Trispecific molecule combining specific tumor targeting and local immune checkpoint inhibition

Abstract

The present invention relates to a novel molecule comprising three binding sites with specificity for a tumor cell, for an effector cell and for a checkpoint molecule, respectively. Moreover, the present invention relates to a pharmaceutical composition comprising such a molecule and to uses of such a molecule.

Claims

1. A trispecific protein comprising: (i) a first domain, wherein said first domain is formed by a Fab domain that specifically binds to a cell surface molecule at the cell surface of a tumor cell, wherein said cell surface molecule at the cell surface of the tumor cell is CD33, wherein the first domain comprises light chain CDR1 having an amino acid sequence of RASESLDNYGIRFLT (SEQ ID NO: 29), light chain CDR2 having an amino acid sequence of AASNQGS (SEQ ID NO: 30), light chain CDR3 having an amino acid sequence of QQTKEVPWS (SEQ ID NO: 31), heavy chain CDR1 having an amino acid sequence of DSNIH (SEQ ID NO: 32), heavy chain CDR2 having an amino acid sequence of YIYPYNGGTDYNQKFKN (SEQ ID NO: 33), and heavy chain CDR3 having an amino acid sequence of GNPWLAY (SEQ ID NO: 34); (ii) a second domain, wherein said second domain is formed by an IgG Fc domain and specifically binds to a cell surface molecule at the cell surface of an immune cell/immune cells; and (iii) a third domain, wherein said third domain is formed by one or more SirpIg domains that specifically bind to a checkpoint molecule at the cell surface of said tumor cell, wherein said checkpoint molecule at the cell surface of said tumor cell is CD47, wherein said third domain is formed by an immunoglobulin-like domain of SIRPα (signal regulatory protein alpha), and wherein said trispecific protein does not comprise any other part of said immune cell checkpoint receptor besides said immunoglobulin-like domain of SIRPα; wherein two copies of the first domain and the second domain form an IgG and wherein the third domain is fused on its C-terminus to the N-terminus of each first domain by a flexible linker; wherein the flexible linker comprises two to eight tandem repeats, each tandem repeat consisting of four or five amino acids with the amino acid sequence GGGS (SEQ ID NO: 6 or GGGGS (SEQ ID NO: 14).

2. The trispecific protein according to claim 1, wherein said tumor cell is an acute myeloid leukemia (AML) cell.

3. The trispecific protein according to claim 1, wherein said trispecific protein further comprises a fourth domain, wherein said fourth domain specifically binds to a cell surface molecule at the cell surface of said tumor cell; and/or said trispecific protein further comprises a second copy of said third domain.

4. The trispecific protein, according to claim 1, wherein the affinity of said first domain for said cell surface molecule at the cell surface of said tumor cell is higher than the affinity of said third domain for said checkpoint molecule by at least a factor of 10, when the affinity is measured by flow cytometry.

5. The trispecific protein, according to claim 1, wherein the affinity of said first domain for said cell surface molecule at the cell surface of said tumor cell is within the range of from 1 to 50 nM; and the affinity of said third domain for said checkpoint molecule is within the range of from 500 nM to 3 μM, when the affinity is measured by flow cytometry.

6. The trispecific protein, according to claim 1, wherein the first domain comprises the sequence of SEQ ID NO: 4.

7. The trispecific protein, according to claim 1, wherein the first domain comprises the sequence of SEQ ID NO: 5.

8. The trispecific protein, according to claim 1, wherein the first domain comprises the sequences of SEQ ID NOs: 4 and 5.

9. The trispecific protein, according to claim 1, wherein the third domain comprises the sequence of SEQ ID NO: 1.

10. A pharmaceutical composition comprising the trispecific protein according to claim 1, and a pharmaceutically acceptable carrier.

Description

(1) In the following, reference is made to the figures:

(2) All methods mentioned in the figure descriptions below were carried out as described in detail in the examples.

(3) FIGS. 1A-1C show a schematic depiction of a conventional mAB and scFv constructs.

(4) (1A) Conventional IgG antibody. 1: Fab (fragment antigen binding); 2: Fc (fragment crystallizable); 3: Fv (fragment variable): variable domain of heavy and light chain; 4: antigen binding region.

(5) (1B) scFv (single chain fragment variable): Heavy and light chain are connected by a flexible linker.

(6) (1C Single chain bispecific molecule: Two different scFvs connected with flexible linker, resulting in a bispecific molecule comprising two scFv.

(7) FIGS. 2A-2B show a schematic overview of autonomous modules within an exemplary embodiment of the molecule according to the invention indicating (2A) the function of the different modules and (2B) examples for domains that may be used for these modules.

(8) Module 1 comprises the first binding site and mediates specific binding to the tumor cell to be eliminated through a cell surface molecule of the tumor cell. Module 1 can for example be an scFv that specifically hinds to a tumor antigen at the cell surface. This domain is connected by a flexible linker to module 2, which mediates effector cell recruitment and can for example be an scFv that specifically binds to an immune cell, such as a macrophage cell.

(9) Module 3 comprises a binding site that specifically binds to a checkpoint molecule on the immune cell or the tumor cell. For example, module 3 may be formed by the endogenous extracellular domain (EED) of an immune checkpoint receptor. For example, the EED of the immune checkpoint receptor SIRPα on immune cells binds specifically to its immune checkpoint ligand, the transmembrane protein CD47, on cells encountering the immune cell. Upon binding of module 1 to the tumor antigen and of module 2 to the effector cell (i.e. the immune cell), the effector cell and tumor cell are brought into close vicinity. Binding of module 3 to the checkpoint molecule prevents checkpoint signaling and thus activation of the effector cell will cause elimination of the tumor cell.

(10) (Note: According to the present invention, different orders of the different modules within the molecule according to the invention are considered, including, but not limited to, the exemplary orders shown in FIG. 2A.)

(11) FIGS. 3A-3C show a schematic illustration of checkpoint signaling and immune surveillance in cancer in the absence or presence of different forms of treatment.

(12) (3A) Checkpoint signaling in the absence of treatment. The tumor cell overexpresses CD47 at its cell surface. Due to binding of CD47 to Sirpα at the cell surface of the effector cell (immune cell), the effector cell receives an antiphagocytic signal and thus does not phagocytose the tumor cell.

(13) (3B) In certain conventional treatments, two antibodies are applied: A tumor-specific mAb binds to a tumor antigen at the surface of the tumor cell. The Fc region of the tumor-specific mAb binds to an Fcγ receptor at the cell surface of the immune cell. This recruits the immune cell to the tumor cell (and also sends, upon engagement of Fey receptor, an additional activation signal to the effector cell). At the same time, an anti-CD47 mAb is administered that binds to CD47 at the cell surface of the tumor cell. This prevents binding of CD47 to Sirpα at the cell surface of the immune cell and thus prevents the tumor cell from sending an antiphagocytic checkpoint signal to the immune cell. Thus, the effector cell destroys the tumor cell. However, the anti-CD47 mAbs bind not only to CD47 molecules on the tumor cell, but also to CD47 molecules on other cell types. Consequently, severe systemic side effects because of effector cell activity against such other cell types might be observed.

(14) (3C shows examples of treatment with molecules according to the present invention.

(15) Left: A molecule according to the present invention contains a module (e.g. a scFv) binding to a tumor antigen, a module (e.g. a scFv) binding to an Fey receptor on the effector cell, and a module (which may e.g. be formed by one, two (or n) Sirp-Ig domains) binding to CD47 on the tumor cell. Simultaneous binding of the molecule to the tumor antigen on the tumor cell and the Fcγ receptor on the effector cell recruits the effector cell to the tumor cell. The binding of the CD47-binding module to CD47 at the cell surface of the tumor cell prevents antiphagocytic checkpoint signaling to the immune cell. Thus, the effector cell kills the tumor cell. Since the Sirp-Ig domains are part of the same molecule as the tumor targeting and the effector cell recruitment/activation modules, inhibition of checkpoint signaling occurs locally. Accordingly, systemic side effects are reduced compared to the treatment in (B).

(16) Right: A molecule according to the present invention contains an antibody with binding specificity to a tumor marker on the tumor cell. The antibody is linked to two domains binding to CD47 on the tumor cell (in this case, each of the antibody “arms” is linked through its light chain to a Sirp-Ig). By binding to a tumor antigen on the cell surface of the tumor cell and to an Fcγ receptor on the effector cell, the effector cell is recruited to the tumor cell. At the same time, binding of the two Sirp-Ig domains to CD47 molecules at the cell surface of the tumor cell efficiently prevents binding of CD47 to the Sirpα receptor and thus prevents the tumor cell from sending an antiphagocytic checkpoint signal to the effector cell. As a consequence, the effector cell destroys the tumor cell. Again, the Sirp-Ig domains are part of the same molecule as the tumor targeting and the effector cell recruitment modules. Hence, inhibition of checkpoint signaling occurs only locally and systemic side effects are reduced compared to the treatment in (B).

