria and incubated for 16 hours at 28°C. The absorbance was then adjusted by adding a second medium supplemented with ampicillin and incubated again at 28° C. until an absorbance value in the range of 0.6-0.8 was reached at a wavelength of 600 nm. Protein expression was induced by adding 50 μM IPTG and incubating cultures at 21-28°C and 200 rpm for 16 h. After incubation, cells were pelleted by centrifugation (30 min, 4°C, 7500 rpm) and stored at -20 °C before further processing.

Purification of anti-CD33 single chain Fv antibodies.

Recombinant scFv was extracted from E. coli periplasm after centrifugation of bacterial cell cultures by resuspending the cell mass in buffer and incubating for 30 min at room temperature with gentle agitation. The cells were pelleted by centrifugation and the supernatants containing the recombinant proteins were kept. The procedure was repeated one more time before pooling the supernatants and homogenizing by sonication. The homogenates were diluted, supplemented with low concentrations of imidazole, and loaded onto a prepackaged Immobilized Metal Affinity Chromatography (IMAC) column (GE Healthcare). The column was washed until it was reset, and then the bound protein was eluted with imidazole buffer. Fractions containing antibody were pooled and then purified by size exclusion chromatography (SEC). Finally, protein eluates were concentrated by ultrafiltration and dialyzed against storage buffer. After treatment at low pH values (incubation at pH 3.0 for 2024 h at 37°C), the samples were neutralized using Tris. Purified proteins were stored in aliquots at -80°C until use.

Example 2

Cloning of expression DNA constructs encoding tandem diabodies (TandAb®).

For expression of the bispecific tandem diabodies in CHO cells, the coding sequences of all molecules were cloned into a mammalian expression vector system. The anti-CD33 scFv domains from Example 1 were used to construct CD33/CD3 tandem diabodies in combination with the anti-CD3 scFv domain, with the domains organized as shown in Table 1. 7 and in FIG. 3. Briefly, gene sequences encoding the VH and VL domains of antiCD33 separated by a peptide linker (VH linker-VL or VL linker-VH) were synthesized and subcloned. The resulting construct was cleaved to generate separate VH and VL coding sequences using a Bam HI restriction site located within the linker sequence. These VH and VL fragments were then ligated to a DNA fragment encoding the VH and VL domains of anti-CD3 (VH linker-VL or VL linker-VH) to obtain the final construct. Variants 1-3 of the order of CD33/CD3 tandem diabody domains are shown in FIG. 3. All expression constructs were designed to contain coding sequences for an N-terminal signal peptide and a C-terminal hexahistidine (6χHis) tag (SEQ ID NO: 122) to promote antibody secretion and purification, respectively.

Expression of tandem diabodies in stably transfected CHO cells.

An expression system based on CHO cells (Flp-In®, Life Technologies) derived from Chinese hamster ovary cells CHO-K1 (ATCC, CCL-61) was used (Kao and Puck, Proc. Natl. Acad Sci USA 1968;60(4 ):1275-81). Adherent cells were subcultured according to standard cell culture protocols from Life Technologies.

To adapt to growth in suspension, cells were detached from tissue culture flasks and introduced into medium without serum. The cells adapted to the suspension were cryopreserved in a medium containing 10% DMSO.

Recombinant CHO cell lines stably expressing secreted tandem diabodies were generated by transfection of suspension-adapted cells. During selection on hygromycin B antibiotic medium, viable cell density was measured twice a week, and cells were centrifuged and resuspended in fresh selection medium at a maximum density of 0.1 x 106 viable cells/ml. Pools of cells stably expressing tandem diabodies were extracted after 2-3 weeks of selection and transferred to a standard culture medium contained in

- 22 040019 shake flasks. Expression of recombinant secreted proteins was confirmed by protein gel electrophoresis or flow cytometry. Stable cell pools were cryopreserved in a medium containing DMSO.

Tandem diabodies were generated in 10 day fed cultures of stably transfected CHO cell lines by secretion into the cell culture supernatant. Cell culture supernatants were harvested after 10 days with culture viability typically >75%. Samples were collected from production cultures every other day and evaluated for cell density and viability. On the day of collection, cell culture supernatants were clarified by centrifugation and vacuum filtration before subsequent use.

Protein expression titers and product integrity in cell culture supernatants were analyzed by SDS-PAGE.

Cleaning of tandem diatels.

Tandem diabodies were purified from CHO cell culture supernatants using a two-step procedure. In a first step, His6-tagged (SEQ ID NO: 122) constructs were chromatographed on Ni-NTA Superflow followed by a second step size exclusion chromatography (SEC) on Superdex 200. The eluted tandem diabodies were characterized for homodimer (tandem antibody) content and pooled. if the homodimer content was 90% or higher. Finally, pooled samples were buffer exchanged and concentrated by ultrafiltration to a typical concentration of >1 mg/mL. Purity and homogeneity (typically >90%) of final samples were assessed by SDS PAGE under reducing and non-reducing conditions followed by immunoblotting with an anti-His-Tag antibody and by analytical SEC, respectively. Purified proteins were stored in aliquots at -80°C until use.

The CD33/CD3 tandem diabody polypeptides are shown in Table 1. 7 and in FIG. 3. Each tandem diabody consists of two identical polypeptides (FIG. 1). Both outer linkers L1 and L3 consist of six amino acids GGSGGS (SEQ ID NO: 95), while the central peptide linker 2 varies in length (4-6 amino acids) with the sequences, respectively, GGSG (SEQ ID NO: 96), GGSGG ( SEQ ID NO: 97) or GGSGGS (SEQ ID NO: 95).

Using a series of anti-CD33 variable domains and anti-CD3 variable domains, a large number of tandem diabody molecules have been generated that can be stably produced in transfected cell lines and that are stable at body temperature as well as after repeated freeze/thaw cycles. To facilitate further development and preclinical toxicological studies, attention has been focused on the selection of tandem diabody molecules that have shown binding to both human CD33 and cynomolgus CD33. Full amino acid sequence examples are shown for single chain tandem diabodies 12 (SEQ ID NO: 109), 14 (SEQ ID NO: 111) and 16 (SEQ ID NO: 113) in FIG. 9M, 9O and 9Q respectively. In this example, the order of variable domains and their linkers for structures is: VL(CD3)-L1-VH(CD33)-L2-VL(CD33)-L3-VH(CD3).

Example 3

Determination of antibody affinity by flow cytometry.

