Inhibitory Anti-cd93 antibodies

Abstract

The current disclosure relates agents, such as binding proteins, that bind vascular CD93, wherein binding of said agent to CD93 results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation The agents or binding proteins, nucleic acid molecules or pharmaceutical compositions thereof, may be used in medical treatments, such as cancer therapies. The binding proteins may also be used for engineering cells to express a chimeric antigen receptor having a binding protein of the present disclosure as antigen binding domain.

Description

[0001] BINDING AGENTS

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to binding agents, being or comprising binding proteins that bind human vascular CD93. In some aspects, the binding proteins constitute CD93 binding antibodies, or fragments thereof. In some embodiments, the agent is a nucleic acid molecule encoding a binding agent, which may be in the form of mRNA to be inserted into a gene vector in a transduced host cell, thereby expressing a binding protein. In some embodiments, the agents, binding proteins, antibodies, or pharmaceutical compositions thereof, are used in medical treatments, such as cancer therapies. In other embodiments, the binding proteins are used for engineering cells to express a chimeric antigen receptor having a binding protein of the present disclosure as antigen binding domain.

[0004] BACKGROUND

[0005] Cancer is one of the most prevalent deadly diseases, which despite recent advances in diagnosis and treatment still accounts for a substantial number of deaths each year. A glioma is tumor that starts in the glial cells of the brain or the spine, and gliomas comprise about 30 percent of all brain tumors and central nervous system tumors, and 80 percent of all malignant brain tumors. Gliomas are very invasive tumors with high post-operative tumor recurrence, and complete surgical resection of the tumor is not possible because of its invasiveness. Thus, at present, no curative treatment for glioma patients exists. Accordingly, novel and enhanced treatments of cancers, such as gliomas, are needed.

[0006] SUMMARY

[0007] An object of the present disclosure is to provide novel and enhanced binding agents, which may be used in medical treatments. This object is obtained by an agent or binding protein that specifically binds vascular CD93.

[0008] In some aspects are provided an agent comprising a binding moiety that specifically binds to vascular Cluster of Differentiation 93, CD93, wherein binding of said agent to CD93 results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation.

[0009] In some embodiments the agent or binding moiety thereof is a binding protein that comprises a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (VH) and a light chain variable domain (VL), each comprising three complementarity determining regions (CDRs), wherein the amino acid sequences of the CDRs are selected from the group comprising: VHCDR1 as defined by SEQ ID NO: 1 ; VHCDR2 as defined by SEQ ID NO: 2; VHCDR3 as defined by SEQ ID NO: 3; VLCDR1 as defined by SEQ ID NO: 4; VLCDR2 as defined by AAS; VLCDR3 as defined by SEQ ID NO: 5; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto, wherein said binding protein has a KD< 21 nM.

[0010] In some embodiments, the CDRs are individually selected from the groups comprising: VHCDR1 is selected from SEQ ID NO: 8-10; VHCDR2 is selected from SEQ ID NO: 11-14; VHCDR3 is selected from SEQ ID NO: 15-18; VLCDR1 as defined by SEQ ID NO: 4; VLCDR2 as defined by AAS; VLCDR3 is selected from SEQ ID NO : 19-21 ; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto.

[0011] In some embodiments, the VH sequence comprises an amino acid sequence selected from the group consisting of SEQ ID NO: 22-25 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, and wherein the VL sequence comprises an amino acid sequence selected from the group consisting of SEQ ID NO: 26-29 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, and wherein the CDR sequences comprise no variations in the amino acid sequence, or wherein the sequence variation of the CDR amino acid sequences is at most 5%, such as 4 %, 3 %, 2 %, 1 % or less.

[0012] In some aspects, the binding protein is a monoclonal antibody, such as an lgG1 I.ALA antibody, or an antigen-binding fragment selected from the group consisting of Fv fragments, such as scFv fragments, Fab-like fragments, such as Fab or F(ab’)2fragments, disulphide- bonded fragments and domain antibodies. Typically, the binding protein is human or of human origin.

[0013] In some aspects, an engineered cell is provided. The cell may be engineered to express a chimeric antigen receptor (CAR), wherein the CAR comprises an antigen-binding domain, a transmembrane domain connected to the antigen binding domain by a hinge region, and an intracellular domain optionally connected to one or more co-stimulatory domains, wherein the antigen binding domain comprises an scFv fragment of a binding protein of the present disclosure.

[0014] According to some aspects, a nucleic acid molecule encoding an agent or binding molecule, or the present disclosure is provided. In some aspect, a vector comprising the nucleic acid molecule is provided. According to some aspects, a pharmaceutical composition is provided. The pharmaceutical composition comprises an agent, a binding protein, a nucleic acid molecule or an engineered cell as described above, and a pharmaceutically acceptable carrier or excipient.

[0015] The binding agent, binding protein, nucleic acid molecule, engineered cell, or pharmaceutical composition as described above may be for use in therapy. In some embodiments, the therapy is cancer therapy, such as glioma therapy. In some embodiments, the cancer is glioblastoma.

[0016] Other objects and advantages will become apparent to those skilled in the art from a review of the ensuing detailed description, which proceeds with reference to the following illustrative drawings, and the attendant claims.

[0017] BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above, as well as additional objects, features and advantages of the present inventive concept, will be better understood through the following illustrative and non-limiting detailed description of different embodiments of the present inventive concept, with reference to the appended drawings, wherein:

[0019] Figure 1 illustrates the effect of a CD93 binding agent generic on peri-vascular invasion of tumor cells.

[0020] Figure 2 shows that CD93 expression is enhanced in glioblastoma patients. CD93 mRNA expression in grade II, grade III glioma and grade IV glioblastoma patients. RNASeq data downloaded from the publicly available database CGGA (http: / / qliovis.bioinfo.cnio.es / ).

[0021] **** p<0.0001 ; *p=0.026 One-way ANOVA, Tukey's multiple comparison test.

[0022] Figure 3 shows that CD93 knockout is associated with reduced tumor growth and improved survival. Figure 3A) Quantification of GL261 area showing a significant reduction of tumor growth in CD93 knockout mice (CD93_ / ) compared to the wild-type group. * p<0.05; unpaired t-test. Figure 3B) Survival curve of mice intracranially injected with GL261 glioma cells showing improved survival of CD93 / _mice compared to the wild-type (WT) group. *p < 0.05; Gehan-Breslow-Wilcoxon test. Langenkamp E. et a! Cancer Res. 2015 Nov 1;75(21):4504-16.

[0023] Figure 4 shows that CD93 knockout is associated with an increased expression of leukocyte adhesion molecule VCAM1 in tumor vessels and increased abundance of cytotoxic T cells in GL261 tumors, and is associated with improved survival after aPD1 -blockade immunotherapy. Figure 4A) Immunofluorescent staining showing increased expression of VCAM1 in glioma tumor vessels in CD93-deficient mice. Figure 4B) Quantification graph showing a significant increase in VCAM1 in tumor vessels in CD93 / _GL261 tumors as compared to the wild-type group. ****p<0.0001 ; unpaired 2-tailed t-test. Figure 4C) Immunofluorescent staining showing increased infiltration of CD8+cytotoxic T cells in gliomas in CD93-deficient mice. Figure 4D) Quantification graph showing a significant increase in the abundance of CD8+T cells in CD93AGL261 tumors as compared to the wild-type group. **p<0.01 ; unpaired 2-tailed t- test. Figure 4E) Survival curve of mice intracranially injected with GL261 glioma cells showing improved survival of CD93 / _mice compared to the wild-type group after aPD1 -blockade immunotherapy. aPD1 (aPD1 antibody), Iso (Isotype control IgG). *p < 0.05; **p<0.01 ; ***p< 0.001 ; ****p<0.0001 ; Gehan-Breslow-Wilcoxon test.

[0024] Figure 5 shows that CD93 is required for fibronectin fibrillogenesis. Figure 5A) Immunofluorescent staining showing the effect of CD93 downregulation (siCD93) on the fibronectin fibrillogenesis in human dermal blood endothelial cells (HDBECs) compared to the control condition (siCtrl). Actin is visualised by phalloidin staining. Figure 5B) Immunofluorescent staining of murine brain sections showing the effect of CD93 deletion (CD93_ / ) on fibronectin in the GL261 glioma model. Dotted lines indicate the tumor area. Nuclei are visualised by Hoechst staining. Figure 5C) Quantification graph showing significant reduction of fibronectin deposition in CD93 / _GL261 tumors compared to the Wild-type group. **** p<0.0001 ; unpaired t-test. Figure 5D) Immunofluorescent images showing disruption of vessel associated fibronectin in CD93 knockout GL261 tumors. Vessels are visualized by CD31 staining. Lugano R. et al J Clin Invest. 2018 Aug 1 ;128(8):3280-3297.

[0025] Figure 6 shows that CD93 is expressed in vessels that are co-opted by tumor cells. Figure 6A) Immunofluorescent image of GL261 tumor in wild-type mice showing CD93 expression in the vessels co-opted by tumor cells at the tumor edge (arrowheads). Dotted line indicates tumor border. Scale bar 40pm. Figure 6B) Immunofluorescent image of CD93 expression in the vessels co-opted by tumor cells (arrowheads) in the ex-vivo mouse brain slice model co-cultured with GL261 tumor cells. Dotted lines indicate tumor border. Scale bar 40pm.