(17) FIGS. 4A-4B show the construct design and an expression analysis of the liCADs prepared in Example 1.

(18) (4A) Schematical view of different constructs used in this study. The constructs consist of an N-terminal hexa-histidine tag (6×HIS), followed by different approaches to target CD47 (shown within the upper black box, the lower box displays control molecules). The central anti-CD16 as well as the anti-CD33 domain remain unchanged in all constructs.

(19) (4B) SDS-PAGE analysis of purified proteins (1) SIRPα-αCD16-αCD33, (2) SIRPα_CV1-αCD16-αCD33, (3) αCD47-αCD16-αCD33 and (4) αCD16-αCD33. (M) molecular weight marker.

(20) FIGS. 5A-5F show experimental data generated by FACS analysis, confirming binding of individual modules of liCADs to their antigens or ligands, respectively.

(21) (5A) HEK cells show binding of SIRPα_CV1, (5B) MOLM-13 cells show binding of anti-CD47 scFv and (5C) THP-1 cells show binding of anti-CD33 scFv. (SD) To demonstrate the binding of the anti-CD16 scFv CHO cells, stably transfected with CD16, were used. (SE) FACS analysis did not allow for the detection of SIRPα binding to CD47 on Jurkat cells.

(22) FIGS. 6A-6C show experimental data generated by FACS analysis to determine the dissociation constant (KD values) using highly over-expressing stable CHO cell lines.

(23) (6A) Top: FACS Analysis of SirpIg.CD16.CD33 and SirpIg.SirpIgCD16.CD33 binding to CHO cells highly over-expressing CD47 (CHO.exCD47). Bottom left: CHO.exCD47 cells were used to determine KD values of SirpIg.SirpIg.CD16.CD33. Bottom right: and KD values of SirpIg.CD16.CD33.

(24) (6B) Top: Binding of anti-CD33 within the SirpIg.CD16.CD33 molecule was analysed by FACS analysis using CD33 over-expressing CHO cells (CHO.exCD33). Bottom: KD-values were determined for anti-CD33 scFV using CHO.exCD33 cells.

(25) (6C) Top: Again binding of anti-CD16 to CD16 over-expressing CHO (CHO.exCD16) cells shown by FACS analysis. Bottom: KD values were determined for anti-CD16 scFv using CHO.exCD16 cells.

(26) FIGS. 7A-7C shows data obtained from a redirected lysis (RDL) assay testing the dose-dependent induction of redirected lysis of MOLM 13 and HEK 293T cells.

(27) (7A) LiCAD dependent cellular cytotoxicity of CD47+/CD33+ MOLM-13 target cells. Killing efficiency correlates with affinities for CD47.

(28) (7B) Calcein-AM labeled HEK CD47 single positive cells, mixed with unlabeled HEK CD47/CD33 double positive cells, were used as targets to compare efficacy of liCADs and control molecules at a constant effector:target ratio of 2:1 at maximal protein concentrations. In a parallel reaction unlabeled HEK CD47 single positive cells, mixed with Calcein-AM labeled HEK CD47/CD33 double positive cells were used as targets to compare efficacy of liCADs and control molecules at a constant effector:target ratio of 2:1 as well at maximal protein concentrations. % specific lysis was analysed and demonstrates the potential of liCADs to preferentially kill double positive target cells.

(29) (7C) Calcein-AM labeled HEK CD47 single positive cells, mixed with unlabeled HEK CD47/CD33 double positive cells, were used as targets to compare efficacy of liCADs and control molecules at a constant effector-target ratio of 2:1 at EC.sub.50 values. In a parallel reaction unlabeled HEK CD47 single positive cells, mixed with Calcein-AM labeled HEK CD47/CD33 double positive cells were used as targets to compare efficacy of liCADs and control molecules at a constant effector:target ratio of 2:1 as well at EC.sub.50 values. % specific lysis was analysed and demonstrates the potential of liCADs to preferentially kill double positive target cells.

(30) FIG. 8 shows the increase in phagocytosis of MOLM-13 target cells in a dose-response manner for liCADs and controls.

(31) FIG. 9 shows some liCAD formations (i.e. combinations of (a) a domain providing a first binding site capable of specifically binding to a cell surface molecule at the cell surface of said tumor cell, (b) a domain providing a second binding site capable of specifically binding to a cell surface molecule at the cell surface of said immune cell(s), and (c) a domain providing a third binding site capable of specifically binding to a checkpoint molecule) considered by the present application. Particularly preferred combinations are indicated by dashed lines.

(32) FIG. 10 shows a summary of experimental data for molecules according to the invention. Additional licCAD and licMAB molecules targeting CD33, interfering with CD47, and effector cell function, according to the invention were expressed and tested in different experiments. The summary includes monoclonal antibody (mAB) formats and Fc-engineered variants thereof (for control, high affinity variants), licMABs as described in Table 4 and Fc-engineered variants thereof, bispecific licMABs and liCADs as indicated in Table 4. Fc-engineered variants of bispecific licMABs are currently under development. The summary includes measurements for thermo stability (thermofluor assays, Tm), binding to the recombinantly expressed extracellular domain of CD16 (exCD16) analyzed by size exclusion chromatography (SEC), binding to MOLM-13 cells (K.sub.D determination), internalization of liCADs and licMABs, antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) assays at 100 nM (shown as % of phagocytosis). n.d.=not determined; “a” in the figure represents “anti” (as in antibody fragment, e.g. scFv).

(33) FIGS. 11A-11B show original data obtained from thermal stability assays. The thermal stability of the indicated molecules was determined by fluorescence thermal shift assays using the CFX96 Touch Real-lime PCR Detection System. 10 μg of protein containing 1× SYPRO Orange were measured using FAM and SYBR Green I filter pairs. All molecules show reasonable thermo stability in the performed assay.

(34) (11A) Thermal stability assays for αCD33 mAB and licMABs (SirpIg.αCD33 and SirpIg. SirpIg.αCD33; upper panel) and Fc-engineered variants thereof (lower panel).

(35) (11B) Thermal stability assays of liCADs (upper panel) and bispecific licMABs (lower panel)

(36) FIG. 12 exemplary shows data on binding of a SirpIg.SirpIg.αCD33 licMAB and a Fc-engineered variant thereof to the recombinantly expressed extracellular domain of CD16 (exCD16) analyzed by size exclusion chromatography (SEC). As summarized in FIG. 10 control mABs, licMABs and Fc-engineered variants thereof were tested for their binding ability to exCD16 by SEC. As indicated in Table 4 conventional antibodies have a rather low affinity to Fc receptors in comparison to Fc-engineered variants which have a much higher affinity. As exemplarily shown in this figure conventional Fc domains do not form a stable complex with exCD16 (upper panel) in contrast to Fc-engineered variants (lower panel) measured by SEC. Co-migration of exCD16 with the Fc-engineered molecules is validated by SDS-PAGE. Conclusively, conventional Fc domains of IgG1 molecules have low affinity to exCD16 and thus do not form a stable complex as shown by SEC. Fc-engineered mABs and licMABs however, can recruit effector cells or cells expressing Fc receptors with higher affinity.

(37) FIGS. 13A-13B show the K.sub.D determination as an avidity value of different molecules indicated in this invention. All molecules tested show binding to target antigen expressing cells in the nM range.

(38) (13A) Binding analysis of αCD33 mAB, SirpIg.CD33 and SirpIg.SirpIg.αCD33 licMABs to MOLM-13 cells measured by flow cytometry (upper panel). K.sub.D determination of Fc-engineered variants of αCD33 mAB and licMABs (lower panel). Mean values and SEM (error bars) are plotted and K.sub.D values are indicated.

(39) (13B) K.sub.D determination of liCADs (upper panel) and bispecific licMABs (lower panel). Mean values and SEM (error bars) are plotted and K.sub.D values are indicated.

(40) FIG. 14 shows data obtained from internalization assays. MOLM-13 cells were incubated with 15 μg/ml of licMABs or mAb either on ice-cold water for 2 h (to prevent internalization) or at 37° C. for 30, 60 or 120 min. Cells were then washed with ice-cold FACS buffer and antibodies remaining on the surface were detected by staining with FITC αHuman IgG Fc. To define the background fluorescence, MOLM-13 cells were directly stained with the secondary antibody. % of CD33 internalization is indicated (upper panel). Internalization was co-evaluated by confocal fluorescence microscopy (lower panel). Scale bar=10 μm. As expected, bivalent binding of CD33 causes internalization of mABs and licMABs. However, internalization can be omitted by monovalent molecules binding CD33 such as bispecific licMABs or liCADs targeting CD33. Thus, these monovalent binders provide a highly promising approach to target CD33 positive cells, while avoiding internalization, locally interfere with an immune checkpoint and recruit effector immune cell.