Cells were incubated with 100 μl serial dilutions of CD33/CD3 tandem diabodies. After washing three times with FACS buffer, the cells were incubated with 0.1 ml of mouse monoclonal anti-His antibody at a concentration of 10 μg/ml in the same buffer for 45 minutes on ice. After the second wash cycle, the cells were incubated with 0.1 ml of FITC-conjugated goat anti-mouse IgG antibodies at a concentration of 15 μg/ml under the same conditions as before. As a control, cells were incubated with anti-His IgG followed by FITC-conjugated goat anti-mouse IgG antibodies without anti-CD33 tandem diabodies. The cells were then washed again and resuspended in 0.2 ml FACS buffer containing 2 μg/ml propidium iodide (PI) to exclude dead cells. Fluorescence of 1x10 ⁴ live cells was measured using a Beckman-Coulter FC500 MPL flow cytometer using MXP software (Beckman-Coulter, Krefeld, Germany) or a Millipore Guava EasyCyte flow cytometer using Incyte software (Merck Millipore, Schwalbach, Germany) . Mean fluorescence intensities of cell samples were calculated using CXP software (Beckman-Coulter, Krefeld, Germany) or Incyte software (Merck Millipore, Schwalbach, Germany). After subtracting the fluorescence intensity values of cells stained with secondary and tertiary reagents only, the obtained values were used to calculate KD values using the single-site binding equation (hyperbola equation) GraphPad Prism (version 6.00 for Windows, GraphPad Software, La Jolla California USA) ).

Tandem diabodies were tested for their binding affinity to human CD3+ and CD33+ cells, and CD3+ and CD33 ⁺ cells of cynomolgus monkeys. Illustrative binding data for selected tandem diabodies are summarized in Table 1. 8.

- 23 040019

Table 8. Characteristics of CD3 and CD33 binding of SVZZ/SVZ-tandem diabodies

TandAb K _D on T cells [nM] K _D at HL60 [nM] K _D per KG-1 [nM] K _D on U937 [nM] K _D ratio of cynomolgus CD33/human POPs EC ₅₀ on HL-60 [nM] 01 94.2 0.6 0.9 7.1 0.7 1.9 02 69.8 0.2 0.3 0.9 1.1 0.5 03 81.9 1.1 1.8 8.9 0.6 3.6 04 79.3 0.5 0.5 1.7 1.1 1.8 05 69.5 1.0 1.2 6.2 0.8 2.7 06 86.3 0.4 0.5 1.6 0.8 1.6 07 49.7 13.7 47.9 47.1 45.8 17.8 08 2.4 0.3 0.5 1.8 0.6 1.8 09 2.4 0.5 0.3 2.2 1.0 6.8 10 1.9 0.5 1.0 1.7 0.8 7.0 eleven 2.6 0.3 0.5 0.6 1.2 5.9 12 1.5 0.3 0.9 0.5 1.7 1.3 13 55.7 0.2 0.3 0.5 1.6 1.1 14 2.1 0.3 0.3 1.2 1.0 1.6 15 1.3 0.4 0.3 0.9 1.1 1.8 16 2.1 0.3 0.2 0.3 1.4 1.5 17 3.3 5.0 52.5 24.4 1.9 18.4 18 1.9 3.4 16.3 15.1 3.1 6.3 19 6.3 2.8 3.6 5.4 37.3 5.7 20 143.8 4.1 7.0 7.2 33.8 10.0 21 2.1 9.7 25.1 80.2 0.9 7.6 22 4.1 0.7 2.0 8.6 0.6 3.2 23 97.2 0.4 1.0 5.1 1.9 2.8 24 2.3 5.6 12.4 39.5 1.8 9.6

#The K _D ratio of Cynomolgus CD33/Human CD33 was calculated based on K _D values measured on CHO cells expressing Cynomolgus CD33 and human CD33, respectively. { Human CD3/human CD3 Kd ratio was calculated based on the K _D values measured on Jurkat (human CD3) cells and the average _Kp value of three human CD3 ⁺ -tumor cell lines (HL-60, KG-1, U937).

CD3 binding affinity and cross-reactivity was assessed in titration and flow cytometry experiments on SVZ ⁺ Jurkat cells (obtained from Dr. Moldenhauer, DKFZ Heidelberg; human acute T-cell leukemia) and cynomolgus monkey CD3 ⁺ HSC-F cell line (JCRB, part number: JCRB1164). CD33 binding and cross-reactivity was assessed on SOP3 ⁺ human tumor cell lines: HL-60 (DSMZ, catalog number: ACC3, human B-cell progenitor leukemia), U-937 (DSMZ, catalog number: ACC5 ; human histiocytic lymphoma) and KG-1 (DSMZ, catalog number: ACC14; acute myeloid leukemia). The cross-reactivity K _D ratio was calculated using K _D values determined on CHO cell lines expressing recombinant human or cynomolgus monkey antigens.

The tandem diabodies showed a relatively high affinity for human CD33 ⁺ on most tumor cell lines tested, which was less than 1 nM. The affinity for human CD3 was 2 nM or less.

Example 4

Cytotoxicity assay.

For cytotoxicity analysis, target cells cultured under standard conditions are collected

-24040019 were washed and resuspended in Diluent C contained in the PKH67 Green Fluorescent Cell Linker Mini Kit to a density of 2x107 cells/mL. Then the cell suspension was mixed with an equal volume of PKH67-labeled solution at a double concentration and incubated for 2-5 min at room temperature. The staining reaction was performed by adding an equal volume of FCS and incubating for 1 min. After washing the labeled target cells with complete RPMI medium, the cells were counted and resuspended to a density of 2x105 cells/ml in complete RPMI medium. Then, 2x104 target cells were plated with enriched human T cells as effector cells in a 5:1 E:T ratio in the presence of increasing concentrations of these tandem diabodies in individual wells of a microtiter plate in a total volume of 200 μl/well. Spontaneous cell death and destruction of targets by T cells in the absence of antibodies was determined at least in triplicate on each plate. After centrifugation, the assay plates were incubated for the indicated times at 37° C. in a humidified atmosphere containing 5% CO2. After incubation, the cultures were washed once with FACS buffer and then resuspended in 150 μl of FACS buffer supplemented with 2 μg/ml PI. The absolute number of live target cells was assessed by positive PKH67 green staining and negative propidium iodide (PI) staining using a FC500 MPL flow cytometer from Beckman-Coulter (Beckman-Coulter) or a Guava EasyCyte flow cytometer from Millipore (Merck Millipore). Based on the measured remaining live target cells, the percentage of specific cell lysis was calculated by the following formula: [1-(number of live targets ( _sample )/number of live targets ( _{spontaneous release} ))] x 100%.