[0026] Figure 7 shows that glioma cells migrate along the fibronectin fibers. Figure 7A) Immunofluorescent image of GL261 -mouse brain endothelial cell (ECs) co-culture showing GL261 cells migrating along the fibronectin fibers produced by the endothelial cells (arrowheads). Scale bar 100pm. Figure 7B) Immunofluorescent image showing GL261 cells migrating along the CD31 positive vessels expressing fibronectin (arrowheads) in a mouse brain slice model. Scale bar: 50pm. Figure 7C) Immunofluorescent image of mouse brain tumor showing GL261 cells migrating along vessels expressing fibronectin in the tumor edge (arrowheads). Scale bar 100pm. Figure 8 shows that CD93 deficiency is associated with reduced tumor cell invasion.

[0027] Figure 8A) Immunofluorescent image showing inhibited GL261 cell spreading when co-cultured in vitro on the CD93 deficient (CD93_ / ) brain endothelial cell monolayer compared to the wildtype condition. Scale bar 200pm. Figure 8B) Quantification of the area of the endothelial cell monolayer that is covered by GL261 cells showing a significant reduction of the GL261 spreading when co-cultured on the CD93 / _endothelial cell monolayer. **p<0.01 ; unpaired t-test. Figure 8C) Immunofluorescent image showing reduced invasion of the GL261 cells along the vessels when co-cultured as a spheroid on a wild-type or CD93 / _mouse brain slice. Scale bar 50pm. Figure 8D) Quantification of the number of GL261 sprouts along the vessels in wild-type and CD93 / _brain slices. Figure 8E) Invasion distance of GL261 cells along the vessels in wildtype and CD93 / _brain slices. *p<0.05; **p<0.01. Unpaired t-test. Figure 8F) H&E staining of the GL261 tumor edge in wild-type and CD93 / _mice showing inhibited tumor invasion in the CD93 / _group. Scale bar 100pm. Figure 8G) Quantification of the invasive tumor area in GL261 gliomas from wild-type and CD93 / _mice. **p<0.01. Unpaired t-test.

[0028] Figure 9 shows that CD93 deficiency reduces tumor cell proliferation. Figure 9A) Proliferation assay of U87 human glioma cells cultured with the conditioned medium derived from control (Mock and siCtrl) or CD93 downregulated (siCD93_1 and siCD93_2) human endothelial cells (HDBECs) showing that lack of soluble CD93 in the supernatant reduces U87 proliferation. **p<0.01 2way ANOVA with Dunnet multicomparison test. Figure 9B) Immunofluorescent image of the proliferation marker Ki67 in GL261 spheroids cultured on wildtype or CD93 / _brain slices. The image shows reduced Ki67 signal in the GL261 spheroid cultured on CD93 / _brain slices as compared to the wild-type brain slices. Scale bar: 100pm. Figure 9C) Quantification of Ki67-positive area normalized by the GL261 spheroid area. The graph is showing significant reduction of Ki67-positive GL261 cells in the CD93 / _group. **p<0.01 ; unpaired t-test.

[0029] Figure 10 shows Surface Plasmon Resonance sensorgrams of selected aCD93 antibodies. Binding of AD169-8, 85, 309 and 324 was measured by capturing the antibodies on a CM5 series S chip with immobilized ahuman Fab antibody mix. A single injection of 50 nM analyte (hCD93, hCD93lectin and mCD93) was performed. Following a dissociation phase, the surfaces were regenerated with 10 mM glycine-HCI, pH 2.1. SPR single cycle kinetics values are dissociation rate constant (KD) in nanomolar.

[0030] Figure 11 shows that CD93-blocking antibodies specifically bind to human endothelial cells expressing CD93. Human endothelial cells (HDBEC; which express CD93) and the human glioma cell line U87 (which does not express CD93) were incubated with aCD93 antibodies (human reactive and human / mouse cross-reactive). Graph is showing the percentage of cells positive for the akappa secondary antibody used to detect aCD93 antibodies by FACS analysis.

[0031] Figure 12 shows that CD93-blocking antibodies specifically bind to mouse endothelial cells expressing CD93. Mouse endothelial cells (MS1 ; which express CD93) and mouse GL261 glioma cells (which do not express CD93) were incubated with aCD93 antibodies (Human / Mouse cross-reactive). Graph is showing the percentage of cells labelled by the akappa secondary antibody used to detect CD93 antibodies by FACS analysis.

[0032] Figure 13 shows that CD93-blocking antibodies reduce fibronectin fibrillogenesis in human endothelial cells in vitro. Figure 13A) Immunofluorescent staining of fibronectin in human endothelial cells HDBECs after 24h treatment with 10pg / ml aCD93 antibodies or control IgG. Scale bar 20pm. Figure 13B) Quantification of the fibronectin covered area normalized to the total cell number. The graph shows the reduction of fibronectin area in cells treated with aCD93 antibodies as compared to control IgG.

[0033] Figure 14 shows that CD93-blocking antibodies reduce fibronectin fibrillogenesis in mouse endothelial cells in vitro. Figure 14A) Immunofluorescent staining of fibronectin in mouse endothelial cells MS1 after 24h treatment with 10pg / ml aCD93 antibodies or control IgG. Scale bar 20pm. Figure 14B) Quantification of the fibronectin covered area normalized to the total cell number. The graph shows the reduction of fibronectin area in cells treated with aCD93 antibodies as compared to the control IgG.

[0034] Figure 15 shows that CD93-blocking antibodies decrease vessel-associated fibronectin in mouse brain slices. Immunofluorescent staining of fibronectin in mouse brain slices treated with aCD93 antibodies or control IgG (10pg / ml during 24h). Images show fibronectin positive signal in the CD31 positive vessels in the brain slices treated with control IgG (arrowheads) and significant reduction of the signal in the brain slices treated with the aCD93 antibodies (arrowheads). Dotted line indicates GL261 tumor area. Scale bar 100pm.

[0035] Figure 16 shows that CD93-blocking antibodies disrupt human endothelial cell-cell junctions in vitro. Figure 16A) Immunofluorescent staining of actin in HDBEC monolayers after 8h treatment with 10pg / ml aCD93 antibodies or control IgG. Gaps between cells are indicated in white. Scale bar 20pm. Figure 16B) Quantification of the gap area normalized by the total cell number. The graph shows increase gap area between cells after aCD93 antibody treatment as compared to the control IgG.

[0036] Figure 17 shows that CD93-blocking antibodies disrupt mouse endothelial cell-cell junctions in vitro. Figure 17A) Immunofluorescent staining of actin showing MS1 monolayer after 8h treatment with 10pg / ml aCD93 antibodies or control IgG. Gaps between cells are indicated in

[0037] 6

[0038] RECTIFIED SHEET (RULE 91) ISA / EP white. Scale bar 20pm. Figure 17B) Quantification of the gap area normalized by the total cell number. The graph shows an increased gap area between cells after aCD93 antibodies treatment as compared to control IgG.

[0039] Figure 18 shows that CD93-blocking antibodies reduce human endothelial cell migration in vitro. Wound healing assay showing significant reduced migration of human endothelial cells (HDBEC) treated with 10pg / ml aCD93 antibodies compared to control IgG treated cells.

[0040] Figure 19 shows that CD93-blocking antibodies do not induce cell apoptosis in vitro. Figure 19A) Analysis of the apoptosis marker cleaved-Caspase 3 (cCasp3) on human endothelial cells (HDBECs) treated with 10pg / ml aCD93 antibodies or control IgG during 24h. Figure 19B) Analysis of the apoptosis marker cleaved-Caspase 3 (cCasp3) on mouse endothelial cells (MS1) treated with 10pg / ml aCD93 antibodies or control IgG during 24h.

[0041] Figure 20 shows that CD93-blocking antibodies inhibit migration of GL261 cells along the vessels in mouse brain slices. Figure 20A) Immunofluorescent staining of GL261 spheroids cultured on mouse brain slices in presence of 10pg / ml aCD93 antibodies or control IgG during 48h. Arrowheads show tumor cells co-opting vessels. Images show significant reduction of tumor co-opted vessels in brain slices treated with aCD93 antibodies as compared to control IgG. Figure 20B) Quantification of the number of tumor co-opted vessels in 100pm of GL261 spheroid border.

[0042] Figure 21 shows that CD93-blocking antibodies inhibit migration of U3013 cells along the vessels in human brain slices. Immunofluorescent staining of U3013 human glioma cells cultured on human brain slices in the presence of 10pg / ml aCD93 antibodies or control IgG during 48h. Arrowheads in Control IgG images indicate elongated tumor cells co-opting CD31 positive vessels. Arrowheads in aCD93 antibodies treated slices indicate aggregates of tumor cells showing a non-migrating phenotype.

[0043] Figure 22 shows in vivo distribution of aCD93 antibodies injected intravenously. GL261 bearing mice (Figure 22A, Figure 22B, Figure 22C) and healthy mice (Figure 22D) were injected with 5mg / Kg of aCD93 antibodies or control IgG intravenously 24h before collecting organs. aCD93 antibodies were detected by using anti-kappa AlexaFluor488 conjugated antibody. Vessels were visualised by CD31 staining. aCD93 antibodies signal was assessed in tumor core area (Figure 22A), tumor border (Figure 22B), tumor contralateral hemisphere (CLH) (Figure 22C), and in healthy mice (Figure 22D). Dotted line in (Figure 22B) indicates tumor edge and T indicate the tumor area. Arrowheads indicate vessels positive for anti-kappa antibody. Scale bar: 50pm. (Figure 22E) Quantification graph showing the % of anti-kappa signal relative to CD31 positive area in brain of GL261 bearing mice and healthy mice as well as in kidneys, liver and spleen.