(41) FIGS. 15A-15B show data obtained from an antibody-dependent cellular cytotoxicity (ADCC) assay testing the dose-dependent induction of specific lysis of MOLM-13 cells. MOLM-13 cells were labeled with Calcein-AM as described for FIG. 7. Calcein-AM labeled MOLM-13 cells were incubated with NK cells (effector:target ratio 2:1) for 4 h at increasing protein concentration. Cytotoxic effects induced by the molecules were analyzed and plotted as a dose-response curve.

(42) (15A) Cytotoxic effects on MOLM-13 cells induced by SirpIg.αCD33 licMAB, SirpIg.SirpIg.αCD33 licMAB, αCD33 mAB, SirpIg.αCD19 licMAB and αCD19 mAB. αCD19 mAB and licMAB were used as a controls (upper panel). Cytotoxic effects on MOLM-13 cells induced by Fc-engineered variants of SirpIg.αCD33 licMAB, SirpIg.SirpIg.αCD33 licMAB and αCD33 mAB (lower panel).

(43) (15B) Cytotoxic effects on MOLM-13 cells induced by liCADs and a control triplebody (αCD47.αCD16.αCD33).

(44) FIG. 16 shows data obtained from an antibody-dependent cellular phagocytosis (ADCP) assay testing the dose-dependent induction of specific phagocytosis of MOLM-13 cells by donor-derived human macrophages.

(45) Phagocytosis of MOLM-13 cells by donor-derived human macrophages stimulated by SirpIg.αCD33 licMAB, SirpIg.SirpIg.αCD33 licMAB and αCD33 mAB at different concentrations was evaluated by imaging flow cytometry. Percentage of macrophages that engulfed MOLM-13 cells was determined with respect to all macrophages, corrected for unspecific phagocytosis (in the absence of licMABs or mAB) and normalized to maximal phagocytosis (obtained with beads). Error bars indicate SEM of three independent experiments using three independent donors and statistical significance was calculated with a t-test with Welch's correction (*p-value <0.05, **p-value <0.01) (upper panel). Control experiments for phagocytosis are shown in the lower panel.

(46) In conclusion, licMABs significantly increase phagocytosis in comparison to a conventional αCD33 mAB in a concentration dependent manner, most likely because of the blockade of the CD47-Sirp interaction.

(47) FIG. 17 shows licMAB induced NK cell-mediated cytotoxicity of AML patient samples. Cytotoxicity of primary, patient-derived AML cells triggered by αCD33 mAB, SirpIg.αCD33 licMAB and SirpIg.SirpIg.αCD33 licMAB at a concentration of 10 nM was analyzed by determining the percentage of remaining CD33 or CD123 positive cells by flow cytometry. Values were normalized to control cultures. Columns represent the mean value and SEM (error bars) of 6 different AML patient samples. Statistical differences were assessed by the Mann-Whitney U test (*p-value <0.05).

(48) FIGS. 18A-18R show the purity and stability of the two liCAD molecules αPD-L1.αCD3.αCD33 and PD1ex.αCD3.αCD33 in comparison to the bispecific control molecule αCD3.αCD33.

(49) (18A) SDS-PAGE analysis of the proteins described. (M) molecular weight marker.

(50) (18B) Melting curves of the molecules with calculated melting temperatures as determined by thermofluor assay.

(51) Conclusively, liCADs can be expressed and purified in suitable amounts and show sufficient stability in thermo stability assays.

(52) FIGS. 19A-19C show binding of the liCAD binding modules to their targets by flow cytometry analysis as well as the determination of the dissociation constants (K.sub.D).

(53) (19A) Binding of the αPD-L1 scFv and PD1ex binding arms (in the molecules αPD-L1.αCD3.αCD33 and PD1ex.αCD3.αCD33) to PD-L1 on a stably transfected HEK293_PD-L1 cell line overexpressing PD-L1 (upper panel). K.sub.D determination studies are shown in the lower panel and K.sub.D values are indicated (n=3, error bars show SEM).

(54) (19B) Binding of the αCD3 scFv in the molecule PD1ex.αCD3.αCD33 to Jurkat cells (upper panel) and determination of K.sub.D values as indicated (n=3, error bars show SEM) (lower panel).

(55) (19C) Binding of the αCD33 scFv in the molecule PD1ex.αCD3.αCD33 to stably transfected HEK293_CD33 cells overexpressing CD33 and determination of K.sub.F values as indicated (n=3, error bars show SEM).

(56) Taken together, the single modules within the liCAD framework are binding their respective antigens or receptors.

(57) FIG. 20 shows experimental data of a redirected lysis (RDL) assay evaluating the dose-dependent specific killing behavior of pre-expanded T cells. The RDL assay was performed with MOLM-13_PD-L1 (stably expressing PD-L1) as target cells at an effector to target ratio of 5 to 1 (n=3, error bars show SEM).

(58) In RDL assays, the recruitment of T cells to their target cells as well as the specificity of target cell killing can be analyzed in vitro using pre-activated T cells that are able to kill their target cells in a short time frame. Briefly, CD33 and PD-L1 double positive target cells are labeled with Calcein-AM and mixed with pre-activated T cells in an effector to target ratio of 5 to 1. After 4 h incubation, the release of the fluorescent dye into the supernatant was analyzed.

(59) The results indicate that targeting of tumor cells is assured by the tumor antigen targeting domain αCD33 scFv. The extracellular domain of PD1 (PD1ex) only has minor effects on targeting the stable MOLM-13_PD-L1 cell line.

(60) FIGS. 21A-21B display experimental data of preferential killing of stably transfected HEK_CD33_PD-L1 cells over stably transfected HEK_PD-L1 cells using pre-expanded T cells as effectors.

(61) (21A) Dose-dependent preferential killing assay on double positive HEK_CD33_PD-L1 cells (++) versus single positive HEK_PD-L1 cells (+) at an effector to target ratio of 2 to 1 (one exemplary dataset is shown out of three).

(62) Calcein-AM labeled HEK_PD-L1 single positive cells, mixed with unlabeled HEK_CD33_PD-L1 double positive cells, were used as targets to compare efficacy of liCADs and control molecules at a constant effector to target ratio of 2 to 1 in a dose-dependent manner. In a parallel reaction, unlabeled HEK_PD-L1 single positive cells, mixed with Calcein-AM labeled HEK_CD33_PD-L1 double positive cells, were used as targets to compare efficacy of liCADs and control molecules at a constant effector to target ratio of 2 to 1 as well in a dose-dependent manner. % specific lysis was analyzed and demonstrates the potential of liCADs to preferentially kill double positive target cells.

(63) (21B) The assay described in (A) is displayed at maximal protein concentrations of 10 nM.

(64) In summary, killing of tumor cells is highly dependent on the tumor targeting module within the liCAD molecules. The PD1ex domain does not contribute to tumor cell targeting.

(65) FIGS. 22A-22B show flow cytometry data of T cell killing using unstimulated T cells as effector cells at an effector to target ratio of 2 to 1. Dose-dependent induction of target cell killing was evaluated.

(66) (22A) Percentage of survival of CD33 and PD-L1 double positive MOLM-13_PD-L1 cells using liCAD molecules in comparison to controls (n=4). Error bars show SEM.

(67) General efficacy of liCAD molecules was already demonstrated using pre-activated T cells (FIG. 20 and FIG. 21). To evaluate the efficacy of liCAD molecules one step further, in the T cell killing assay shown here, unstimulated T cells were used as effector cells to analyze the induction of T cell effector functions by the liCAD molecules without additional stimuli. Freshly isolated human T cells were incubated with CD33 and PD-L1 double positive stable MOLM-13_PDL1 cells at a constant effector to target ratio of 2 to 1 in a dose dependent manner. After 72 h the percentage of surviving target cells was analyzed by flow cytometry by analyzing CD33 positive living target cells.

(68) The results clearly show that liCAD molecules are able to induce specific killing of double positive target cells in a dose-dependent manner and that PD1ex plays a minor role in targeting the stable MOLM-13_PDL1 cell line.

(69) (22B) Percentage of survival in a direct comparison of liCAD molecules on CD33 single positive MOLM-13 cells versus CD33 and PD-L1 double positive MOLM-13_PD-L1 cells (n=4). Error bars show SEM.

(70) The results demonstrate that liCAD molecules lead to more efficient killing of CD33 and PD-L1 double positive target cells in comparison to CD33 single positive target cells.

(71) FIGS. 23A-23B show flow cytometry based T cell assays with CD33 single positive MOLM-13 cells versus CD33 and PD-L1 double positive MOLM-13_PD-L1 cells using unstimulated T cells as effector cells at an effector to target ratio of 2 to 1.

(72) (23A) Dose-dependent T cell proliferation assay (n=3).

(73) One readout of T cell activation is the analysis of the T cell proliferation behavior. Freshly isolated T cells were labeled with CFSE and mixed with either CD33 single positive MOLM-13 cells or, in a parallel reaction, with CD33 and PD-L1 double positive stable MOLM-13_PDL1 cells at a constant effector to target ratio of 2 to 1 in a dose dependent manner.