Sigmoidal dose-response curves and EC ₅₀ values were analyzed using non-linear regression/4-parameter logistic regression using GraphPad software. Lysis values obtained for a given antibody concentration were used to construct sigmoidal dose-response curves using 4-parameter logistic regression using Prism software.

EC ₅₀ values were determined in a 20-24 hour assay on CD33+ U-937 target cells (DSMZ, catalog number: ACC5; human histiocytic lymphoma) with enriched human T cells as effector cells at a ratio of 5:1. Some tandem diabodies were also tested in cytotoxicity assays on CD33+ KG-1 (DSMZ, catalog number: ACC14; acute myeloid leukemia) and HL-60 target cells. In particular, HL-60 cells were chosen as an AML model with a relatively high expression of CD33 on the cell surface (mean fluorescence intensity (MFI) in arbitrary units [mean±SEM]: 3133±215; n=3), and KG-1a was chosen as a model of AML with very limited expression of CD33 (MFI in arbitrary units: 277±11; n=3). Illustrative data on cytotoxicity for selected tandem diabodies are summarized in table. 9. Additional data on cytotoxicity for HL-60 cell lines are presented in table. 8, last column.

Table 9. In vitro potency of CD33/CD3 tandem antibodies on various CD33+ tumor cell lines

tandem diabody ECso [pM (pg/ml)] on human CD33 ⁺ target cell lines HL-60 U-937 KG-1 average 12 1.3 (137) 0.8 (84) 1.2 (126) 1.1 (116) 14 1.6 (168) 3.6 (378) 2.6 (273) 2.6 (273) 16 1.5 (158) 1.9 (200) 1.8 (189) 1.7 (179)

_EC50 values were determined in FACS-based cytotoxicity assays with primary human T cells as effector cells at an E:T ratio of 5:1 on the indicated target cell lines incubated for 20-24 h. Each tandem diabody was tested on each tumor cell. cell lines in at least two independent experiments. Average values are shown.

Example 5

Additional screening experiments for cytotoxicity in human CD33+ AML cell lines at 48 h.

As described above, significant cytotoxicity was detected as early as 24 hours, however higher levels of toxicity could be detected after 48 hours. A 48 hour time point was chosen for subsequent assays. The effect of T cell selection on tandem diabody-induced cytotoxicity was tested. To do this, unstimulated PBMCs were obtained from healthy volunteer donors and CD3 ⁺ cells were isolated by simple positive enrichment using CD3 MicroBeads, as well as by more complex negative selection using a MicroBead Cocktail antibody cocktail against CD14, CD15, CD16, CD19, CD34, CD36 , CD56, CD123 and CD235a. As shown in FIG. 4, tandem diabody-induced cytotoxicity was higher with negatively selected T cells from healthy donors compared to positively selected T cells. However, the T cell selection method did not affect the relative

- 25 040019 cytotoxic activities of individual tandem diabodies. Therefore, subsequent analyzes were performed using positively enriched T cells from healthy donors.

Unstimulated mononuclear cells were collected from healthy adult volunteers by leukapheresis at the Hematopoietic Cell Processing Core (Core Center of Excellence) at the Fred Hutchinson Cancer Research Center (FHCRC) according to study protocols approved by the FHCRC Institutional Review Board. The cells were enriched by magnetic cell sorting with CD3 MicroBeads (positive enrichment) or with the Pan T-Cell Isolation Kit (negative selection; both from Miltenyi Biotec, Auburn, CA) and then frozen in aliquots and stored in liquid nitrogen. Thawed cell aliquots were labeled with 3 μM CellVue Burgundy (eBioscience, San Diego, CA) according to the manufacturer's instructions. Purified PBMCs were cultured in the presence of various concentrations of tandem diabody molecules.

To quantify drug-induced cytotoxicity, cells were incubated at 37° C. (in 5% CO2 and air) as in Example 4 at various E:T cell ratios. After 24-72 hours, cell numbers and drug-induced cytotoxicity using DAPI to detect non-viable cells were determined using an LSRII cytometer (BD Biosciences) and analyzed using the FlowJo software. AML cells were identified by their forward/side scatter properties and, in experiments in which T cells from healthy donors were added, by the negativity of CellVue Burgundy staining (FIG. 5). Drug-induced specific cytotoxicity is presented as follows:

% cytotoxicity = 100 x (1 - live target cells _treated /live target cells _control ).

The results of cytotoxicity analyzes were presented as means ± standard error of the mean (SEM). Spearman's nonparametric correlation was used to calculate correlations between characteristics of a continuous sample. All P-values are two-sided. Statistical analyzes were performed using GraphPad Prism software.

In the absence of T cells from healthy donors, none of the CD33/CD tandem diabodies showed any appreciable cytotoxic effect on AML cell lines in the absence of T cells, supporting the mandatory requirement for T cells with respect to their cytotoxic effects (data not shown). In the presence of T cells, the degree of specific cytotoxicity induced by the tandem diabodies depended on the concentration of the tandem diabody as well as the cellular E:T ratio. Direct comparative studies of CD33/CD3-targeted tandem diabody molecules and one control tandem diabody (00) showed significant differences in antibody-induced cytotoxicity in HL-60 cells (Fig. 6A/B and Table 10) and KG-1a cells (Fig. 6C/D and Table 10), with results having a high level of reproducibility in repeated experiments. Overall, the degree of cytotoxicity induced by tandem diabodies correlated with binding affinity for CD3 on primary human T cells (for cytotoxicity in KG-1a cells at 25 pM (approximately 2.5 ng/mL) and E:T=5:1: r=-0.542, p=0.009; for cytotoxicity in HL-60 cells at 25 pM and E:T=5:1: r=-0.391, p=0.07). Tandem diabodies 12, 14, 16 were highly cytotoxic to both HL-60 and KG-1a cells.