[0044] Figure 23 shows that repeated doses of CD93-blocking antibodies administered systemically in GL261 bearing mice reduce the perivascular invasion of glioma cells. Figure 23A) Timeline of the in vivo tumor study (i.v., intravenous). Figure 23B) Representative immunofluorescent staining of the GL261 tumor border in mice that received the aCD93 antibody AD169-85 or AD169-309 or the control IgG. The nuclei are visualized by Hoechst staining, the vessels by CD31 and the tumor cells, GL261-GFP by GFP signal. The dotted line indicates the tumor invasive front, while the continuous line indicates the tumor bulk margin. Scale bar: 50pm. Figure 23C) Quantification graph shows the tumor invasion area corresponding to the area comprised between the continuous line (tumor bulk margin) and the dotted line (tumor invasion border). The data values represent arbitrary units (AU). *p<0.05; **p<0.01 , One-way ANOVA with Dunnet multicomparison test.

[0045] Figure 24 shows that repeated intravenous injections of CD93-blocking antibodies in GL261 bearing mice inhibit vessel-associated fibronectin. Figure 24A) Representative immunofluorescent staining of the GL261 tumors isolated from mice receiving aCD93 antibodies AD169-85 or AD169-309 or the control IgG. The images show the nuclei, visualized by Hoechst staining; the fibronectin and the vessels visualized by CD31. Scale bar: 50pm. Figure 24B) Quantification graph shows the percentage of fibronectin covered area relative the total CD31 area in the field of view. *p<0.05; ***p<0.001. One-way ANOVA with Dunnet multicomparison test.

[0046] Figure 25 shows that treatment with aCD93 blocking antibody AD169-309 in GL261 bearing mice increases VCAM1 expression in tumor vessels, promotes intratumor infiltration of CD8 and CD3 positive T-cells and increases survival in combination with aPD1 immunotherapy. Figure 25A) Representative immunofluorescent staining of tumor vasculature (visualized by CD31) and VCAM1 in mice treated with control IgG or with aCD93 antibodies AD169-85 or AD169-309. Scale bar 50um. Figure 25B) Quantification graph shows the percentage of VCAM1 covered area relative the total CD31 area. Figure 25C) Representative immunofluorescent staining of CD8 positive T-cells in control IgG, AD169-85 and in AD169-309 groups. Nuclei are visualized by Hoechst staining. Scale bar 50um. Figure 25D) Quantification graph shows the percentage of CD8 covered area relative the total nuclei area. Figure 25E) Immunofluorescent images showing CD3 positive T-cells. Scale bar 20um. Figure 25F) Quantification graph shows the percentage of CD3 covered area in the field of view. *p<0.05; ns=not significant. One-way ANOVA with Dunnet multicomparison test. Figure 25G) Survival curve of mice intracranially injected with GL261 glioma cells showing improved survival of AD169-309 treated mice compared to the Control IgG group after aPD1 -blockade immunotherapy (AD169-309+aPD1 vs Control lgG+aPD1). Rat lgG2 was used as an isotype IgG control for immunotherapy treatment. *p<0.05; ***p<0.005. Log-Rank Mantel-Cox test.

[0047] Figure 26 shows that treatment with aCD93 blocking antibody AD169-85 in GL261 bearing mice decreases the VE-cadherin levels in tumor vessels and increases leakage of endogenous fibrinogen. Figure 26A) Representative immunofluorescent staining of tumor vasculature (visualized by CD31) and VE-Cadherin in mice treated with control IgG or with aCD93 antibody AD169-85 or AD169-309. Scale bar 20um. Figure 26B) Quantification graph showing the percentage of VE-Cadherin covered area relative to the total CD31 area. Figure 26C) Immunofluorescent staining showing the signal of total endogenous fibrinogen (inside the vessels and extravasated in the surrounding tissue) and the extravasated fibrinogen (only the fibrinogen signal detected in the surrounding tissue excluding the signal overlapping with the vessel area). Vessels are visualized by CD31. Scale bar 25um. Figure 26D) Quantification graph showing the percentage of extravasated fibrinogen relative to the total tissue area. *p<0.05; **p<0.01 , One-way ANOVA with Dunnet multicomparison test.

[0048] Figure 27 shows the constructs of hCD93 used for determining the binding region of aCD93 antibodies.

[0049] The figures are not necessarily to scale, and generally only show parts that are necessary in order to elucidate the inventive concept, wherein other parts may be omitted or merely suggested.

[0050] DETAILED DESCRIPTION

[0051] The present disclosure relates to new agents, binding proteins or nucleic acid molecules encoding said agents / binding proteins, such as monoclonal antibodies or antigen binding fragments thereof, which selectively bind Cluster of Differentiation 93 (CD93) and inhibit perivascular tumor cell migration and cell invasion. The agents, binding proteins, and antibodies may be used in medical treatments, such as cancer therapies. The binding proteins may also be used for engineering cells to express a chimeric antigen receptor having a binding protein of the present disclosure as antigen binding domain.

[0052] The aim of the present disclosure is to provide new and enhanced binding agents specific for CD93, which also affect / impair the ability for tumor cells to migrate along vessels in the vasculature, invade into the normal brain tissue and proliferate, which may be used in therapy and cell engineering. Aspects of the present disclosure will be described more fully hereinafter with reference to the accompanying drawings. The agents, binding proteins, nucleic acids, and methods disclosed herein can, however, be realized in many different forms and should not be construed as being limited to the aspects set forth herein. Like numbers in the drawings refer to like elements throughout.

[0053] The terminology used herein is for the purpose of describing particular aspects of the disclosure only, and is not intended to limit the disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0054] In some embodiments a non-limiting term “agent” is used. The term “agent” is used herein to denote an entity that comprise a binding moiety, or encode a binding moiety, wherein the binding moiety binds CD93. Binding of the agent to CD93 typically provides an antagonistic effect. This may prevent functional activation of CD93 by its natural ligands, such as by preventing binding of the natural ligands to CD93. The binding moiety of the agent may be a binding protein, such as an antibody, or binding fragment thereof.

[0055] In some embodiments a non-limiting term “binding protein” is used. The term “binding protein” is used herein to denote a binding protein comprising a binding domain of an antibody (that is to say, a binding domain obtained or derived from an antibody, or based on a binding domain of an antibody). Thus, the binding protein is an antibody-based, or antibody-like, molecule comprising the binding site of, or a binding site derived from, an antibody. It is thus an immunological binding agent.

[0056] In some embodiments non-limiting terms “antibody” or “antigen binding fragment thereof’ is used. The term “antibody” is used herein in its broadest sense, including both monoclonal and polyclonal antibodies. As is well known, antibodies are immunoglobulin molecules capable of specific binding to a target (an antigen), such as a protein, carbohydrate, polynucleotide, lipid, polypeptide or other, through at least one antigen recognition site located in the variable region of the immunoglobulin molecule. As used herein, the term ’’antibody” or “an antigen binding fragment thereof” encompasses not only full-length or intact polyclonal or monoclonal antibodies, but also antigen-binding fragments thereof, such as Fab, Fab’, F(ab’)2, Fab3, Fv and variants thereof, fusion proteins comprising one or more antibody portions, humanized antibodies, chimeric antibodies, minibodies, diabodies, triabodies, tetrabodies, linear antibodies, single chain antibodies, multispecific antibodies (e.g. bispecific antibodies) and any other modified configuration of the immunoglobulin molecule that comprises an antigen recognition site of the required specificity, including glycosylation variants of antibodies, amino acid sequence variants of antibodies and covalently modified antibodies.

[0057] As is known to the skilled person, antibodies are proteins which comprise four polypeptide chains: two heavy chains and two light chains. Typically, the heavy chains are identical to each other, and the light chains are identical to each other. The light chains are shorter (and thus lighter) than the heavy chains. The heavy chains comprise four or five domains: at the N-terminus a variable (VH) domain is located, followed by three or four constant domains (from N-terminus to C-terminus CH1 , CH2, CH3 and, where present, CH4, respectively). The light chains comprise two domains: at the N-terminus a variable (VL) domain is located and at the C-terminus a constant (CL) domain is located. In the heavy chain an unstructured hinge region is located between the CH1 and CH2 domains. The two heavy chains of an antibody are joined by disulphide bonds formed between cysteine residues present in the hinge region, and each heavy chain is joined to one light chain by a disulphide bond between cysteine residues present in the CH1 and CL domains, respectively. In mammals two types of light chain are produced, known as lambda (A) and kappa (K). For kappa light chains, the variable and constant domains can be referred to as VKand CKdomains, respectively. Whether a light chain is a A or K light chain is determined by its constant region: the constant regions of A and K light chains differ, but are the same in all light chains of the same type in any given species. Depending on the amino acid sequence of the constant domain of its heavy chains, antibodies are assigned to different classes. There are six major classes of antibodies: IgA, IgD, IgE, IgG, IgM and IgY, and several of these may be further divided into subclasses, e.g., IgG 1 , lgG2, lgG3, lgG4, lgA1 and lgA2. The term ’’full-length antibody” as used herein, refers to an antibody of any class, such as IgD, IgE, IgG, IgA, IgM or IgY (or any sub-class thereof). The term ’’antigen binding fragment” refers to a portion or region of an antibody molecule, or a derivative thereof, that retains all or a significant part of the antigen binding of the corresponding full-length antibody. In some embodiments, the heavy chain of the antibodies may comprise VH+CH1+Hinge+CH2+CH3, and the light chain VL+CL. In preferred embodiments, the antibodies have the lgG1 LALA format, and the CH1 is defined by SEQ ID NO: 30, the CH2 is defined by SEQ ID NO: 31 , the CH3 is SEQ ID NO: 32, the CL is defined by SEQ ID NO: 33 and the hinge by SEQ ID NO: 34.