(74) With every T cell division, the CFSE cell dye is diluted, which can be monitored by flow cytometry. By this, after 96 h the percentage of proliferated living T cells was evaluated.

(75) The data clearly indicates that, in line with the results from T cell killing assays, liCAD molecules lead to more efficient T cell proliferation when incubated with CD33 and PD-L1 double positive target cells in comparison to CD33 single positive target cells.

(76) (23B) IFNγ release at 5 nM liCAD concentration, displayed as ratio of IFNγ release on MOLM-13 versus MOLM-13_PDL1 cells (n=4). Error bars show SEM.

(77) The activation of T cells correlates with their secretion of cytokines like IFNγ into the supernatant. To evaluate IFNγ release, freshly isolated human T cells were incubated with either CD33 single positive MOLM-13 cells or, in a parallel reaction, with CD33 and PD-L1 double positive stable MOLM-13_PDL1 cells at a constant effector to target ratio of 2 to 1 at a constant liCAD concentration of 5 nM. After 72 h, the supernatant of the reactions was analyzed using a bead-based flow cytometry method, in which the cytokine is captured on pre-coated beads.

(78) The results indicate that both liCAD molecules tested here lead to similar IFNγ release, independent of the presence of PD-L1 on the respective cell line, whereas the bispecific control molecule shows reduced IFNγ levels in the presence of PD-L1.

(79) FIGS. 24A-24B display the flow cytometry analysis of T cell activation using unstimulated T cells as effector cells and double positive MOLM-13_PD-L1 cells as targets at an effector to target ratio of 2 to 1.

(80) (24A) T cell activation without target cells at 5 nM liCAD concentration (n=3).

(81) During the process of T cell activation, different characteristic surface molecules are upregulated. To analyze the effect of liCAD molecules on T cells without target cells, unstimulated human T cells were incubated with liCADs at a concentration of 5 nM. After 96 h, the T cells were analyzed by flow cytometry regarding their expression of the two activation markers CD25 and CD69.

(82) Taken together, liCAD molecules alone only lead to a slight upregulation of activation markers and thereby to minor T cell activation.

(83) (24B) T cell activation with MOLM-13_PD-L1 target cells at 5 nM liCAD concentration (n=3). Error bars show SEM.

(84) Unstimulated human T cells were incubated with either CD33 single positive MOLM-13 cells or with CD33 and PD-L1 double positive stable MOLM-13_PDL1 cells at a constant effector to target ratio of 2 to 1 at 5 nM liCAD concentration. After 96 h the T cells were analyzed by flow cytometry regarding their expression of the two activation markers CD25 and CD69 as well as the upregulation of PD-1.

(85) In summary, liCAD molecules are able to activate T cells in the presence of both MOLM-13 and MOLM-13_PD-L1 cells.

(86) FIGS. 25A-25C schematically show exemplary antibody formats used in clinical therapy or studies. 25A) An IgG antibody targeting a tumor antigen that is expressed on tumor cells. Examples from clinical therapy or studies are Herceptin or Rituximab. This type of antibody has an effector (cell) function and a tumor specificity. 25B) An IgG antibody targeting an immune checkpoint such as CTLA4 or PD-1/L1. Examples from clinical therapy or studies are Ipilimumab, Tremelimumab, CT-011, BMS-936558 or MPDL3280A. This type of antibody has an effector (cell) function, no tumor specificity but inhibits an immune checkpoint by binding to the respective immune checkpoint. 25C) A tumor specific antibody and an immune checkpoint antibody can be combined as combination therapy in clinical applications.

(87) FIGS. 26A-26D schematically show different exemplary antibody formats, fragments and derivatives thereof such as Fc-fusions. 26A) A bispecific antibody in the IgG format with tumor specificity on one binding site and effector (cell) function on the other binding side, e.g. anti-CD3. Examples are Catumaxomab and Ertumaxumab. 26B) A F(ab′)2 fragment with tumor specificity. 26C) A minibody is composed of single-chain Fv (scFv) fragments that are linked via to CH2-CH3 domains (also known as Fc domain). The Fc domain heterodimerizes to fulfill an effector function. 26D) Example for a minibody composing a protein domain, e.g. the extracellular domain of SIRPalpha (or SirpIg), fused to the CH2-CH3 (Fc domain), which upon heterodimerization built the minibody.

(88) FIGS. 27A-27B schematically show exemplary antibody derivatives in preclinical or clinical development. 27A) Bispecific T cell engager (BiTE) consists of two single-chain Fv (scFv) fragments with different specificities. One scFv has a strong effector function mediated by the anti-CD3 binding, the other scFv has a tumor specificity, e.g. CD19 or CD33). ScFvs consist of the variable domains of the heavy and light chain of an IgG antibody, connected by a linker. 27B) A triplebody format consists of three scFvs that are connected by flexible linkers. Triplebodies can have up to three different specificities (indicated by solid lines, dashed lines, no fill). For example, SPM-2 has three different specificities, one tumor specificity (anti-CD123), one effector (cell) specificity (anti-CD16) and one second tumor specificity (anti-CD33).

(89) FIGS. 28A-28B schematically show the trispecific antibodies examples according to the present invention. 28A) Local inhibitory checkpoint antibody derivatives (liCAD) are composed of the extracellular domain of a protein interfering with an immune checkpoint, e. g. SIRPalpha or SirpIg interfering with CD47, a scFv specific for an effector (cell) function and a scFv specific for a tumor antigen. Thus, liCADs combine three functions in one molecule. 28B) A liCAD with two repeats of the extracellular domain of a protein interfering with an immune checkpoint, e.g SIRPalpha or SirpIg or PD-1, which are connected by a linker.

(90) FIGS. 29A-29C schematically show the trispecific antibodies examples according to the present invention. 29A) Local inhibitory checkpoint monoclonal antibody (licMAB) are composted of a tumor specific antibody, preferably in the IgG1 format, which comprises a fusion of the extracellular domain of a protein interfering with an immune checkpoint, e. g. SIRPalpha or SirpIg interfering with CD47. The extracellular domain is fused by a flexible linker. 29B) The licMAB according to A) can comprise mutations in the amino acid sequence of the Fc domain to obtain a stronger effector (cell) function (licMAB Fc engineered) as indicated by the “star.” 29C) A licMAB according to A) can comprise two or more repeats of the extracellular domain of a protein interfering with an immune checkpoint, e. g. SIRPalpha or SirpIg.

(91) In the following, reference is made to the examples, which are given to illustrate, not to limit the present invention.

EXAMPLES

Example 1

(92) In this example, methodology is described that was used by the present inventors for embodiments of the present invention. More specifically, the construction and production of exemplary molecules according to the present invention is described as well as details on cell lines, patients, the preparation of various cells, assays for thermal stability, the detection of binding by flow cytometry, the quantitative determination of cell surface antigens, assays for K.sub.D determination, internalization assays by flow cytometry and/or confocal microscopy, assays for determining antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP) and antibody-dependent cellular cytotoxicity in primary AML patient samples.

(93) SIRPα-antiCD33 licMAB, 2×SIRPα-antiCD33 licMAB and antiCD33 mAb Construction and Production

(94) The antiCD33 variable light (VL) and variable heavy (VH) (clone hP67.6) were generated using custom gene synthesis (GeneArt, Thermo Fisher Scientific). The antiCD33 VL was subeloned into the pFUSE2-CLIg-hk vector (InvivoGen) and the antiCD33 VH into the pFUSE-CHIg-hG1 vector (InvivoGen). To generate the SIRPα-antiCD33 licMAB, the N-terminal Ig-like V-type domain of SIRPα (residues 1-120) was synthesized using custom gene synthesis (GeneArt, Thermo Fisher Scientific) and subcloned into the N-terminus of the antiCD33 light chain (LC) together with a (G.sub.4S).sub.4 linker. A second SIRPα-(G.sub.4S).sub.4 linker cassette was cloned N-terminal of the SIRPα-antiCD33 LC to obtain the 2×SIRPα-antiCD33 light chain. A cassette of SIRPα-(G.sub.4S).sub.2 linker, containing a PreScission protease cleavage site (PreSc) at the C-terminus, was cloned N-terminal of the antiCD33 LC to generate an SIRPα-PreSc-antiCD33 antibody with a cleavable SIRPα. The corresponding plasmids were transfected into Expi293F cells (Thermo Fisher Scientific) according to the manufacturer's protocol. After five to seven days, the cell culture supernatant was harvested and licMABs were purified by protein A affinity chromatography. To obtain the antiCD33 mAb, SIRPα-PreSc-antiCD33 was incubated with PreScission protease for 4 h followed by a second round of protein A affinity chromatography. LicMABs and mAb were dialyzed against Phosphate Buffered Saline (PBS) and size exclusion chromatography (SEC) of the purified molecules was performed using a Superdex 200 increase 10/300 column (GE Healthcare Life Sciences, Little Chalfont, Buckinghamshire, United Kingdom). LicMABs and mAb were then analyzed by 4-20% SDS-PAGE (Expedeon) under reducing conditions and visualized by Coomassie Brilliant Blue staining. Protein concentration was measured with a spectrophotometer (NanoDrop, GE Healthcare Life Sciences, Little Chalfont, Buckinghamshire, United Kingdom) and aliquots were stored at −80° C.