- 26 040019

Table 10. Induction and cytotoxicity of CD25 and CD69 after 48 hours of CD33/CD3 tandem diabodies

Tand. diabody ¹ CD3 Co (nM) human cells CD33 Co (nM) HL-60 cells CD25 Induction ECzo (pM) ² CD69 Induction ECzo(pM) ² T-cell proliferation in PBMC ECzo (pM) ³ Cytotoxicity, HL-60 cells (% ± SEM) ⁴ Cytotoxicity, KG-la cells (% ± SEM) ⁴ 15 1.3 0.4 6 7 7 82.9±3.7 80.2±1.9 12 1.5 0.3 6 3 2 84.7±2.3 85.6±1.6 10 1.9 0.5 10 6 6 48.0±2.4 78.6±2.3 14 2.1 0.3 10 7 6 86.0±0.4 69.8±5.7 21 2.1 9.7 ND 225 500 12.4 ± 1.0 0.0±0.2 24 2.3 5.6 ND 57 264 24.5 ± 1.9 1.1±0.2 09 2.4 0.5 eleven 7 9 43.2 ± 15.8 74.6±3.2 eleven 2.6 0.3 eleven 5 6 52.7±8.1 84.7±1.4 17 3.3 5.0 thirty 114 thirty 4.2±0.2 0.7±0.4 22 4.1 0.7 10 4 7 74.2±7.4 44.4±5.3 16 5.1 0.3 1 2 3 86.0 ± 1.4 81.3±1.5 19 6.3 2.8 9 5 6 79.4±3.5 83.8±2.9 07 49.7 13.7 134 65 50 6.3±3.3 2.1±0.7 13 55.7 0.2 thirty 22 23 70.4±2.5 1.3±0.4 05 69.5 1 116 74 74 23.8±6.9 0.3±0.3 02 69.8 0.2 42 27 4 80.9±3.6 4.6±2.1 04 79.3 0.5 94 62 44 24.1±4.0 0.7±0.8 03 81.9 1.1 117 87 63 13.1±3.6 0.0±0.5 06 86.3 0.4 39 21 48 45.7±6.4 1.4±0.2 01 94.2 0.6 92 91 89 8.0 ± 1.6 0.4±0.4 23 97.2 0.4 41 17 37 73.7±2.6 1.5±0.3 20 143.8 4.1 98 75 38 31.2±3.9 1.1±0.3

'Tandem diabodies (TandAbs) are listed in order of increasing affinity for CD3. ² CD25 and CD69 induction was measured after 24 hours in unfractionated PBMC cultures.

³ T cell proliferation induced by CD33/CD3 tandem diabodies in unfractionated PBMCs with CD33+ cells present.

⁴ Cytotoxicity (%) at 48 h of DAPI+ cells at 25 pM tandem diabodies in the presence of T cells from healthy donors at an E:T ratio of 5:1, based on the results of 3 independent experiments performed in double wells.

ND: No detectable CD25 activation.

Example 6

Additional characterization of tandem diabodies in human primary AML samples.

To compare the cytotoxic properties of these candidates, samples from AML patients were obtained for testing from the FHCRC sample repository.

Frozen aliquots of Ficoll-isolated mononuclear cells from pretreated (diagnostic) peripheral blood or bone marrow samples from adult AML patients were obtained from the FHCRC specimen repositories. We used the 2008 World Health Organization (WHO) AML criteria (Vardiman et al.; Blood. 2009; 114(5):937-951) and the updated United Kingdom Medical Research Council (MRC) criteria for cytogenetic risk assessment (Grimwalde et al., Blood 2010;116(3):354-365). Patients gave written informed consent to the collection and use of their biological samples for research purposes according to protocols approved by the FHCRC Institutional Review Board. Clinical data was de-identified in accordance with the requirements of the Health Insurance Portability and Accountability Act. After thawing, cells were directly stained with antibodies recognizing CD33 (clone P67.6; PE-Cy7-conjugated), CD3 (clone SK7; PerCP-conjugated), CD34 (clone 8G12; APC-conjugated; all from BD Biosciences, San Jose, CA) And

- 27 040019

CD45 (clone HI30; APC-eP1ior® 780-conjugated; eBioscience). To identify non-viable cells, the samples were stained with 4',6-diamidino-2-phenylindole (DAPI). At least 10,000 events were collected on a Canto II flow cytometer (BD Biosciences) and DAPI cells were analyzed using the FlowJo program (Tree Star, Ashland, OR).

After thawing, samples contained >58% AML blast cells by flow cytometry based on CO45/side scatter properties. The samples contained >50% viable cells immediately after thawing and >50% viable cells after 48 h in cytokine-containing liquid culture (Fig. 7). The mean age of patients was 58.1 years (range: 23.9-76.2 years); the risk of cytogenetic disease was favorable in 2 patients, moderate in 18 patients, and severe in 7 patients. Information on the mutational status of NPM1, FLT3, and SEBRA was incomplete; however, one sample was known to contain the cgBPA ^double - ^mutant _and the other sample was NPMI ^pos /FLT3-ITD ^neg . The average percentage of myeloid blasts and CD3 T cells in the studied samples was 86.1% (range: 58.4-97.0%) and 2.0% (range: 0-11.9%), respectively, and the average viability sample in culture at 48 h was 80.1% (range: 53.6-93.6%). At the time of sample collection, fifteen patients had newly diagnosed AML, while 12 had relapsed (n=7) or refractory (n=5) disease. As indicated in Table. 11, the main characteristics of samples from patients with newly diagnosed AML and patients with relapsed/refractory disease were similar in terms of CD33 expression on myeloid blasts, autologous T cell count, proportion of myeloid blasts, and culture viability.

The addition of tandem diabody molecules to cultures of AML samples resulted in moderate dose-dependent cytotoxicity (Figure 8A), indicating that autologous T cells contained in samples from patients with active AML can be captured for lysis of leukemic cells. In the presence of T cells from healthy donors, the cytotoxic activity of individual tandem diabodies was strongly dependent on drug dose and E:T cell ratio (Figure 8B/C). However, high activity of tandem diabodies was observed even in some samples with very low expression of CD33 on AML blasts. Of the tandem diabody molecules, 12 were the most active, as they had the highest cytotoxicity at low concentrations (2.5 pM (approximately 250 ng/mL) and, to a lesser extent, also 10 pM (approximately 1 ng/mL)) at both ratios E: T= 1:3 and E: T= 1:1.

POPs/POPs tandem diabodies were screened in representative blood samples from AML patients, which differed in age, patient gender, disease stage (newly diagnosed, relapsed, refractory), CD33 expression level, and cytogenetic risk (Table 11). Notably, a number of tandem diabodies tested (eg, 02, 08, 09, 11, 12, 14, 16, 19, 22, and 23) were highly active in almost all samples from patients with various diseases shown in FIG. 15. In addition, the degree and extent of activity are similar across all stages of AML, including newly diagnosed, relapsed, and refractory patients.