[0058] As briefly listed above, examples of antigen binding fragments include, but are not limited to: (1) a Fab fragment, which is a monovalent fragment having a VL-CL chain and a VH-CH chain; (2) a Fab’ fragment, which is a Fab fragment with the heavy chain hinge region, (3) a F(ab’)2 fragment, which is a dimer of Fab’ fragments joined by the heavy chain hinge region, for example linked by a disulfide bridge at the hinge region; (4) an Fc fragment; (5) an Fv fragment, which is the minimum antibody fragment having the VL and VH domains of a single arm of an antibody; (6) a single chain Fv (scFv) fragment, which is a single polypeptide chain in which the VHand VLdomains of an scFv are linked by a peptide linker; (7) an (scFv)2, which comprises two VH domains and two VL domains, which are associated through the two VH domains via disulfide bridges and (8) domain antibodies, which can be antibody single variable domain (VH or VL) polypeptides that specifically bind antigens. Antigen binding fragments can be prepared via routine methods. For example, F(ab’)2 fragments can be produced by pepsin digestion of a full- length antibody molecule, and Fab fragments can be generated by reducing the disulfide bridges of F(ab’)2 fragments. Alternatively, fragments can be prepared via recombinant technology by expressing the heavy and light chain fragments in suitable host cells (e.g., E. coli, yeast, mammalian, plant or insect cells) and having them assembled to form the desired antigen-binding fragments either in vivo or in vitro. A single-chain antibody can be prepared via recombinant technology by linking a nucleotide sequence coding for a heavy chain variable region and a nucleotide sequence coding for a light chain variable region. For example, a flexible linker may be incorporated between the two variable regions. Accordingly, throughout the description the generic terms “binding protein”, or “antibody” is used. These terms are used in their broadest sense and thus also incorporate all variants and fragments described above and below. In some embodiments the binding protein is a monoclonal antibody, or an antigen binding fragment selected from the group consisting of Fv fragments (e.g. single chain Fv and disulphide-bonded Fv), Fab-like fragments (e.g. Fab fragments, Fab' fragments and F(ab)2 fragments) and domain antibodies (e.g. single VH variable domains or VL variable domains).

[0059] Thus, the constant regions of the heavy chains are the same in all antibodies of any given isotype in a species, but differ between isotypes. The specificity of an antibody is determined by the sequence of its variable region. The sequence of variable regions varies between antibodies of the same type in any individual. In particular, both the light and heavy chains of an antibody comprise three hypervariable complementarity-determining regions (CDRs). In a pair of a light chain and a heavy chain, the CDRs of the two chains form the antigen-binding site. The CDR sequences determine the specificity of an antibody. A pair of a light chain variable region and a heavy chain variable region, comprising an (antigen) binding site, is known as an (antigen) binding domain. The three CDRs of a heavy chain are known as VHCDR1 , VHCDR2 and VHCDR3, from N-terminus to C-terminus, and the three CDRs of a light chain are known as VLCDR1 , VLCDR2 and VLCDR3, from N-terminus to C-terminus. In an antibody, as described above, the CDR sequences are located in the variable domains of the heavy and light chains. The CDR sequences sit within a polypeptide framework, which positions the CDRs appropriately for antigen binding. Thus, the remainder of the variable domains (i.e. the parts of the variable domain sequences which do not form a part of any one of the CDRs) constitute framework regions. The N-terminus of a mature variable domain forms framework region 1 (FR1); the polypeptide sequence between CDR1 and CDR2 forms FR2; the polypeptide sequence between CDR2 and CDR3 forms FR3; and the polypeptide sequence linking CDR3 to the constant domain forms FR4. In a binding protein of the invention the variable region framework regions may have any appropriate amino acid sequence such that the binding protein binds to CD93 via its CDRs.

[0060] If the binding protein is an antibody, the antibody may be of any isotype and sub-type. Thus, it may be an IgA, IgD, IgE, IgG, or IgM antibody. The heavy-chain constant domains that correspond to the different isotypes of immunoglobulins are termed a, 5, E, y and p, respectively. The subunit structures and three-dimensional configurations of different isotypes of immunoglobulins are well known. Preferably the antibody is an IgG antibody. As discussed above, there are four sub-types of IgG antibody: IgG 1 , I gG2 , lgG3 and lgG4. The IgG ocCD93 antibody of the invention may be of any IgG sub-type, i.e. it may be an lgG1 , lgG2, lgG3 or lgG4 antibody. In a preferred embodiment, the antibody is an lgG1 antibody, such as an lgG1 LALA antibody. In IgG 1 LALA the leucines (L) have been substituted for alanines (A) in amino acid positions 234 and 235 in the Fc region. The LALA mutation removes Fc mediated binding to the Fey receptor of immune cells which diminishes effector functions. Removing the binding thereby evades immune reactions, i.e. decreases immunogenicity mediated by Fc-effector functions such as ADCC and CDC and the risk of the biopharmaceutical causing unwanted off target and on target side effects.

[0061] Alternatively, the binding protein may be a binding fragment of an antibody (i.e. an antibody fragment), that is a fragment which retains the ability of the antibody to bind specifically to CD93. Such fragments are well-known, and examples include Fab’, Fab, F(ab’)2, Fv, Fd, or dAb fragments, which may be prepared according to techniques well known in the art.

[0062] A Fab fragment consists of the antigen binding domain of an antibody, i.e. an individual antibody may be seen to contain two Fab fragments, each consisting of a light chain and its conjoined N-terminal section of a heavy chain. Thus, a Fab fragment contains an entire light chain and the VHand CH1 domains of the heavy chain to which it is bound. Fab fragments may be obtained by digesting an antibody with papain. F(ab’)2fragments consist of the two Fab fragments of an antibody, plus the hinge regions of the heavy domains, including the disulphide bonds linking the two heavy chains together. In other words, a F(ab’)2fragment can be seen as two covalently joined Fab fragments. F(ab’)2fragments may be obtained by digesting an antibody with pepsin. Reduction of F(ab’)2fragments yield two Fab’ fragments, which can be seen as Fab fragments containing an additional sulfhydryl group which can be useful for conjugation of the fragment to other molecules.

[0063] Alternatively, the binding protein may be a synthetic or artificial construct, i.e. an antibody-like molecule which comprises a binding domain, but which is genetically engineered or artificially constructed. This includes chimeric or CDR-grafted antibodies, as well as single chain antibodies and other constructs, e.g. scFv, dsFv, ds-scFv, dimers, minibodies, diabodies, single domain antibodies (DABs), TandAbs dimers and heavy chain antibodies such as VHH, etc. In a particular embodiment the artificial construct is a single chain variable fragment (scFv). An scFv is a fusion protein in which a single polypeptide comprises both the VHand VLdomains of an antibody. scFv fragments generally include a peptide linker covalently joining the VH and VL regions, which contributes to the stability of the molecule. The linker may comprise from 1 to 20 amino acids, such as for example 1 , 2, 3 or 4 amino acids, 5, 10 or 15 amino acids, or other intermediate numbers in the range 1 to 20 as convenient. The peptide linker may be formed from any generally convenient amino acid residues, such as glycine and / or serine, as familiar to the person skilled in the art. However, it is not essential that a linker be present, and the VLdomain may be linked to the VHdomain by a peptide bond. An scFv typically comprises, N- terminal to C-terminal, a VHregion linked to a VLregion by a linker sequence. The preparation of scFv molecules is well known in the art.

[0064] A binding domain of an antibody is composed of a light chain variable domain and a heavy chain variable domain (a classical bivalent antibody has two binding domains). A binding protein may thus be a native antibody or a fragment thereof, or an artificial or synthetic antibody, or an antibody construct, or derivative (e.g. a single chain antibody, as discussed further below). In summary, the binding protein of the invention comprises a binding domain of an antibody, said binding domain of an antibody comprising a light chain variable domain and a heavy chain variable domain.

[0065] As used herein, the term ’’capable of binding X”, wherein X is an antigen, refers to a property of an antibody or binding fragment thereof which may be tested for example by ELISA, by use of surface plasmon resonance (SPR) technology, by use of the Kinetic Exclusion Assay (KinExA®) or by bio-layer interferometry (BLI). The skilled person is aware of said methods and others. The term “antagonistic”, as in “antagonistic agent” or “antagonistic antibody”, refers to an agent or a monoclonal antibody that block effector response (antagonistic effect) by binding to the receptor. Antagonistic antibodies mediate antagonist activity in a manner that blocks or dampen the binding of the natural ligand to the receptor.

[0066] The term “specificity”, sometimes referred to as “selectivity,” of the agent, binding moiety of the agent or binding protein, for a target refers to an agent, binding moiety or binding protein which will bind to the target with high affinity, but typically not to other antigens. A selective or specific binding agent / moiety / protein / antibody will not, or to a low extent, cross-react with other targets than the intended antigen. Thus, by binding “specifically” it is meant that e.g. the binding protein binds to its target (i.e. CD93) in a manner that can be distinguished from binding to nontarget molecules, more particularly that the binding protein binds its target (CD93) with greater binding affinity than with which it binds other molecules. That is, the binding protein does not bind to other, non-target, molecules, or does not do so to an appreciable or significant degree, or binds with lower affinity to such other molecules than with which it binds CD93. A binding protein “that specifically binds” CD93 may alternatively be referred to as “directed against” or “that recognises” CD93. In other words, CD93 is the antigen of the binding protein of the present invention, and the binding protein is thus an “antigen binding protein” in the sense that it binds CD93 as its antigen.