(95) Cell Lines

(96) The MOLM-13 cell line was purchased from the ‘Deutsche Sammlung von Mikroorganismen und Zellkulturen’ (DSMZ, Leibniz-Institut DSMZ, Braunschweig, Germany) and the Flp-IN™-CHO and Flp-IN™ T-Rex™-293 cell lines from Thermo Fisher Scientific (Waltham, Mass., USA). The THP-1 and the Jurkat cell line was kindly provided by collaborators and cultured in RPMI 1640+GlutaMAX (Gibco, Thermo Fisher Scientific) and supplemented with 10% fetal bovine serum (FBS, Gibco, Thermo Fisher Scientific). MOLM-13 cells were cultured in RPMI 1640+GlutaMAX (Gibco, Thermo Fisher Scientific) and supplemented with 20% FBS. Flp-IN™-CHO was maintained in Ham's F-12 (Biochrom) media supplemented with 10% FBS and Flp-IN™ T-Rex™-293 cell line was cultured in DMEM+GlutaMAX and supplemented with 10% FBS and 2 mM L-glutamine. The Flp-IN™-CHO or Flp-IN™ T-Rex™-293 cell lines were engineered to stably express either human CD33 (referred to as CHO_CD33 or CHO.exCD33, HEK.exCD33 or HEK293_CD33), human CD16 (referred to as CHO.exCD16), human CD47 (referred to as CHO_CD47 or CHO.exCD47) or human PD-L1 (referred to as HEK293_PD-L1) and maintained in selection media according to manufacturers' instruction. The HEK293_CD33 cell line has further been engineered to additionally stably express human PD-L1 (HEK_CD33_PD-L1 cell line). The MOLM-13 cell line was engineered to stably expressed human PD-L1 and is referred to MOLM-13_PD-L1 cell line. The Expi293F™ and Freestyle™ 293-F cell line was obtained from Thermo Fisher Scientific and cultured in Expi293™ Expression Medium or FreeStrylc™ 293 Expression Medium, respectively.

(97) Patients

(98) After written informed consent in accordance with the Declaration of Helsinki and approval by the Institutional Review Board of the Ludwig-Maximilians-Universität (Munich, Germany), peripheral blood (PB) or bone marrow (BM) samples were collected from healthy donors (HDs) and patients with AML at initial diagnosis or relapse. PB or BM samples from AML patients were cryoconserved at ≤−80° C. in 80% FCS and 20% dimethyl sulfoxide (Serva Electrophoresis) until usage. PB from IIDs was obtained on the day of the experiment.

(99) Preparation of Peripheral Blood Mononuclear Cells (PBMCs), NK Cells and Monocytes from Whole Human Blood PBMCs from AML patients and HDs were separated from PB by density gradient using the Biocoll separating solution (Biochrom), according to the manufacturer's protocol. NK cells were either expanded ex vivo by culturing PBMCs under IL-2 stimulus as described previously (Carlens et al., 2001; Hum. Immunol.; 62(10); 1092-1098) or freshly isolated by magnetic separation using a human NK cell isolation kit (MACS Miltenyi Biotech) according to the manufacturer's protocol. Monocytes were isolated from PBMCs with human CD14 MicroBeads (MACS Miltenyi Biotech) by magnetic separation following the manufacturer's instructions.

(100) Thermal Stability

(101) The thermal stability of the licMABs and mAb was determined by fluorescence thermal shift assays using the CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Munich, Germany) (Boivin et al., 2013; Protein Expr. Purif.; 91(2); 192-206). 10 μg of protein containing 1×SYPRO Orange (Thermo Fisher Scientific) were measured using FAM and SYBR Green I filter pairs.

(102) Detection of Binding by Flow Cytometry

(103) If not otherwise stated, flow cytometry analyses were performed on a Guava easyCyte 6HT instrument (Merck Millipore, Billerica, Mass., USA) and data was plotted with GuavaSoft software version 3.1.1 (Merck Millipore, Billerica, Mass., USA).

(104) MOLM-13, SEM, CHO_CD33 and CHO_CD47 cells were stained with 15 μg/ml of licMABs or mAb followed by staining with a secondary FITC antiHuman IgG Fc antibody (clone HP6017, BioLegend). The median fluorescence intensity (MFI) ratio was calculated dividing MFI of the antibody by the MFI of the isotype control.

(105) K.sub.D Determination

(106) CD33 equilibrium binding constants (K.sub.L, as an avidity measurement) of the licMABs and mAb were determined by calibrated flow cytometry analyses as previously described (Benedict et al., 1997; J. Immunol. Methods; 201(2); 223-231). Briefly, MOLM-13 cells were incubated with licMABs or mAb in concentrations ranging from 0.01 to 15 μg/ml and stained with a FITC antiHuman IgG Fc (clone HP6017, BioLegend) secondary antibody by flow cytometry. The instrument was calibrated with 3.0-3.4 μm Rainbow Calibration particles of 8 peaks (BioLegend), the maximum MFI value was set to 100% and all data points were normalized accordingly. The assay was performed in quadruplicates and the values were analyzed by non-linear regression using a one-site specific binding model.

(107) Internalization Assay by Flow Cytometry

(108) MOLM-13 cells were incubated with 15 μg/ml of licMABs or mAb either on ice-cold water for 2 h (to prevent internalization) or at 37° C. for 30, 60 or 120 min. Cells were then washed with ice-cold FACS buffer and antibodies remaining on the surface were detected by staining with FITC antiHuman IgG Fc (clone HP6017, BioLegend). To define the background fluorescence, MOLM-13 cells. were directly stained with the secondary antibody. The internalization was calculated as follows:

(109) $\mathrm{Internalization}\left(\%\right)=\frac{\left({\mathrm{MFI}}_{4°C.}-{\mathrm{MFI}}_{\mathrm{background}}\right)-\left({\mathrm{MFI}}_{37°C.}-{\mathrm{MFI}}_{\mathrm{background}}\right)}{\left({\mathrm{MFI}}_{4°C.}-{\mathrm{MFI}}_{\mathrm{background}}\right)}\times 100$

(110) Internalization Assay by Confocal Microscopy

(111) MOLM-13 cells were grown on a poly-L-lysine (Sigma-Aldrich) coated 96-well plate. Subsequently, cells were incubated with 15 μg/ml of licMABs or mAb directly labeled with Alexa Fluor 488 (Antibody Labeling Kit, Thermo Fisher Scientific), either on ice-cold water for 2 h or at 37° C. for 30, 60 or 120 min. Then, cells were fixed and permeabilized in 20 mM PIPES pH 6.8, 4% formaldehyde, 0.2% Triton X-100, 10 mM EGTA, 1 mM MgCl.sub.2 at room temperature for 10 min, followed by incubation in blocking solution (3% bovine serum albumin in PBS). Cells were washed three times with 0.05% Tween 20 in PBS and stored in PBS until examination on a fully automated Zeiss inverted microscope (AxioObserver Z1) equipped with a MS-2000 stage (Applied Scientific Instrumentation, Eugene, Orlando, USA), a CSU-X1 spinning disk confocal head (Yokogawa) and a LaserStack Launch with selectable laser lines (Intelligent Imaging Innovations, Denver, Colo.). Images were acquired using a CoolSnap HQ camera (Roper Scientific, Planegg, Germany), a 63× oil objective (Plan Neofluoar 63×/1.25) and the Slidebook software (version 6.0; Intelligent Imaging Innovations, Denver, Colo.). Images were processed with Adobe Photoshop CS4 (Adobe Systems, Mountain View, Calif., USA).

(112) Antibody-Dependent Cellular Cytotoxicity (ADCC)

(113) Target cells (MOLM-13 or SEM) were labeled with calcein AM (Thermo Fisher Scientific) according to manufacturer's protocol. Calcein-labeled target cells were incubated with freshly isolated or IL-2 expanded NK cells in an effector-to-target (E:T) ratio of 2:1 and licMABs or mAb at different concentrations for 4 h. Target cells were cultured in 10% Triton X-100 to assess the maximum unspecific lysis. Calcein release was measured by fluorescence intensity with an Infinite® M100 plate reader instrument (TECAN, Männedorf, Switzerland) and specific lysis was calculated as follows:

(114) $\mathrm{Specific}\mathrm{lysis}\left(\%\right)=\frac{{\mathrm{Fluorescence}}_{\mathrm{Sample}}-{\mathrm{Fluorescence}}_{\mathrm{Spontaneous}\mathrm{lysis}}}{{\mathrm{Fluorescence}}_{\mathrm{Maximum}\mathrm{lysis}}-{\mathrm{Fluorescence}}_{\mathrm{Background}}}\times 100$

(115) Averaged specific lysis of triplicates or quadruplicates were plotted according to a dose-response curve and analyzed using the integrated four parameter non-linear fit model.