Table 11 Characteristics of primary AML samples

All patients (n=27) Newly diagnosed AML(n=15) Recurrent/refractory AML (n=12) Average age (range), years 58.1 (23.9-76.2) 64.0 (40.2-76.2) 44.4 (23.9-67.4) Cytogenetic/molecular risk Favorable 2 2 — Average 18 10 8 ^ppp^double-mutant 1 1 — NPMX^FLTS-ITiy^ 1 — 1 NPMl^/FLTl-ITD ¹ ™ or NPMl ^neg /FLT3I _TD p°s 10 5 5 Serious 7 3 4 Sample Source Bone marrow eleven 4 7 peripheral blood 16 eleven 5 Mean % of blasts (range) 86.1 (58.4-97.0) 86.1 (66.7-95.5) 86.7 (58.4-97.0) Average expression of CD33 on blasts (range) 849 (5-5356) 849 (5-5356) 788(7-2242) Mean % T cells (range) 2.0 (0-11.9) 1.6 (0-11.9) 2.1 (0.7-8.7) Average % viability at 48 hours (range) 80.1 (53.6-93.6) 76.0 (53.6-93.6) 83.5 (63.9-93.1)

Example 7

The potency and efficacy of POPs/POPs tandem diabody 12 and tandem diabody 16 on different CD33 cell lines of different origins expressing at different levels of CD33.

To assess the dependence of potency and efficacy of POPs/POPs tandem diabodies on CD33 density on target cells, various human POPs ⁺ tumor cell lines and CHO cells expressing recombinant human CD33 were tested for their CD33 expression levels using the QIFIKIT kit. quantification kit and anti-CD33 mAb WM53. The results in the table. 12 show that CD33 densities on tumor cell lines range between ~1300 SABC (standardized antibody binding capacity) and ~46000 SABC. Expression on CHO-CD33 cells was ~197,000 SABC, significantly higher than on tumor cell lines. All CD33 ⁺ cell lines tested were used as target cells in at least 3 independent FACS-based cytotoxicity assays with human T cells as effector cells at a 5:1 effector to target ratio in the presence of serial dilutions of CD33/CD3 tandem diabody 12 and tandem diabody 16. In each assay, EC50 and tandem diabody-mediated lysis were calculated using non-linear regression. The results show that neither potency (EC50 values) nor potency (% lysis) of 12 and 16 correlate with CD33 density on the target cell surface.

Notably, at least 12 and 16 show their cytotoxic activity also against SEM-like cells with very low CD33 densities below 1500 SABC.

Table 12. CD33 expression on target cell surfaces and cytotoxic activity of CD33/CD3 tandem diabody 12 and tandem diabody 16

cell line Density CD33 [SABC] 12 16 EU ₅₀ [pM] EU ₅₀ [pM] average SD average SD average SD sleepCD33 196990 28053 11.8 AND 2 24.0 19.5 HL-60 45948 4478 1.4 0.5 1.6 0.4 KG-1 42828 6923 1.0 0.6 1.9 2.0 KASUMI1 25922 6484 1.3 0.6 2.4 1.4 TNR-1 22065 415 1.9 0.2 6.0 1.2 RPMI- 8226 19931 2604 14.0 17.8 2.8 2.0 U-937 17669 4593 0.9 0.1 1.3 0.6 K562 13789 2156 4.5 1.3 4.8 2.7 BV-173 8518 1231 1.4 0.6 3.2 1.6 SEM 1306 144.2 2.2 0.5 5.1 3.0

Standardized antibody-binding capacity (SABC) on CD33 ⁺ cell lines was determined using QIFIKIT and anti-CD33 mAb WM53. _EC50 values for tandem diabody 12 and tandem diabody 16 redirecting target cell lysis were determined in FACS-based cytotoxicity assays with human primary T cells as effector cells at E:T ratios of 5:1 and incubation for 20-24 h; samples with CD33-expressing CHO cells were incubated for 40-48 hours. Means and standard deviation (SD) from at least 3 independent experiments are shown.

Example 8

TandAb activation of T cells and in vitro killing of AML cells.

TandAbs were incubated with purified human T cells and VPD-450-labeled CD33+ human leukemia cell line KG-1 or human CD33 ^- β-ALL cell line G2 (E:T 5:1). Flow cytometry was used to evaluate target cell lysis of TandAbs (10 ^-15 to 10 ^-8 M; 24 h, 37°C).

Incubation of TandAbs 12, 16, and 19 with human T cells resulted in effective lysis of KG-1 cells (IC50 ~0.01, 0.5, and 5 pM, respectively). There was an activation of up to 40% of T cells (CD25+), which increased simultaneously with cytotoxic activity. The control TandAb with an irrelevant target, 00 (>10 ^-7 M), did not cause significant killing of KG-1 in vitro. 16 alone induced lysis of KG-1 cells (IC50 = 5 x 10 ^-12 M), while at 1 x 10 ^-8 M there was no effect on CD33 ^- G2 cells. The results indicate that T cells become activated and efficiently lyse tumor cells when targeted to CD33 ⁺ leukemic cells (KG-1) and primary CD33+ AML blasts by CD33/CD3 TandAbs.

Example 9

Epitope mapping.

Tandem diabodies containing various CD33-binding fragments were subjected to epitope mapping using CLIPS Technology (Pepscan) to identify CD33-binding epitopes.

CLIPS technology facilitates the structuring of peptides into single loops, double loops, triple loops, sheeting, coiling, and combinations thereof, enabling continuous epitope mapping of the target molecule.

An array of over 7000 independent peptides was synthesized and the binding of each antibody to

- 29 040019 peptides were tested by ELISA.

Tandem diabodies 12, 14, 16 and 22 bind to region ₆₂ of DQEVQEETQ ₇₀ (SEQ ID NO: 94) in the first Ig-like domain of human CD33. The corresponding amino acid regions are not underlined and are shown in bold in FIG. 9A. Tandem diabodies 01, 02, 04, 06, 08, 09, 13 and 23 are also expected to bind to this epitope as they contain the same CD33 binding domains (SEQ ID NOS: 2 and 12, 3 and 13, 5 and 15 , 9 and 19) as tandem diabodies 12, 14 16 and 12.

Example 10

Dose-dependent effect in an in vivo prophylactic tumor model.

Tandem diabodies 12 and 16 were compared at various dose levels in a prophylactic HL-60 tumor xenograft model in NOD/scid mice reconstituted with human T cells. Three dose levels at 10, 1 and 0.1 μg (0.5, 0.05 and 0.005 mg/kg) were chosen to achieve a dose-dependent effect.