[0067] By “therapy” as used herein is meant the treatment of any medical condition. Such treatment may be prophylactic (i.e. preventative), curative (or treatment intended to be curative), or palliative (i.e. treatment designed merely to limit, relieve or improve the symptoms of a condition). Thus, “therapy” or “treating” of a disorder, such as cancer / cancer tumors, by means of a binding protein as used herein, are referring to preventing or ameliorating a certain disorder or medical condition, or to cure it. In the case of cancer and tumors, the treatments may shrink or abolish the present tumors, or they may halt or prevent the further spread of the tumors. An amount adequate to accomplish this is defined as a ’’therapeutically effective amount”. Effective amounts for a given purpose will depend on the disease or condition to be treated, its severity and the size / weight and general state of the subject. Thus, the agents or binding proteins as described herein may be used in the treatment / therapy of any condition in which the target antigen is expressed / overexpressed in a subject, such as in tumor vessels, to ameliorate said conditions, and may be administered systemically or locally, and by any suitable method known in the art. A subject, as defined herein, refers to any mammal, e.g. a farm animal such as a cow, horse, sheep, pig or goat, a pet animal such as a rabbit, cat or dog, or a primate such as a monkey, chimpanzee, gorilla or human. Most preferably the subject is a human being. By the term “immunotherapy” is referred to any type of treatment intended to stimulate the immune response against a tumor, i.e. targeting the immune system. That could include e.g. checkpoint blockade antibodies, vaccines, CAR-T cells, AAV vectors etc. A tumor targeting antibody is considered as a type of immunotherapy, but immunotherapy may also refer to the more common use of checkpoint blockade antibodies or similar ways to stimulate immune response.

[0068] Prophylactic treatment may include the prevention of a condition, or a delay in the development or onset of a condition. For example, the binding proteins may be used to prevent to prevent, delay or reduce the extent of a cancer developing, or recurring, or for example to prevent or reduce the extent of metastasis, by affecting tumor proliferation, tumor cell migration and perivascular invasion of tumor cells.

[0069] The present agents or binding proteins may, upon binding the target CD93, inhibit the ability of the tumor cells to sustain their blood supply, by e.g. mitigating vessel co-option in a subject. The present agents or binding proteins may, upon binding CD93, inhibit “tumor cell invasion”. With “invasion” is referred to the (peri)vascular migration of tumor cells along the abluminal side of the vasculature. This is illustrated in Figure 1 , where the effect of a CD93 binding agent generic on peri-vascular invasion of tumor cells is shown. On top is shown a nontreated invasive tumor, where tumor cell migration is also promoted by the interaction of tumor cells with proteins in the extracellular matrix, such as fibronectin, and below a tumor treated with an aCD93 antibody, inhibiting invasion of tumor cells in the vasculature in vicinity of the tumor cells. The present agents or binding proteins may, upon binding CD93, inhibit tumor cell migration. Glioma tumor cells have been shown to migrate along vessels, which is beneficial for their spread and for blood supply. Thus, “with tumor cell migration” is meant the migration of tumor cells along the abluminal side of the blood vessels in the vasculature in vicinity of the tumor. This migration is thus mitigated when the present agents / binding proteins are administered. The present agents or binding proteins may also, upon binding CD93, inhibit tumor cell proliferation, i.e. multiplication of the tumor cells by division. Proliferation of any tumor cells that grow by using vascular / vessel co-option (instead of new vessels / angiogenesis) will be affected.

[0070] The present disclosure also comprises nucleic acids encoding an agent or binding protein, which thus refers to a nucleic acid molecule which, when translated, will give rise to the agent, the binding moiety of the agent or the binding protein, or parts thereof, such as a binding domain of an antibody. The nucleic acid molecule may be in the form of mRNA to be expressed in mammalian cells, and may be is inserted into a mammalian cell using the CRISPR / Cas method. A “vector”, such as an expression vector, may comprise the nucleic acid molecule, wherein the vector may be a plasmid or a viral vector. Viral vectors are tools commonly used by molecular biologists to deliver genetic material into cells. This process can be performed inside a living organism (in vivo) or in cell culture (in vitro). Delivery of genes or other genetic material by a vector is termed transduction and the infected cells are described as transduced. In some embodiments, the viral vector(s) is comprised in a “host cell”, which is a cell capable of translation of the nucleic acid molecule comprised in the vector.

[0071] All solid tumors require a vascular supply in order to progress. Although the ability to induce angiogenesis (new blood vessel growth) has been seen as important to this purpose, it has also been shown that tumors can grow instead by hijacking pre-existing blood vessels of the surrounding non-malignant tissue, a process called vessel co-option. Vessel co-option is thus an important mechanism of tumor vascularization that can influence disease progression, metastasis and response to treatment. Thus, methods for inhibiting said process could give rise to effective treatment of invasive cancers that use said process.

[0072] Cluster of Differentiation 93 (CD93) is a C-type lectin transmembrane receptor which plays a role not only in cell-cell adhesion processes but also in host defense. CD93 is expressed by a wide variety of cells such as platelets, monocytes, microglia and endothelial cells. CD93 is also expressed on neutrophils, activated macrophages, B cell precursors, and a subset of dendritic cells and of natural killer cells in the immune system. CD93 is involved in endothelial cell-cell adhesion, cell spreading, cell migration, cell polarization as well as tubular morphogenesis. Recently it has been found that CD93 is able to control endothelial cell dynamics through its interaction with an extracellular matrix glycoprotein MMRN2.

[0073] CD93 is upregulated in the tumor vessels of glioblastoma patients and has been found to be of importance in glioma development. CD93 affects the ability of endothelial cells to regulate their cytoskeleton, and is important for blood vessel formation, angiogenesis. CD93 is highly expressed in glioma vessels and high levels of vascular CD93 is associated with poor patient survival. CD93 knockout mice with glioma show smaller tumor size and improved survival. It has been shown that it is possible to block angiogenesis in vitro either by knockdown of CD93 using siRNA or shRNA or through treatment with a neutralizing antibody that disrupts the CD93- MMRN2 complex, murine antibody 4E1 (M. Orlandini et al. Oncotarget 2014 May 15; 5(9):2750- 60). It has further been shown that CD93 interacts with IGFBP7 (US2022 / 0235136), and that CD93 interacts with MMRN2 and organizes fibronectin in extracellular matrix through regulating the activity of ct5p1 integrin. Thus, targeting MMRN2 (WO2018 / 020222) and activity of ct5p1 integrin has been an aim in the field. In spite of all this progress, efficient therapies of glioma are still lacking. Glioma is hard to treat partly due to being a very invasive type of cancer, and thus it was hypothesized that agents affecting its invasiveness would provide treatment benefits. Peri-vascular invasion is one of the route used by tumor cells to spread into the brain, and extracellular matrix, including fibronectin, is essential for tumor cell migration and invasion. It was shown that CD93-MMRN2 complex promotes the formation of perivascular fibronectin (FN) fibers and that knockout of CD93 disrupt perivascular FN fibers. Surprisingly, current data from the inventors show that CD93 knockout or downregulation inhibits tumor invasiveness and tumor cell proliferation, and that the inhibition of tumor invasiveness in response to CD93 deficiency is associated with less peri-vascular fibronectin fibers. Initially, it was found that CD93 has a role in regulating invasion of glioma cells, that murine glioma cells can migrate along vessels in wild-type mice, but not CD93 / _mice, and it was interpreted that antibodies or peptides that inhibit ct5|31 -integrin could decrease fibronectin fibrillogenesis and glioma invasion. However, it was found that even if CD93 is required for peri-vascular invasion of tumor cells, a role independent of the suggested complex with a5pi integrin seems to exist.

[0074] It has also been found that a neutralizing antibody to CD93 can disrupt endothelial junctions and already formed fibronectin fibers. This data shows that a neutralizing or antagonistic antibody to CD93 not only inhibits the active process of angiogenesis, but actually breaks endothelial junctions and leads to retraction of the fibronectin network. In addition, it was found that leukocyte adhesion molecules expression was enhanced in tumor vessels and T cell abundance was increased in gliomas in CD93 / _mice. Consistent with this, CD93-deficiency was associated with improved response to aPD1 immunotherapy. Thus, new data presented by the current inventors is also suggesting a previously unknown function of CD93 in regulating endothelial activation and expression of adhesion molecules required for recruitment of immune cells to the glioma tumor microenvironment. Thus, CD93 can also be a target to enhance cancer immunotherapy.

[0075] The fact that an antibody to CD93 can break up junctions that are already formed and alter remodeling of extracellular matrix is surprising. CD93 is up-regulated during active angiogenesis, and its role in maintaining endothelial junctions is unexpected. Importantly, this may suggest that a neutralizing antibody to CD93 may have a significant effect on vessel permeability, and may e.g. be used together with other cancers drugs, such as temozolomide.

[0076] It was thus concluded that agents targeting vascular CD93 and that block cell invasion and proliferation by inhibiting perivascular cell migration would be beneficial in the treatment of cancers / tumors. An agent that blocks the glioma cell invasion and proliferation by targeting vascular CD93 could be used in the treatment of glioma. The agent may e.g. be a specific blocking antibody, blocking the function of human endothelial CD93, with the above mentioned properties.

[0077] Accordingly, 27 CD93 targeting mAb were generated (human (Hu) or Murine (Mu) or Human / Murine (Hu / Mu)) (Table 1) and 26 of them successfully converted in hlgG1 LALA (Table 2) and tested for their ability to block tumor perivascular invasion. Surprisingly, based on in vitro and ex-vivo assays, it was found that only 4 antibodies (3 Mouse(Ms) / Human (Hu) and 1 Hu) exhibited the desired properties, as illustrated in Table 3.