(116) Antibody-Dependent Cellular Phagocytosis (ADCP)

(117) Phagocytosis assay was performed as described previously (Blume et al., 2009; J. Immunol.; 183(12); 8138-8147). Briefly, isolated monocytes were stained with PKH67 (Sigma-Aldrich) according to the manufacturer's instructions and differentiated to macrophages by 20 ng/ml Macrophage-Colony Stimulator Factor (M-CSF) (R&D Systems) in X-VIVO 10 medium (Lonza) supplemented with 10% autologous serum. MOLM-13 cells were stained with PKH26 (Sigma-Aldrich) following the manufacturer's instructions and incubated in a 1:2 E:T ratio with licMABs or mAb concentrations ranging from 0.01 nM to 100 nM for 2 h. Polybead® Carboxylate Red Dye Microspheres 6 μm (Polysciences) were used as a positive control and incubation either at 4° C. or at 37° C. in the presence of 10 μM Cytochlasin D (Sigma-Aldrich) served as a negative control. Cells were harvested, measured by imaging flow cytometry using an ImageStream®.sup.X Mark II instrument (Merck Millipore, Billerica, Mass., USA) and analyzed with IDEAS® and INSPIRE® Software (Merck Millipore, Billerica, Mass., USA). The maximum phagocytosis value was set to 100% and all data points were normalized accordingly. Mean values and standard errors of triplicates were calculated and plotted.

(118) Antibody-Dependent Cellular Cytotoxicity in Primary AML Patient Samples

(119) Ex vivo expanded primary AML cells were co-cultured with freshly isolated healthy donor NK cells, at an E:T ratio of 5:1 in an ex vivo long term culture system as described by Krupka and coworkers (Krupka et al., 2016; Leukemia; 30(2); 484-491) (Krupka et al., 2014; Blood; 123(3); 356-365). Antibodies were added at a final concentration of 10 nM. After 24 hours, cells were harvested, stained for CD16 (clone B73.1), CD56 (clone HCD 56), CD33 (clone WM53) and in some cases CD123 (clone 6H6; all antibodies from Biolegend) and analyzed by flow cytometry with a BD LSR 11 (Becton Dickinson, Heidelberg, Germany). The percentage of residual CD33 or CD123 positive cells in treated cultures relative to control cultures was used to determine licMAB-mediated cellular cytotoxicity.

(120) Plotting and Statistical Analysis

(121) Unless stated otherwise, data were analyzed and plotted with GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla, Calif., USA).

(122) Differences in phagocytosis were calculated using an unpaired, parametric Student's t-test with Welch correction and statistical differences of patient characterization and responses were assessed by the Mann-Whitney U test. Statistical significance was considered for p-value <0.05 (*), <0.01 (**), <0.001 (***) and <0.0001 (****).

Example 2

(123) In this example, a molecule comprising a binding site with specificity for CD33 (to bind to CD33-positive leukemic cells), a binding site with specificity for CD16 (for recruitment of immune cells as effector cells), and a binding site with specificity for the checkpoint molecule CD47 (for inhibition of antiphagocytic checkpoint signaling) and several related constructs were generated and tested. Since these molecules included antibody-derived binding domains, they are referred to as a “local inhibitory checkpoint antibody derivatives” (liCADs).

(124) Design, Expression and Purification of liCADs

(125) To target CD47, the extracellular N-terminal Ig variable domain of SIRPα (herein called SIRP-Ig or SirpIg) was used, which has been shown to be sufficient for CD47 binding (Barclay et al., 2009). In order to modulate binding affinities, molecules carrying two copies of SIRP-Ig were designed. Apart from the varying N-terminal module 1 all constructs had a central anti-CD16 scFv (derived from murine hybridoma 3G8) (Fleit et al., 1982), recruiting immune effector cells (module 2). CD16, also known as Fc gamma receptor IIIa (FcγRIIIa) is expressed on NK cells, dendritic cells (DCs) and macrophages and mediates antibody dependent cellular cytotoxicity (ADCC) or antibody dependent phagocytosis (ADCP), respectively (Guilliams et al., 2014). Using a scFv specific for CD16 allows to exclude side effects generated by the Fc part of a conventional mAb that would activate far more Fc receptor expressing immune cells, thus leading to fatal side effects like the cytokine release syndrome (Brennan et al., 2010). On the C-terminus of the molecule an anti-CD33 scFv (derived from gemtuzumab ozogamicin) is expressed (module 3). CD33 is a tumor specific marker that is highly overexpressed in acute myeloid leukemia (AML) and has successfully been used as tumor target before (Larson et al., 2005; Krupka et al., 2014). As control molecules, an anti-CD47 scFv (triplebody control) was included, as well as a high affinity version of SIRP-Ig (SirpIg_CV1) that had been published before (FIG. 4 A). SirpIg_CV1 binds to CD47 with much higher affinity (1 μM) compared to SIRP-Ig (1 μM) (Weiskopf et al., 2013).

(126) The protein domains used had the following sequences:

(127) Amino Acid Sequence of Sirp-Ig (SEQ ID NO: 1):

(128) TABLE-US-00001 EEELQVIQPDKSVLVAAGETATLRCTATSLIPVGPIQWFRGAGPGRELIY NQKEGHFPRVTTVSDLTKRNNMDFSIRIGNITPADAGTYYCVKFRKGSPD DVEFKSGAGTELSVRAKPS

(129) Amino Acid Sequence of Vh CD16scFv (3G8 Clone) (SEQ ID NO: 2):

(130) TABLE-US-00002 QVTLKESGPGILQPSQTLSLTCSFSGFSLRTSGMGVGWIRQPSGKGLEWL AHIWWDDDKRYNPALKSRLTISKDTSSNQVFLKIASVDTADTATYYCAQI NPAWFAYWGQGTLVTVSA

(131) Amino Acid Sequence of Vl CD16scFv (3G8 Clone) (SEQ ID NO: 3):

(132) TABLE-US-00003 DTVLTQSPASLAVSLGQRATISCKASQSVDFDGDSFMNWYQQKPGQPPKL LIYTTSNLESGIPARFSASGSGTDFTLNIHPVEEEDTATYYCQQSNEDPY TFGGGTKLEIK

(133) Amino Acid Sequence of Vl CD33scFv (SEQ ID NO: 4):

(134) TABLE-US-00004 DIQLTQSPSTLSASVGDRVTITCRASESLDNYGIRFLTWFQQKPGKAPKL LMYAASNQGSGVPSRFSGSGSGTEFTLTISSLQPDDFATYYCQQTKEVPW SFGQGTKVEVK

(135) Extracted CDRS for light chain are:

(136) CDRL1: RASESLDNYGIRFLT (SEQ ID NO: 29)

(137) CDRL2: AASNQGS (SEQ ID NO: 30)

(138) CDRL3: QQTKEVPWS (SEQ ID NO: 31)

(139) Amino Acid Sequence of Vh CD33scFv (SEQ ID NO: 5):

(140) TABLE-US-00005 EVQLVQSGAEVKKPGSSVKVSCKASGYTITDSNIHWVRQAPGQSLEWIGY IYPYNGGTDYNQKFKNRATLTVDNPTNTAYMELSSLRSEDTAFYYCVNGN PWLAYWGQGTLVTVSS

(141) Extracted CDRS for heavy chain are:

(142) CDRH1: DSNIH (SEQ ID NO: 32)

(143) CDRH2: YIYPYNGGTDYNQKFKN (SEQ ID NO: 33)

(144) CDRH3: GNPWLAY (SEQ ID NO: 34)

(145) Between the checkpoint binding module (third binding site, e.g. Sirp-Ig, PD1ex, CTLA4ex, αPDL1), Vh und VI domains and between the scFvs, GGGS-based linkers were included.

(146) Linker Sequences:

(147) TABLE-US-00006 (SEQ ID NO: 6) Gly Gly Gly Ser

(148) and tandem-repeats thereof, n=2-8, (SEQ ID NO: 7-13)

(149) TABLE-US-00007 (SEQ ID NO: 14) Gly Gly Gly Gly Ser

(150) and tandem-repeats thereof, n=2-8, (SEQ ID NO: 15-21)

(151) Sequence Constant Region IgG1 Format (CH1, Hinge, CH2, CH3): (SEQ ID NO: 22)

(152) TABLE-US-00008 ASTKGPSVFPLAPSSKSTSGGTAALGCLVKDYFPEPVTVSWNSGALTSGV HTFPAVLQSSGLYSLSSVVTVPSSSLGTQTYICNVNHKPSNTKVDKKVEP KSCDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVVDVS HEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGK EYKCKVSNKALPAPIEKTISKAKGQPREPQVYTLPPSREEMTKNQVSLTC LVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLYSKLTVDKSRW QQGNVFSCSVMHEALHNHYTQKSLSLSPGK

(153) Sequence Constant Region of the Light Chain (CL): (SEQ ID NO: 23)

(154) TABLE-US-00009 VAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVDNALQSGNS QESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSF NRGEC

(155) Sequence Constant Region IgG1 Format; Fc-Engineered (SEQ ID NO: 24)

(156) Examplarily used mutations: S239D and 1332E (shown below in bold) EU numbering according to Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S. & Foeller, C. (1991) Sequences of Proteins of Immunological Interest (U.S.Dept.ofHealthandHum. Serv., Bethesda)

(157) The used mutations in this example were S239D and I332E, but others may also be used.