Eight experimental groups of immunodeficient NOD/scid mice were xenotransplanted by subcutaneous injection of a suspension of 4x106 HL-60 cells. Prior to injection, cells were mixed with 3x106 T cells isolated from buffy coats (healthy donors) using negative selection. To account for potential donor variability in T cells, each of the experimental groups was divided into three cohorts, each of which received T cells from only one individual donor. All animals of the experimental groups transplanted with tumor cells and T cells received an intravenous bolus on days 0, 1, 2, 3 and 4 (qdxd5) of the vehicle (control) or 16 or 12 at three different dose levels, as indicated (0.1, 1 and 10 µg). One group without treatment with effector cells and vehicle served as an additional control. In table. 13 summarizes the grouping and dosing regimen.

Table 13

Scheme Cell concentration / introductions Group Treatment Dose animal Cohort (iv) η 1 Carrier — 4x106 HL-60 4 4x10 ^b HL-60 + 3x10 ^b T-cells Cohort 1 Day 0, 1, 2,3,4 3 2 Carrier — 4x10 ^b HL-60 + 3x10 ^b T-cells Cohort 2 3 4x10 ⁶ HL-60 + 3x10 ⁶ T cells Cohort 3 3 4x10 ^b HL-60 + 3x10 ^b T-cells Cohort 1 Day 0, 1, 2,3,4 3 3 16 10 mcg 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 2 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 3 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 1 Day 0, 1, 2,3,4 3 4 16 1 mcg 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 2 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 3 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 1 Day 0, 1, 2,3,4 3 5 16 0.1 µg 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 2 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 3 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 1 Day 0, 1, 2,3,4 3 6 12 10 mcg 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 2 3 4x10 ⁶ HL-60 + 3x10 ⁶ T-cells Cohort 3 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 1 Day 0, 1, 2,3,4 3 7 12 1 mcg 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 2 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 3 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 1 Day 0, 1, 2,3,4 3 8 12 0.1 µg 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 2 3 4x10 ^b HL-60 + 3x10 ⁶ T-cells Cohort 3 3

Treatment groups for an in vivo dose-response study in the HL-60 xenograft model. All animals in the control groups reliably developed the tumor and showed uniform tumor growth. The presence of T cells did not affect the development of tumors. No difference in HL-60 growth was observed in the presence or absence of T cells in the vehicle control groups.

Treatment with both tandem diabodies tested showed a clear dose-dependent antitumor effect (FIG. 10). No significant difference was found between the two tandem diabodies. Plotting mean tumor volumes in FIG. 10 was limited to day 29, when most of the treatment groups were completed. The study continued until day 45 and the animals were observed for tumor-free survival. In the 10 µg or 1 µg 16 treated groups, 6 of 9 animals were tumor-free at the end of the observation period and 5 of 9 animals treated with 10 µg 12 were tumor-free at day 45. One animal that received 1 µg 12, remained tumor-free.

All animals in the control groups reliably developed the tumor and showed uniform tumor growth. Treatment with either of the tandem diabodies showed a dose-dependent antitumor effect and no significant difference between the two tandem diabodies until day 29.

Detectable differences were only present after long-term observation (day 45), which was discontinued for the low dose and control groups due to the growth of large tumors. In groups receiving 16, there were more animals that did not have a tumor.

Example 11.

Model of the developed tumor.

To prove the correctness of the concept, a xenograft model in mice was developed.

NOD/scid with pre-evolved HL-60 tumors using 16.

Briefly, immunodeficient female NOD/scid mice were sublethally irradiated (2 Gy) and subcutaneously inoculated with 4×10 ⁶ HL-60 cells. On day 9, animals received a single bolus injection of anti-asialo GM1 rabbit antibody (Wako, Neuss, Germany) to deplete murine natural killer (NK) cells. On day 10, when the tumor had reached a volume in the range of 50-150 mm ³ (mean 73±11 mm ³ ), the animals were divided into 3 treatment groups. Groups 2 and 3 (8 animals in each group) were intraperitoneally injected with 1.5 x 10 ⁷ activated human T cells. Before injection, T cells were isolated from buffy coats (healthy donors) using negative selection. T cells were expanded and activated with the T-Cell Activation/Expansion Kit according to the manufacturer's instructions (Miltenyi Biotech). To assess potential donor variability, groups 2 and 3 were further divided into two cohorts, each receiving expanded and activated T cells from an individual donor. Each cohort received T cells from only one individual T cell donor.

Table 14. Treatment Groups for the HL-60 Evolved Tumor Xenograft Model

Group Animals (η) Inoculum Day 0, subcutaneous (sc.) >bath cages Day 10, intraperitoneally (ip) Cohort Treatment Day 13 to 21, once a day 1 5 4x10 ^b HL-60 Carrier (iv) 2 4 4 4x10 ^b HL-60 4x10 ^b HL-60 1.5x10 ⁷ T cells (Donor 1) 1.5x10 ⁷ T cells (Donor 2) 1 2 Carrier (iv) 3 4 4 4x10 ^b HL-60 4x10 ^b HL-60 ^1.5x107 T cells (Donor 1) 1.5x107 T cells (Donor 2) 1 2 TandAb 16 (iv) 50 mcg

From day 13, animals in group 3 showed a mean tumor volume of 105 mm ³ and were treated with a total of 9 intravenous doses of 50 μg of tandem diabody 16 (qdx9d). In table. 14 shows the grouping and dosing regimen. Groups 1 and 2 received vehicle only. Body weight and tumor volume were measured up to day 27.

All animals reliably developed a tumor that was palpable on day 6. The mean tumor volume in animals of groups 1 and 2 (HL-60) that received vehicle increased continuously until the end of the study on day 27 (FIG. 11). In group 2 animals that received primary activated human T cells in addition to HL-60 tumor cells, the mean tumor volume increased faster compared to group 1 (HL-60 only).

Repeat intravenous treatment from days 13 to 21 (qdxd9) with tandem diabody 16 (50 μg/animal; 2.5 mg/kg) in the presence of human T cells (group 3) rapidly retarded tumor growth compared to group 1 and group 2. Tandem diabody 16 delayed tumor growth in group 3 by approximately 4-5 days compared to the vehicle-treated control group (group 2). Statistically significant differences between days 6 and 27 were found between group 2 (HL-60, T cells, vehicle) and group 3 (HL-60, T cells, 16) on day 22 (p<0.05), day 23 (p<0.01) and day 27 (p<0.01) (two-way ANOVA with repeated measures ANOVA with Bonferroni corrections applied post-test). No statistically significant difference was found between group 1 and group 2 due to the unusually slow tumor growth in group 1.

Donor variability in T cell activity was not observed when tumor progression was compared between cohort 1 and cohort 2 in the group that received T cells from different donors (see Table 14).