[0078] Table 1

[0079]

[0080] Table 1 shows the results from scFv and Fab clone interaction studies. Binding studies of the most promising purified scFv and Fab clones to hCD93 (H), hCD93lectin (L), mCD93 (M) and a non-relevant control protein (C). For ELISA and HTRF (Homogeneous Time Resolved Fluorescence) the colour coding is: black: maximum or near maximum binding signal; striped: weak binding signal; white: no binding. SPR single cycle kinetics values are dissociation rate constant (KD) in nanomolar. Flow cytometry (FACS) data shows the percentage (%) of cells positive for the different clones. Human endothelial cells (H) and mouse endothelial cells (M). Hu Neg (human U87 cells, negative control); M Neg (mouse GL261 cells, negative control). Interaction studies of converted aCD93 antibodies were performed, to assess the bindings properties, as illustrated in Table 2 below.

[0081] Table 2

[0082] Shown in Table 2 above is binding studies after conversion of the most promising purified scFv and Fab clones to hlgG1 LALA. Binding to hCD93 (H), hCD93lectin (L), mCD93 (M) and a non- relevant control protein (C) was measured by ELISA and SPR. For ELISA the colour coding is: black: maximum or near maximum binding signal; striped: weak binding signal; white: no binding. SPR single cycle kinetics values are dissociation rate constant (KD) in nanomolar. Flow cytometry (FACS) data shows the percentage (%) of cells positive for the different clones. Human endothelial cells (H) and mouse endothelial cells (M). Hu Neg (human U87 cells, negative control); M Neg (mouse GL261 cells, negative control). Studies of aCD903 antibodies functional properties where performed, as illustrated in Table 3 below, to show which antibodies that have the desired functionality / properties / effects.

[0083] Table 3

[0084] Twenty-six aCD93 antibodies (10|jg / ml) were tested for their functional effect by using in vitro and ex vivo models. For in vitro assays, human and mouse endothelial cells (ECs) were used. Intercellular gaps formation (Gaps), fibronectin deposition (FN) and cell migration (Migr.) were assessed. For ex vivo assays, co-culture of glioma cells on human and mouse brain slices models were used. Perivascular invasion of tumor cells (PV invasion) and perivascular deposition of fibronectin (PV FN deposition) were assessed. Scoring criteria and symbols used to assess aCD93 antibodies effect in different assays are the follow: for “Gaps”, increased fold change relative to control isotype IgG was calculated: x (2-4 fold change); xx (4-8 fold change); xxx (>8 fold change). For “FN”, % of fibronectin reduction relative to control isotype IgG was measured: x (20-40% reduction); xx (40-60% reduction); xxx (>60% reduction). For EC migration, % of inhibition relative to isotype control IgG was measured: x (20-40% inhibition); xx (40-60% inhibition); xxx (>60% inhibition). For “PV invasion” and “PV FN deposition”: x (weak inhibition); xx (partial inhibition); xxx (strong inhibition). - (no effect); nd (not tested). Four antibodies were identified as having the desired effect(s), these being referred to as AD169-85, AD169-309, AD169-324, and AD169-8, as illustrated in top of Table 3. As shown in Table 2, the KDof these antibodies range from 2-21 nM.

[0085] The resulting four binding proteins / mAb (AD169-85, AD169-309, AD169-324, and AD169-8) all bind vascular CD93 with a high affinity, having a KD<21 nM, and all show an inhibiting effect on tumor cell proliferation and tumor (glioma) cell invasion, resulting in inhibition of perivascular tumor cell migration, i.e. migration of tumor cells along the vasculature. Thus, the administration of the four binding proteins to a subject, and their subsequent binding to vascular CD93, results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation. Thus, vessel co-option, migration and proliferation of the tumor cells may be mitigated / inhibited using the present binding molecules, or nucleic acid molecules encoding said binding molecules. These effects mediated by the binding proteins results in increased survival of mice with cancer, as shown in the example section below. The structures of the four binding molecules (antibodies) also show a common CDR motif, as defined below. They are also all of human origin, as compared to murine or humanized.

[0086] Accordingly is provided herein binding proteins that specifically bind vascular CD93 and inhibiting perivascular tumor cell migration. The binding proteins comprise light and heavy chain variable domains. The light and heavy chain variable domains comprise 3 CDRs each: the light chain variable domain comprises VLCDR1 , VLCDR2 and VLCDR3, and the heavy chain variable domain comprises VHCDR1 , VHCDR2 and VHCDR3. The six CDRs have the following amino acid sequences:

[0087] VHCDR1 as defined by SEQ ID NO: 1 ;

[0088] VHCDR2 as defined by SEQ ID NO: 2;

[0089] VHCDR3 as defined by SEQ ID NO: 3;

[0090] VLCDR1 as defined by SEQ ID NO: 4;

[0091] VLCDR2 defined as AAS;

[0092] VLCDR3 as defined by SEQ ID NO: 5. wherein VLCDR1 and VLCDR2 are identical in all 4 binding protein variants, while VHCDR1 , VHCDR2, VHCDR3, and VLCDR3 comprises some variations, as indicated in Table 4 below. Besides the CDRs according to SEQ ID NO: 1-5 and AAS, the invention also encompasses CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto that also show the same effect upon tumor proliferation, cell invasion and tumor cell migrations upon binding of CD93.

[0093] Table 4

[0094] The sequences of the CDRs of the binding proteins are shown in Table 4 in one letter amino acid code, where VHCDR1 , VHCDR2, VHCDR3 and VLCDR3 comprises variations in the sequence between the 4 binding protein variants, as indicated by X. In some cases, the sequence of one of the CDRs comprises fewer residues than the motif, where the X in case no residue exists in said position is indicated as a “gap” (i.e. none). In VHCDR1 , X5may be G, S or Y, Xe may be S or Y, X7may be Y or S, and X8may be S or G. In VHCDR2 X2may be S or Y, X3 may be G, S or Y, X4may be S, Y, or G, X7may be S, G or Y. In VHCDR3 X3may be Y, S or P, X4may be G, D, or S, X5may be W, Y, or G, X6may be T, or gap, X7may be Y, or gap, X8may be P, or gap, X9may be D, Y, or gap, X10 may be Y, V, or G, Xn may be I, L, or F. In VLCDR3 X3may be S or R, X5may be S, or gap, X6may be Y, T, or F, X8may be P, or Y.

[0095] In some aspects, some CDRs are surrounded by specific amino acids, referred to as framework amino acids (faa), which are conserved between the different binding proteins, with some small variations. Accordingly, the VHCDR1 and VHCDR2 of the binding protein may be present next to specific framework amino acids, wherein the CDR and framework amino acid sequences are selected from the group comprising:

[0096] VHCDR1 and faa defined by SEQ ID NO: 6;

[0097] VHCDR2 and faa as defined by SEQ ID NO: 7, as indicated in Table 5 below. Besides the CDRs+faa according to SEQ ID NO: 6-7, the invention also encompasses CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto that also show the same effect when binding CD93. Table 5

[0098] In VHCDR1+faa, X10 may be S or Y, where the remaining unknowns correspond to the sequences of Table 4 above. In VHCDR2+faa, there is inserted a faa at the beginning and the end, where the first one, Xi may be A or G, while the remaining unknowns correspond to the sequences of Table 4 above but with altered numbering due to the inserted amino acid residue at the beginning, such that X3corresponds to X2of Table 4, etc.

[0099] In some aspects, 4 different binding proteins are presented, having the individual CDR combinations as presented in Tables 6-7 below, where Table 6 shows the CDRs H1 , H2 and H3 and Table 7 shows the CDRs L1 , L2 and L3, for the respective 4 different binding proteins.

[0100] Table 6

[0101] Table 7

[0102] In some aspects, 4 different binding proteins are presented, having the individual CDR combinations as described above, and surrounding framework amino acids for VHCDR1 and VHCDR2 as presented in Table 8 below. Table 8

[0103] The binding proteins of the invention may be synthesised by any method known in the art. Preferably, the binding proteins are synthesised using a protein expression system, such as a cellular expression system using prokaryotic (e.g. bacterial) cells or eukaryotic (e.g. yeast, fungus, insect or mammalian) cells. An alternative protein expression system is a cell-free, in vitro expression system, in which a nucleotide sequence encoding the binding protein is transcribed into mRNA, and the mRNA translated into a protein, in vitro. Cell-free expression system kits are widely available, and can be purchased from e.g. Thermo Fisher Scientific. Alternatively, binding proteins may be chemically synthesised in a non-biological system. Liquidphase synthesis or solid-phase synthesis may be used to generate polypeptides which may form or be comprised within the binding protein of the invention.

[0104] The skilled person can readily produce binding proteins using appropriate methodology common in the art. In particular, the binding proteins may be recombinantly expressed in mammalian cells, such as CHO cells. A binding protein synthesised in a protein expression system may be purified using standard techniques in the art, e.g. it may be synthesised with an affinity tag and purified by affinity chromatography. If the binding protein is an antibody, it can be purified using affinity chromatography using one or more antibody-binding proteins, such as Protein G, Protein A, Protein A / G or Protein L.

[0105] As noted above, the binding proteins are typically antibody-based, or antibody-like, molecules. Thus, a binding protein may be a native antibody or a fragment thereof, or an artificial or synthetic antibody, or an antibody construct or derivative (e.g. a single chain antibody). In a preferred embodiment, the binding protein is a human protein (of human origin as compared to humanized), in particular a human monoclonal antibody, antibody fragment or scFv. A human binding protein may comprise VH and VL regions in which both framework and CDR regions are derived from human germline immunoglobulin sequences, and also a human constant region, if a constant region is contained in the protein. Such proteins may however include amino acids not encoded by human germline Ig sequences, for example mutations introduced by random or site-specific mutagenesis.