(158) TABLE-US-00010 ASTKGPSVFPLAPSSKSTSGGTAALGCLVKDYFPEPVTVSWNSGALTSGV HTFPAVLQSSGLYSLSSVVTVPSSSLGTQTYICNVNHKPSNTKVDKKVEP KSCDKTHTCPPCPAPELLGGPDVFLFPPKPKDTLMISRTPEVTCVVVDVS HEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGK EYKCKVSNKALPAPEEKTISKAKGQPREPQVYTLPPSREEMTKNQVSLTC LVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLYSKLTVDKSRW QQGNVFSCSVMHEALHNHYTQKSLSLSPGK

(159) Fc-engineered mutations may be produced and used for example in accordance with: Engineered antibody Fc variants with enhanced effector function Greg A. Lazar, Wei Dang, Sher Karki, Omid Vafa, Judy S. Peng, Linus Hyun, Cheryl Chan, Helen S. Chung, Araz Eivazi, Sean C. Yoder, Jost Vielmetter, David F. Carmichael, Robert J. Hayes, and Bassil I. Dahiyat

(160) Sequence Used for PD1ex (Extracellular Domain of PD1) (SEQ ID NO: 25)

(161) TABLE-US-00011 NPPTFSPALLVVTEGDNATFTCSFSNTSESFVLNWYRMSPSNQTDKLAAF PEDRSQPGQDCRFRVTQLPNGRDFHMSVVRARRNDSGTYLCGAISLAPKA QIKESLRAELRVTERRA

(162) Sequence Used for αPDL1 (Anti-PDL1; SEQ ID NO: 26 & 27)

(163) TABLE-US-00012 V1 PDL1 scFv (SEQ ID NO: 26) DIQMTQSPSSLSASVGDRVTITCRASQDVSTAVAWYQQKPGKAPKLLIYS ASFLYSGVPSRFSGSGSGTDFTLTISSLQPEDFATYYCQQYLYHPATFGQ GTKVEIKR Vh PDL1 scFv (SEQ ID NO: 27) EVQLVESGGGLVQPGGSLRLSCAASGFTFSDSWIHWVRQAPGKGLEWVAW ISPYGGSTYYADSVKGRFTISADTSKNTAYLQMNSLRAEDTAVYYCARRH WPGGFDYWGQGTLVTVSA

(164) SEQ ID NO: 28 is the combination of SEQ ID NO: 26 and 27 with a (GGGGS).sub.4-linker in between.

(165) SEQ ID NO: 22 is the sequence for the constant regions of the IgG1 format and includes the domains CH1, hinge, CH2 and CH3; this is the wildtype sequence. In contrast thereto, SEQ ID NO: 24 shows the same sequence for the constant regions of the IgG1 format but with mutations S239D and 1332E; this is the mutated sequence for producing an Fc-engineered fragment. This is also the sequence which was used in embodiments of the present disclosure; the mutations are located in the CH2 domain. The Fc-fragment and the Fc-engineered fragment consist of domains CH2 and CH3 only. In all of these constant regions of the IgG1 format, the numbering according to Kabat is used. In SEQ ID NO: 24 the entire constant region including domains CH1, hinge, CH2 and CH3 is shown; however, it is clear to a person skilled in the art that the corresponding Fc-engineered fragment only contains the respective CH2 and CH3 domains.

(166) For the various molecules according to the present invention a number of linkers were used, shown herein as SEQ ID NO: 6-SEQ ID NO: 21. SEQ ID NO: 6 is a GGGS-linker and SEQ ID NO: 7-13 are tandem repeats thereof, wherein the linker occurs 2-8 times, SEQ ID NO: 14 is a GGGGS-linker and SEQ ID NO: 15-21, again, are tandem repeats thereof, wherein the linker occurs 2-8 times. These linkers allow for an occurrence of respective domains/binding sites (first binding site, second binding site and third binding site) within molecules (liCADs and licMABs) of the present invention.

(167) PDL1 as used herein refers to the programmed death-ligand 1 which binds to its corresponding receptor PD1. “PD1ex” is the extracellular domain of PD1.

(168) The liCADs were expressed in Drosophila melanogaster Schneider 2 (S2) cells and purified from the insect cell medium after secretion. The purification strategy included a capture step (Ni-NTA affinity chromatography) via the N-terminal hexa-histidine (6×HIS) tag, followed by anion exchange (IEC) and size exclusion chromatography (SEC), resulting in monomeric soluble protein (FIG. 4B).

(169) For binding tests using flow cytometry, cells expressing CD47, CD33 or CD16, respectively, were incubated with the purified liCADs for 30 minutes on ice. Unbound protein was washed away and bound protein was detected using an Alexa488-conjugated antibody specific for the 6×HIS tag. Cells were again incubated for 30 minutes on ice, washed twice and subsequently analysed in a Guava easyCyte 6HT (Merck Millipore).

(170) These experiments confirmed binding of SirpIg, anti-CD16 scFv and anti-CD33 scFv to their respective binding partners/antigens (i.e. binding of Sirp-Ig to CD47, binding of anti-CD16 scFv to CD16 and binding of anti-CD33 scFv to CD33) (see FIGS. 5 and 6).

(171) liCAD-Induced Redirected Lysis of Tumor Cells

(172) In this experiment, it was tested whether the prepared liCAD molecules would indeed induce tumor cell killing by the recruitment of NK cells in vitro. To this end, a redirected lysis (RDL) assay was carried out with the MOLM 13 cell line, which expresses CD33 and CD47 at high level. The RDL assay functions analogous to an antibody dependent cellular cytotoxicity (ADCC) assay, but recruitment and activation of NK cells is not mediated by the Fc domain of an antibody, but by the scFv against CD16. As effector cells, isolated peripheral blood mononuclear cells (PBMCs) that had been expanded as described previously (Alici et al., 2008) were used. Effector cells and calcein labeled target cells were mixed in a ratio of 2:1 and incubated with increasing protein concentrations for 4 hours at 37° C./5% CO.sub.2. Afterwards fluorescence intensity of calcein was measured from the cell supernatant using the Infinite M1000 PRO (Tecan) plate reader.

(173) The results are shown in FIG. 7. As expected, molecules targeting CD47 and CD33 simultaneously show improved cell lysis (FIG. 7A) compared to monospecific molecules only targeting CD33. Further, we determined the EC.sub.50 values (concentrations of half maximum lysis) by dose response curves. EC.sub.50 values achieved for the liCADs were 1.5 μM and 22 μM for the double SIRP-Ig and single SIRP-Ig, respectively. Thus, it is possible to regulate the degree of checkpoint inhibition. This is advantageous for systemic administration in vivo. In comparison to the control molecules (triplebody) the liCADs achieved a similar range of specific lysis.

(174) As CD47 is a marker of self and thus expressed on every cell, it is necessary to avoid killing all CD47 positive cells. To this end, a preferential RDL assays was carried out to show that liCADs preferentially eliminate CD47/CD33 double positive cells over CD47 single positive cells (FIGS. 7B and 7C).

(175) The preferential lysis assay was carried out using CD47+ single positive HEK cells mixed with CD47+, CD33+ double positive HEK cells. Effector cells and calcein stained target cells (one reaction with single positive stained and one reaction with double positive stained) were mixed in a 2:1 ratio again and incubated with the maximal used protein concentration in the redirected lysis assay or with the evaluated EC50 value for 4 hours at 37° C./5% CO.sub.2. Afterwards fluorescence intensity of calcein was measured from the cell supernatant using the Infinite M1000 PRO (Tecan) plate reader.

(176) As shown in FIGS. 7B and 7C, CD47+CD33+ HEK cells are preferentially killed in case of the Sirp-Ig-CD16-CD33 and Sirp-Ig-Sirp-Ig-CD16-CD33 liCAD, but not in case of a control triplebody that targets CD47 with high affinity. Moreover, our low affinity molecules are comparable to a bispecific control that does not target CD47 and at the EC50 value these molecules do not redirect killing of CD47+ cells at all in contrast to the triplebody control.

(177) Phagocytosis Assay

(178) Besides expression on NK cells, CD16 is also expressed on macrophages. Therefore, it was investigated if the liCADs can recruit macrophages as effector cells. To support the results seen in the RDL assays, it was analyzed if the prepared liCADs affect phagocytosis.