Example 10 shows that the xenograft model in NOD/scid mice with pre-evolved HL-60 tumor (AML) and intraperitoneally grafted human T cells developed successfully. Repeated dosing of tandem diabody 16 at the single dose level resulted in a statistically significant delay in tumor growth compared to the corresponding vehicle-treated control group. The data obtained are comparable to those published for a similar study using CD33/CD3 BiTE™ (Aigner et al., 2012; Leukemia, 2013, Apr;27(5):1107-15).

- 31 040019

Example 12.

Efficacy of CD33/CD3 tandem diabodies in a PDX AML model in NSG mice.

Cryopreserved cells from an AML patient whose CD33 ⁺ leukemia contained 2-4% CD3+ T cells were used to develop the PDX AML model in NSG mice. One hour after tumor cell injection, irradiated (250 cGy) NSG mice were injected with CD33/CD3 tandem diabodies 16 or 12 at one of two intravenous (iv) doses (50 μg or 5 μg; n=8 mice/group) as a bolus of 200 µl. Additional injections of tandem diabodies were performed every 4 days. Mice were weighed once a week and then sacrificed on day 38 to collect peripheral blood, bone marrow and spleen for analysis by flow cytometry (huCD33, huCD34, huCD45, muCD45, huCD14, huCD3, huCD4, huCD8 and 7AAD). The results are shown in FIG. 12.

In FIG. 12 shows that untreated mice had a significant number of human blasts in the bone marrow and spleen after 38 days. In contrast, mice treated with daily intravenous injections of tandem diabodies 12 and 16 showed significantly lower levels of human AML blasts in the bone marrow and spleen. A strong antiAML effect of CD33/CD3 tandem diabody was observed at both dose levels (5 and 50 µg/injection).

The observed anti-AML effect for both CD33/CD3 tandem diabodies 12 and 16 was much stronger than that of the CD123/CD3 DART® antibody targeting AML in an identical mouse model (Hussaini et al.: Targeting CD123 In Leukemic Stem Cells Using Dual Affinity Re-Targeting Molecules (DARTs®) November 15, 2013; Blood: 122 (21)). In contrast to CD33/CD3 tandem diabodies, which destroyed almost all AML blasts in the bone marrow and spleen, Hussaini et al. CD123/CD3 DART® only reduces AML blasts in bone marrow and spleen in the PDX model at 2.5 and 0.25 mg/kg, and the authors also report that CD123/CD3 DART® reduces the number of AML blasts in the bone marrow and spleen in the PDX model is only 40-78% at 0.5 mg/kg.

Example 13

Rapid onset of target cell lysis mediated by CD33/CD3 tandem diabody 16.

To evaluate the kinetics of target cell lysis mediated by the CD33/CD3 tandem diabody, cytotoxicity assays were performed by detecting calcein release at various incubation times. Calcein-labeled CD33 ⁺ target HL-60 cells were incubated with serial dilutions of tandem diabody 16 in the presence of primary human T cells as effector cells at an E:T ratio of 25:1 for 30 min, 1, 2, 3, 4 or 5 hours At each time point, the calcein that was released from the lysed target cells was used to calculate the EC ₅₀ value and target cell lysis mediated by tandem diabody 16 using a non-linear regression/sigmoidal dose-response curve. In FIG. 13 shows an unexpectedly rapid onset of target cell lysis mediated by the tandem diabody, with over 40% lysis after 30 minutes of incubation at saturating concentrations of the tandem diabody. After 4 hours of incubation, more than 90% lysis of target cells was achieved. In table. 15 and in FIG. 14 shows the EC ₅₀ values and specific lysis values determined for tandem diabody 16 at incubation times ranging from 30 minutes to 5 hours. In addition, the results demonstrate that under the analytical conditions used, the maximum force (lowest EC ₅₀ value) was reached after 2 hours of incubation and that after 5 hours of incubation almost all target cells were lysed. Overall, these results demonstrate a very fast, strong and efficient target cell lysis mediated by CD33/CD3 tandem diabodies.

Table 15. EC50 Kinetics and Lysis Values Determined for Tandem Diabody 16

Incubation time [min] EU ₅₀ [pM] Lysis mediated by tandem diabody [%] thirty 4.8 44.1 60 2.5 59.8 120 1.6 75.1 180 1.6 88.8 240 1.5 93.7 300 1.6 97.4

Example 14

Protocol for a pilot clinical study on the administration of CD33/CD3 tandem diabodies in patients with AML.

This phase I/II clinical trial is aimed at studying the CD33/CD3 tandem diabody 16 for use in the treatment of acute myeloid leukemia (AML).

Research results.

Primary. The maximum tolerated dose of CD33/CD3 tandem diabody is 16.

Secondary. Determining whether the in vitro response of CD33/CD3 tandem diabody 16 is associated with clinical

- 32 040019 reply.

Phase 1

The maximum tolerated dose (MTD) will be determined in the phase I section of the study.

1.1. The maximum tolerated dose (MTD) will be determined in the phase I section of the study.

1.2. Eligible patients will be included in the study using the CD33/CD3 tandem diabody 16.

1.3. The aim is to identify the highest dose of CD33/CD3 tandem diabody 16 that can be safely administered without severe or unmanageable side effects in participants. The dose administered will depend on the number of participants who were included in the study and how well the dose is tolerated. Not all participants will receive the same dose.

Phase II

2.1. The next phase II section will be performed in MTD to determine if therapy with CD33/CD3 tandem diabody 16 results in at least a 20% positive clinical response.

Primary outcome for phase II. Determining whether therapy with CD33/CD3 tandem diabody 16 results in a clinical response in at least 20% of patients (blast cell response, minor response, partial response, or complete response).

Selection.

AML documented by WHO criteria in peripheral blood and bone marrow, excluding patients with acute promyelocytic leukemia (APL).

Patients with AML who are refractory to primary induction chemotherapy, have recurrent disease, or are >60 years of age and are not eligible for standard cytotoxic therapy due to age, general condition, and/or adverse risk factors as judged by the treating physician.

Age > 18 years.

Karnofsky General Condition > 50% or ECOG General Condition 0-2.

Lifespan > 6 weeks.

Although some embodiments are described and shown herein, it will be apparent to those skilled in the art that such embodiments are presented by way of example only. Numerous variations, changes and substitutions will occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments described herein may be used in the implementation of these embodiments in practice. It is intended that the following claims define the scope of the invention and that methods and structures falling within the scope of those claims and their equivalents are hereby covered.