[0106] As detailed above, the binding proteins of the invention comprises a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (or variable region) and a light chain variable domain. Thus, in a particular embodiment the binding protein of the invention comprises:

[0107] (i) a heavy chain variable domain (VH) comprising (or consisting of) the amino acid sequence set forth in any one of SEQ ID NO: 22-25, or a variant thereof; and

[0108] (ii) a light chain variable domain (VL) comprising (or consisting of) the amino acid sequence set forth in any one of SEQ ID NO: 26-29, or a variant thereof. The variant may define sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, wherein the CDR sequences comprise no variations in the amino acid sequence, or wherein the sequence variation of the CDR amino acid sequences is at most 5%, such as 4 %, 3 %, 2 %, 1 % or less. Thus, the defined variants as sequences having 80 % or more, such as 85 %, 90 %, 95 % or more identity thereto, comes with the proviso that the CDR sequences of the variants are unaltered in view of the antibody variant defined by the VH or VL domain, such that sequence variation of the CDR amino acid sequences is at most 5 %, such as 4 %, 3 %, 2 %, 1 % or less, such that the sequence that form the CDRs have at least 95 %, such as 96 %, 97 %, 98 %, 99 % identity thereto. Binding proteins with variants of the sequences of the variable and / or constant domains are functional variants, having the activities described above (i.e. they specifically bind CD93 and inhibit one or more of tumor cell proliferation, invasion and migration). Variant sequences may be modified relative to the native sequences by substitution, insertion and / or deletion of one or more amino acids.

[0109] Sequence identity may be assessed by any convenient method. However, for determining the degree of sequence identity between sequences, computer programmes that make pairwise or multiple alignments of sequences are useful, for instance EMBOSS Needle or EMBOSS stretcher (both Rice, P. et al., Trends Genet., 16, (6) pp276 — 277, 2000) may be used for pairwise sequence alignments while Clustal Omega (Sievers F et al., Mol. Syst. Biol. 7:539, 2011) or MUSCLE (Edgar, R.C., Nucleic Acids Res. 32(5): 1792-1797, 2004) may be used for multiple sequence alignments, though any other appropriate programme may be used. Whether the alignment is pairwise or multiple, it must be performed globally (i.e. across the entirety of the reference sequence) rather than locally. Sequence alignments and % identity calculations may be determined using for instance standard Clustal Omega parameters: matrix Gonnet, gap opening penalty 6, gap extension penalty 1. Alternatively, the standard EMBOSS Needle parameters may be used: matrix BLOSUM62, gap opening penalty 10, gap extension penalty 0.5. Any other suitable parameters may alternatively be used.

[0110] In some embodiments, the binding proteins are antibodies that comprises a heavy chains and a light chains of comprising variable and constant regions, wherein the binding protein comprises a heavy chain and a light chain, the heavy chain comprising the VH domain and three constant domains, CH1 , CH2, and CH3, wherein CH1 and CH2 are joined via a hinge region, and the light comprising the VL domain and constant domain, CL. The binding protein may be a monoclonal antibody of the lgG1 isotype, such as an lgG1 LALA antibody, wherein the amino acid sequences of the domains and hinge region are: CH1 as defined by SEQ ID NO: 30, CH2 as defined by SEQ ID NO: 31 , CH3 as defined by SEQ ID NO: 32, CL as defined by SEQ ID NO: 33, Hinge region as defined by SEQ ID NO: 34.

[0111] The binding protein may be a monoclonal antibody, or an antigen-binding fragment selected from the group consisting of Fv fragments, Fab-like fragments, disulphide-bonded fragments and domain antibodies, wherein the Fv fragment may be an scFv fragment and the Fab-like fragment may be a Fab or F(ab’)2 fragment.

[0112] The binding proteins have been shown to bind CD93 with a high affinity, and may thus be used as a stand-alone cancer therapeutic as described above. The binding proteins of the present disclosure may also be used for designing chimeric antigen receptors (CARs) against CD93 to obtain CD93-targeted CAR cells, such as CAR-T cells. It is also possible that the antibody would be produced as a separate part in a CAR, to bring the T cells to the tumor, i.e. , by designing a cell engineered to express a CAR wherein the antibodies or antigen binding domains are produced in the cell to direct it towards tumor vessels, via binding of CD93.

[0113] These CD93-targeted CAR T-cells can then be used in therapy to bind CD93 expressing cells in the tumor vessel and eliminate surrounding malignant cells or to eliminate CD93 expressing cells, such as cancer cells. For example, a CAR T-cell is made by isolating T cells from a subject, inserting a gene for the CAR in the T-cells to create a CAR T-cell expressing a CAR protein, where the CARs are hybrids of T-cell and antibody receptors comprising 4 distinct regions; an extracellular domain which recognizes the antigen (typically an scFv fragment of an antibody) connected to a transmembrane domain by a hinge (spacer), where the transmembrane domain has a hydrophobic alpha-helix structure, and wherein the transmembrane domain is connected to an endodomain (intracellular domain), which undergoes conformational changes following antigen recognition, which triggers downstream signalling pathways to induce immune responses. The endodomain may also comprise one or more costimulatory domains to enhance the anti-tumor activity. Thus, the present disclosure provides cell engineered to express a chimeric antigen receptor (CAR), wherein the CAR comprises an antigen-binding domain, a transmembrane domain connected to the antigen binding domain by a hinge region, and an intracellular domain optionally connected to one or more co-stimulatory domains, wherein the antigen binding domain comprises an scFv fragment of the binding protein. The cell may be a human cell, an immune effector cell, such as a T cell, an NK cell or a macrophage. Thus, an aspect of the present disclosure comprises a cell engineered to express a chimeric antigen receptor (CAR), wherein the CAR comprises an antigen-binding domain, a transmembrane domain connected to the antigen binding domain by a hinge region, and an intracellular domain optionally connected to one or more co-stimulatory domains, wherein the antigen binding domain comprises the binding protein of the present disclosure, wherein the cell may be a human cell. The cell may also be an immune effector cell, such as a T cell, an NK cell or a macrophage.

[0114] Further provided herein is a nucleic acid molecule encoding an agent or binding protein of the present disclosure. The nucleic acid molecule may encode a binding protein comprising a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (VH) and a light chain variable domain (VL), according to the present disclosure. Thus, the agents or binding proteins may be administered in protein form, or they may be administered in nucleic acid form to be expressed inside the body of a subject. Thus, any nucleotide sequence translating into any one of the agents or binding proteins herein may be used. The nucleic acid molecule comprises a nucleotide sequence in the form of mRNA to be expressed in mammalian cells. In some embodiments, a nucleotide sequence of the nucleic acid molecule is inserted into a mammalian cell using the CRISPR / Cas method. The nucleic acid molecule may also be comprised in a vector, such as an expression vector. The vector may be a plasmid or a viral vector. The vector may be transfected into an isolated host cell for translation of the nucleic acid.

[0115] Also provided herein are an agents, which may comprise a binding moiety that specifically binds to vascular CD93. The binding of the agent to CD93 may provide an antagonistic effect, such as by preventing functional activation of CD93 by its natural ligands. The agent, or the binding moiety thereof, may be a binding protein, or encode a binding protein, wherein the encoded binding protein binds CD93.

[0116] The present invention provides new enhanced binding proteins, which bind human vascular CD93. The binding proteins of the invention has been evaluated for binding affinity and therapeutic effect, and it has been determined that they provide enhanced performance compared to other binding proteins in the art, as they provide the effect or inhibited tumor invasiveness. As discussed above, several CD93 binding antibodies were made, but only 4 exhibited the desired properties of inhibiting glioma invasiveness. Thus, it is demonstrated that binding CD93 is not enough to mediate this effect, but that some other interaction, such as steric hindrance or other, may be present. Accordingly, the present disclosure relates to agents, such as binding proteins, which specifically bind CD93, and which binding results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation. One important factor is that the binding will affect the tumor cell migration, without damaging / killing the endothelial cells. As vessel co-option may, besides angiogenesis, play an important role in tumor growth, inhibiting said mechanism provides new possibilities for treatment.

[0117] The present disclosure relates to agents comprising a binding moiety that specifically binds to vascular CD93, wherein the agent or binding moiety thereof is a binding protein. Binding of the agent to CD93 may provide an antagonistic effect by preventing functional activation of CD93 by its natural ligands. It is hypothesized that this may lead to that interaction with MMRN2 is altered, as blockage of fibronectin fibrillogenesis is seen. The agents or binding proteins may inhibit the ability of the tumor cells to sustain their blood supply, by e.g. mitigating vessel cooption. Thus, binding of said agent or binding protein to CD93 results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation.

[0118] One object of the present disclosure is to provide novel enhanced tumor targeting agents for improved therapy, particularly cancer therapy. The binding proteins are introduced into the body by various means (such as injection or ingestion) localize the specific antigen (CD93), bind the receptor, stay on cell surface or become internalized into the cell.

[0119] The binding proteins of the present disclosure may be used to treat a number of cancers wherein vessel co-option is seen, and / or which are linked to expression / overexpression of the present target antigen, CD93, as discussed above and below. Any types of cancer where CD93 is expressed in the tumor vessels / vasculature of the cancer may be targeted. Particularly, subject suffering from brain cancer, especially glioma, including glioblastoma, could benefit enormously from these treatments, as often no efficacious treatments are available. As discussed above, glioma are very invasive tumors, where complete surgical resection of the tumor is not possible because of its invasiveness. Thus, being able to reduce this invasiveness provides an efficient way of treating these tumors, as is shown in the examples below.