(179) Regarding the tri-specific molecules (SIRP-Ig-αCD16-αCD33 and SIRP-Ig-SIRP-Ig-αCD16-αCD33) it was hypothesized that macrophages may be activated through CD16 signaling. Hence, an increase in phagocytosis should mainly be dependent on the SIRPα-CD47 interaction. Consequently, the liCADs combine tumor cell targeting via CD33 together with a local immune checkpoint inhibition through their low binding affinity for CD47.

(180) A phagocytosis assay was performed generating M2 macrophages for 5 days in culture and incubation of macrophages with MOLM13 target cells in a 1:2 ratio. Cells were mixed and incubated with increasing amount of LiCAD concentration in serum free conditions for 2 hours at 37° C./5% CO.sub.2. Afterwards cells were collected and FACS analyzed for macrophages that had taken up target cells.

(181) As shown in FIG. 8 the Sirp-Ig-Sirp-Ig-CD16-CD33 liCAD is better in mediating phagocytosis compared to the Sirp-Ig-CD16-CD33 molecule, which suggests indeed an additive effect of the blocking by Sirp-Ig. Overall both liCAD molecules perform considerably better than a conventional used mAB against CD47. As a control, Sirp-Ig only was used to test if blocking of the immune checkpoint alone is enough to induce phagocytosis, which is not the case.

(182) Tables

(183) TABLE-US-00013 TABLE 1 Exemplary antibody formats used in clinical therapy or studies, respectively Effector Tumor Checkpoint Format function specific inhibition Example IgG + + − Herceptin (α-Her2) targeting Rituximab (α-CD20) tumor antigen IgG + − + Ipilimumab, Tremelimumab targeting (α-CTLA4) immune CT-011 (α-PD1) checkpoint BMS-936558, MPDL3280A (α-PDL1) IgG + + + Ipilimumab + Bevaeizumab (α-CTLA4 + α-VEGF) Galiximab + Rituximab (α-PDL1 + α-CD20) CT-011 + Rituximab (α-PD1 + α-CD20) bisepcific + + − Catumaxotnab IgG (α-CD3 × α-EpCAM) Ertumaxumab (α-CD3 × α-Her2) F(ab′)2 − + − MDX-2120, MDX-H210 (α-CD64 × α-HER2) MDX-447 (α-CD64 × α-EGFR) Minibody + − + Sirpα - Fc fusions BiTE ++ + − Blinaturnomab α-CD3 × α-CD19 AMG330 α-CD3 × α-CD33 Triplebody ++ ++ − SPM-2 (α-CD123 × α-CD16 α-CD33)

(184) In FIGS. 25-27, antibody constant domains are shown as white rectangles. Variable heavy and light chains are shown in white (tumor antigen specific), dotted (effector cell specific) and dashed line rounded rectangles (immune checkpoint specific). White spikes represent endogenous extracellular domains of immune checkpoint receptors. BiTE, bispecific T cell engager; F(ab′)2, Fragment antigen binding; IgG, immunoglobulin G. Single chain fragment variables (scFv) are depicted as two rounded rectangles with a diagonal black line across (e.g. the triplebody at the bottom of the table includes three scFv).

(185) TABLE-US-00014 TABLE 2 Examples of tumor specific makers, related disease and available immunotherapy format Antigen Disease CD19 NHL, B-ALL CD20 B cell lymphoma Her2/neu Breast cancer CD123 AML CEA Gastrointestinal cancer, lung cancer EPCAM Ovarian cancer, colorectal cancer

(186) TABLE-US-00015 TABLE 3 Cell surface molecules on different immune effector cells Immune cells Surface receptors T cells CD3, TCRαβ, Nk cells CD16, NKG2D, NKp30, NKp40, LFA1 Macrophages CD89, CD64, CD32a, CD15a, CD16 Monocytes CD89, CD64, CD32a, CD15a, CD16 Neutrophilic Granulocytes CD89, CD64, CD32a, CD16

(187) TABLE-US-00016 TABLE 4 Examples for molecules according to the invention Effector Tumor Checkpoint Effector Format function specificity function recruitment liCAD ++ ++ + + liCAD ++ ++ ++ + licMAB + ++ ++ + licMAB ++ ++ + Fc engineered licMAB + ++ +++ +

(188) In FIGS. 28-29, variable heavy and light chains are shown in white (tumor antigen specific) and black (effector cell specific). White spikes represent endogenous extracellular domains of immune checkpoint receptors (FIGS. 28A-28B and 29A-29C).

(189) The first molecule comprises a SIRPα-Ig linked to a tumor cell-specific and an immune cell-specific scFv. The second molecule comprises two SIRPα-Igs linked to a tumor cell-specific and an immune cell-specific scFv. The third molecule consists of an IgG antibody with variable domains having binding specificity for the tumor cell and further linked to two SIRPα-Igs.

REFERENCES

(190) Aigner, M. et al. T lymphocytes can be effectively recruited for ex vivo and in vivo lysis of AML blasts by a novel CD33/CD3-bispecific BiTE antibody construct. Leukemia 27, 1107-1115, doi:leu2012341 [pii] 10.1038/leu.2012.341 (2013). Alici, E. et al. Autologous antitumor activity by NK cells expanded from myeloma patients using GMP-compliant components. Blood 111, 3155-3162, doi:blood-2007-09-110312 [pii]10.1182/blood-2007-09-110312 (2008). Barclay, A. N. & Van den Berg, T. K. The interaction between signal regulatory protein alpha (SIRPalpha) and CD47: structure, function, and therapeutic target. Annu Rev Immunol 32, 25-50, doi:10.1146/annurev-immunol-032713-120142 (2014). Barclay, A. N. Signal regulatory protein alpha (SIRPalpha)/CD47 interaction and function. Curr Opin Immunol 21, 47-52, doi:S0952-7915(09)00004-1 [pii] 10.1016/j.coi.2009.01.008 (2009). Brennan, F. R. et al. Safety and immunotoxicity assessment of immunomodulatory monoclonal antibodies. MAbs 2, 233-255, doi:11782 [pii] (2010). Chao, M. P. et al. Anti-CD47 antibody synergizes with rituximab to promote phagocytosis and eradicate non-Hodgkin lymphoma. Cell 142, 699-713, doi:S0092-8674(10)00892-5 [pii]10.1016/j.cell.2010.07.044 (2010). Fleit, H. B., Wright, S. D. & Unkeless, J. C. Human neutrophil Fc gamma receptor distribution and structure. Proceedings of the National Academy of Sciences of the United States of America 79, 3275-3279 (1982). Guilliams, M., Bruhns, P., Saeys, Y., Hammad, H. & Lambrecht, B. N. The function of Fcgamma receptors in dendritic cells and macrophages. Nat Rev Immunol 14, 94-108, doi:nri3582 [pii] 10.1038/nri3582 (2014). Krupka, C. et al. CD33 target validation and sustained depletion of AML blasts in long-term cultures by the bispecific T-cell-engaging antibody AMG 330. Blood 123, 356-365, doi:blood-2013-08-523548 [pii] 10.1182/blood-2013-08-523548 (2014). Larson, R. A. et al. Final report of the efficacy and safety of gemtuzumab ozogamicin (Mylotarg) in patients with CD33-positive acute myeloid leukemia in first recurrence. Cancer 104, 1442-1452, doi:10.1002/cncr.21326 (2005). Majeti, R. et al. CD47 is an adverse prognostic factor and therapeutic antibody target on human acute myeloid leukemia stem cells. Cell 138, 286-299, doi:S0092-8674(09)00650-3 [pii] 10.1016/j.cell.2009.05.045 (2009). Pardoll, D. M. The blockade of immune checkpoints in cancer immunotherapy. Nat Rev Cancer 12, 252-264, doi:nrc3239 [pii] 10.1038/nrc3239 (2012). Petersdorf, S. H. et al. A phase 3 study of gemtuzumab ozogamicin during induction and postconsolidation therapy in younger patients with acute myeloid leukemia. Blood 121, 4854-4860, doi:blood-2013-01-466706 [pii] 10.1182/blood-2013-01-466706 (2013). Weiskopf, K. et al. Engineered SIRPalpha variants as immunotherapeutic adjuvants to anticancer antibodies. Science 341, 88-91, doi:science.1238856 [pii]10.1126/science.1238856 (2013). Willingham, S. B. et al. The CD47-signal regulatory protein alpha (SIRPα) interaction is a therapeutic target for human solid tumors. Proc Natl Acad Sci USA 109, 6662-6667, doi:1121623109 [pii] 10.1073/pnas.1121623109 (2012). Wines, B. D. et al. The IgG Fc contains distinct Fc receptor (FcR) binding sites: the leukocyte receptors Fc gamma RI and Fc gamma RIIa bind to a region in the Fc distinct from that recognized by neonatal FcR and protein A. J Immunol. 164(10), 5313-5318 (2000).