Claims (23)

CLAIM 1. A bispecific antigen-binding tandem diabody specific for human CD33 and human CD3, said tandem diabody comprising a first polypeptide and a second polypeptide, each polypeptide having at least four sequentially linked chain variable domains, each polypeptide containing : (i) a heavy chain variable domain (VH) specific for human CD33; (ii) a light chain variable domain (VL) specific for human CD33; (ii i) a VH domain specific for human CD3; and (iv) a VL domain specific for human CD3, wherein each polypeptide contains four chain variable domains selected from the group consisting of: (a) SEQ ID NOs: 3, 13, 64 and 68; (b) SEQ ID NOs: 2, 12, 64 and 68; (c) SEQ ID NOs: 4, 14, 64 and 68; (d) SEQ ID NOs: 5, 15, 64 and 68; (e) SEQ ID NOs: 6, 16, 64 and 68; or (f) SEQ ID NOs: 9, 19, 64, and 68, wherein in each polypeptide, the four variable chain domains are sequentially linked to each other via peptide linkers L1, L2, and L3 in the following order: VL(CD3)-L1-VH(CD33)-L2-VL(CD33)-L3-VH(CD3); VH(CD3)-L1-VL(CD33)-L2-VH(CD33)-L3-VL(CD3); VL(CD33)-L1-VH(CD3)-L2-VL(CD3)-L3-VH(CD33); or VH(CD33)-L1-VL(CD3)-L2-VH(CD3)-L3-VL(CD33). 2. A bispecific tandem diabody according to claim 1, characterized in that the linkers L1, L2 and L3 consist of about 12 amino acid residues or less. 3. The bispecific tandem diabody according to claim 1, wherein the L1, L2 and L3 linkers are each independently selected from GGSGGS (SEQ ID NO: 95), GGSG (SEQ ID NO: 96) or GGSGG (SEQ ID NO: 97) . - 33 040019 4. A bispecific tandem diabody according to claim 1, characterized in that the L1 and L3 linkers are GGSGGS (SEQ ID NO: 95) and the L2 linker is GGSG (SEQ ID NO: 96) or GGSGG (SEQ ID NO: 97). 5. A bispecific tandem diabody according to claim 1, characterized in that the four variable chain domains are sequentially linked to each other via peptide linkers L1, L2 and L3 in the order VL(CD3)-L1-VH(CD33)-L2-VL(CD33 )-L3-VH(CD3); and wherein linkers L1 and L3 are GGSGGS (SEQ ID NO: 95) and linker L2 is GGSG (SEQ ID NO: 96) or GGSGG (SEQ ID NO: 97). 6. A bispecific tandem diabody according to claim 1, wherein the tandem diabody is a tandem diabody 12 (SEQ ID NO: 109), 14 (SEQ ID NO: 111), 15 (SEQ ID NO: 112), 16 (SEQ ID NO: 113), 19 (SEQ ID NO: 116) or 22 (SEQ ID NO: 119). 7. The bispecific antigen-binding tandem diabody according to claim 1, wherein the tandem diabody has binding with a dissociation constant K _D of 10 nM or less to CD33 on CD33 ⁺ tumor cells selected from HL-60, KG- 1 and U-937. 8. The bispecific antigen-binding tandem diabody according to claim 1, wherein the tandem diabody specifically binds to a human CD33 epitope that is within ₆₂ DQEVQEETQ ₇₀ (SEQ ID NO: 94) (amino acid residues 62-70 of SEQ ID NO: 93) human CD33. 9. A composition for the treatment of CD33+ cancer, comprising a bispecific antigen-binding tandem diabody according to any one of claims 1 to 8 and a pharmaceutically acceptable carrier. 10. The composition of claim 9, wherein the CD33+ cancer is acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), progenitor B-cell lymphoblastic leukemia, myeloid sarcoma, multiple myeloma, acute lymphoma, acute lymphoblastic lymphoma, or chronic myelomonocytic leukemia (CMML), optionally CD33+ cancer being AML with recurrent genetic abnormalities, AML with changes associated with myelodysplasia, myeloid neoplasms associated with therapy; myeloid sarcoma; myeloid proliferations associated with Down syndrome; or neoplasm of blast plasmacytoid dendritic cells. 11. The composition of claim 9, wherein the AML is newly diagnosed, relapsed, or refractory. 12. Composition according to claim 9, additionally containing cytarabine, azacitidine, decitabine, anthracycline, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast, epipodophyllotoxin, anthracenedione, an anti -CD20 or combinations thereof. 13. A composition for the treatment of myeloid dysplastic syndrome (MDS) comprising a bispecific antigen-binding tandem diabody according to any one of claims 1 to 8 and a pharmaceutically acceptable carrier. 14. Composition according to claim 13, additionally containing cytarabine, azacitidine, decitabine, anthracycline, amsacrine, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast, epipodophyllotoxin, anthracenotoxin an anti-CD20 agent; or combinations thereof. 15. A composition for the treatment of immune suppression by myeloid-derived suppressor cells (MDSCs), comprising a bispecific antigen-binding tandem diabody according to any one of claims 1 to 8 and a pharmaceutically acceptable carrier. 16. The use of a bispecific tandem diabody according to any one of claims 1 to 8 for the treatment of acute myelogenous leukemia (AML). 17. Use according to claim 16, further comprising administering cytarabine, azacitidine, decitabine, anthracycline, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast, epipodophyllotoxin, anthracene toxin, agent anti-CD20 or combinations thereof. 18. Use according to claim 16, characterized in that the AML is an AML with recurrent genetic abnormalities; AML with changes associated with myelodysplasia; myeloid neoplasia associated with therapy; myeloid sarcoma; myeloid proliferations associated with Down syndrome; neoplasm of blast plasmacytoid dendritic cells; or AML not otherwise classified. 19. Use according to claim 16, wherein the AML is AML-MO, AML-M1, AML-M2, AML-M3, AML-M4, AML-M5, AML-M6 or AML-M7. 20. Use according to claim 16, characterized in that the AML is newly diagnosed, recurrent or refractory. 21. Use according to claim 16, further comprising administering cytarabine, azacitidine, decitabine, anthracycline, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast, epipodophyllotoxin, anthracenedione, agent anti-CD20 or combinations thereof. - 34 040019 22. The use of a bispecific tandem diabody according to any one of claims 1 to 8 for the treatment of CD33+ myeloid dysplastic syndrome (MDS). 23. Use according to claim 22, further comprising administering cytarabine, azacitidine, decitabine, anthracycline, amsacrine, fludarabine, clofarabine, cladribine, nelarabine, methotrexate, bortezomib, carfilzomib, melphalan, ibrutinib, thalidomide, lenalidomide, pomalidomide, apremilast, epipodophyllotoxin , an anti-CD20 agent, or combinations thereof.