[0120] In a further aspect, the invention provides a pharmaceutical composition comprising an agent or binding protein of the invention, as described above, or a nucleic acid molecule encoding said agent or binding protein, as described above, or an engineered CAR-cell, as described above. In addition, the pharmaceutical composition also comprises at least one pharmaceutically acceptable carrier or excipient. As used herein, ’’pharmaceutically acceptable carrier or excipient” includes any and all solvents, dispersion media, coatings, antibacterial and antifungal agents, isotonic and absorption delaying agents and the like that are physiologically compatible.

[0121] Preferably, the carrier or excipient is suitable for parenteral, e.g. intradermal, intravenous, intramuscular or subcutaneous administration (e.g. by injection or infusion). Depending on the route of administration, the binding protein or conjugated binding protein may be coated in a material to protect them from the action of acids and other natural conditions that may inactivate or denature it. Preferred pharmaceutically-acceptable carriers comprise aqueous carriers or diluents. Examples of suitable aqueous carriers that may be employed in the pharmaceutical compositions, kits and products include water, buffered water and saline. Examples of other carriers include ethanol, polyols (such as glycerol, propylene glycol, polyethylene glycol, and the like), and suitable mixtures thereof, vegetable oils and injectable organic esters, such as ethyl oleate. Proper fluidity can be maintained, for example, by the use of coating materials, such as lecithin, by the maintenance of the required particle size in the case of dispersions, and by the use of surfactants. In many cases, it will be preferable to include isotonic agents, for example, sugars, polyalcohols such as mannitol, sorbitol, sodium chloride and the like.

[0122] The present disclosure thus relates to an agent, binding protein, cell, nucleic acid molecule or pharmaceutical composition as described herein, for use in therapy, such as cancer therapy. The agent, binding protein, cell, or pharmaceutical composition thereof may be for use in inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and / or iii) inhibition of tumor cell proliferation, and / or for use in combination with one or more additional treatment regime, wherein the additional treatment regime is selected from i) use of antineoplastic agents, i.e. agents known to be able to treat cancer by the skilled person, ii) chemotherapy, iii) immunotherapy, including tumor targeting antibodies, iv) irradiation and v) use of pathway blockers for treating cancer, wherein pathway blockers may include tyrosine kinase inhibitors or neutralizing antibodies towards growth factors / cytokines or their receptors.

[0123] Thus, herein is described a method of treating a subject in need thereof, comprising administering a therapeutically effective amount of an agent, a binding protein, a cell or a nucleic acid molecule, as described herein. Also, use of an agent, a binding protein, a cell, a nucleic acid molecule, or a pharmaceutical composition, for the treatment or prevention of cancer, preferably glioblastoma, is provided. The agents, binding proteins, nucleic acid molecules, cells, or pharmaceutical compositions thereof may be administered via one or more routes of administration using one or more of a variety of methods known in the art. As will be appreciated by the skilled artisan, the route and / or mode of administration will vary depending upon the desired results. Preferred routes of administration include intravenous, intramuscular, intradermal, intraperitoneal, subcutaneous, spinal or other parenteral routes of administration, for example by injection or infusion, e.g. directly to the site of a tumor. The phrase ’’parenteral administration” as used herein means modes of administration other than enteral and topical administration, usually by injection. Alternatively, a non-parenteral route may be used, such as a topical, epidermal or mucosal route of administration. Local administration is preferred, including peritumoral, juxtatumoral, intratumoral, intralesional, perilesional, intra cavity infusion, intravesicle administration, and inhalation. However, the antigen binding protein, conjugate or engineered cells may also be administered systemically.

[0124] A suitable dosage of a specific binding protein, conjugated binding protein, or pharmaceutical composition of the invention may be determined by a skilled medical practitioner. Actual dosage levels of the active ingredients in the pharmaceutical compositions and products of the present invention may be varied so as to obtain an amount of the active ingredient which is effective to achieve the desired therapeutic response for a particular subject, i.e. patient, without being toxic to the patient. The selected dosage level will depend upon a variety of pharmacokinetic factors including the activity of the particular protein / conjugate employed, the route of administration, the time of administration, the rate of excretion of the protein, the duration of the treatment, other drugs, compounds and / or materials used in combination with the particular compositions employed, the age, sex, weight, condition, general health and prior medical history of the patient being treated, and like factors well known in the medical arts.

[0125] Doses may be given at certain intervals, such as bi-weekly (every fortnight), or shorter intervals, such as daily doses, may be more used if suitable. Suitable intervals for different types of treatments would be apparent to the skilled person.

[0126] Dosage regimens may be adjusted to provide the optimum desired response (e.g. a therapeutic response). For example, a single bolus may be administered, several divided doses may be administered over time, or the dose may be proportionally reduced or increased as indicated by the exigencies of the therapeutic situation. It is especially advantageous to formulate parenteral compositions in dosage unit form for ease of administration and uniformity of dosage. Dosage unit form as used herein refers to physically discrete units suited as unitary dosages for the subjects to be treated; each unit contains a predetermined quantity of active compound calculated to produce the desired therapeutic effect in association with the required pharmaceutical carrier.

[0127] The agent, binding protein or composition may be administered in a single dose or in multiple doses. The multiple doses may be administered via the same or different routes and to the same or different locations. Alternatively, they can be administered as a sustained release formulation, in which case less frequent administration is required. Dosage and frequency may vary depending on the half-life of the administered species in the patient and the duration of treatment that is desired. The dosage and frequency of administration can also vary depending on whether the treatment is prophylactic or therapeutic. In prophylactic applications, a relatively low dosage may be administered at relatively infrequent intervals over a long period of time. In therapeutic applications, a relatively high dosage may be administered, for example until the patient shows partial or complete amelioration of symptoms of disease. In an exemplary dosage regime, the conjugated binding protein is administered to the subject once a week, once a fortnight or once every three weeks, in a cycle repeated from 2 to 10 times.

[0128] The content of this disclosure thus enables treatment of disorders linked to tumor invasiveness and CD93 expression, such as cancer, by administering binding proteins of the present invention. In the drawings and specification, there have been disclosed exemplary aspects of the disclosure. However, many variations and modifications can be made to these aspects without substantially departing from the principles of the present disclosure. Thus, the disclosure should be regarded as illustrative rather than restrictive, and not as being limited to the particular aspects discussed above. Accordingly, although specific terms are employed, they are used in a generic and descriptive sense only and not for purposes of limitation.

[0129] The description of the example embodiments provided herein have been presented for purposes of illustration. The description is not intended to be exhaustive or to limit example embodiments to the precise form disclosed, and modifications and variations are possible in light of the above teachings or may be acquired from practice of various alternatives to the provided embodiments. The examples discussed herein were chosen and described in order to explain the principles and the nature of various example embodiments and its practical application to enable one skilled in the art to utilize the example embodiments in various manners and with various modifications as are suited to the particular use contemplated. The features of the embodiments described herein may be combined in all possible combinations of methods, products, and systems. It should be appreciated that the example embodiments presented herein may be practiced in any combination with each other. It should also be noted that the word “comprising” does not necessarily exclude the presence of other elements or steps than those listed and the words “a” or “an” preceding an element do not exclude the presence of a plurality of such elements. It should further be noted that any reference signs do not limit the scope of the claims, that the example embodiments may be realized in the broadest sense of the claims.

[0130] EXAMPLES

[0131] In brief, CD93-binding antibodies have been selected. The antibody production and validation workflow started with performing phage display selection, followed by cloning and expression of binders. Screening for binding was performed by ELISA, followed by sequencing of hits (154 sequences of unique clones). A secondary ELISA was performed to verify binding (selection of clones based their binding towards hCD93, hCD93 lectin domain and mCD93), followed by and HTRF assay (Homogeneous Time Resolved FRET). Affinity screening using SPR of 30 scFv clones, 14 Fab clones (total 44 clones), followed by small scale protein purification (Kingfisher Flex). Kinetics measurements were performed using SPR, followed by FACS study of binding to human and mouse endothelial cells, selecting 27 clones. Conversion was made to hlgG 1 LALA, followed by in vitro functional assays in human and mouse endothelial cells (15 hlgG selected). Ex vivo functional assays were performed on human and mouse brain slices, where 4 hlgG selected (3 cross-reactive with mouse, 1 binds only human CD93) were selected for showing the desired functionality. 3 hlgG cross-reactive with mouse were converted in mouse backbone, and in vivo functional assays in mice were performed.

[0132] Example 1 , CD93 expression in glioma patients

[0133] The clinical relevance of CD93 in glioma was assessed by analyzing CD93 expression in low-grade glioma (grade II), high-grade glioma (grade III) and glioblastoma (grade IV) patients from the publicly available CGGA database.

[0134] Methods:

[0135] Expression of CD93 RNA in human glioma samples was performed using the CGGA (Chinese Glioma Genome Atlas) RNASeq glioma datasets (Zhao Z et al. Genomics Proteomics Bioinformatics. 2021 Feb;19(1):1-12.), downloaded from the GlioVis data portal (http: / / qliovis.bioinfo.cnio.es). Glioma samples were graded according to the 2007 WHO classification. Each dot in the graph represents the CD93 mRNA level in a tumor obtained from one individual patient (651 patients in total). The p value was determined using a one-way ANOVA followed by Tukey's multiple comparison test. P < 0.05 was considered statistically significant. Results:

[0136] The graph in Figure 2 shows increased CD93 expression in patients diagnosed with high-grade glioma (grade III) and glioblastoma (grade IV) as compared to patients with low- grade glioma (grade II).

[0137] Conclusion:

[0138] This data indicates that CD93 is highly expressed in glioma patients and its expression increases with the severity of the disease. Indeed, CD93 mRNA levels are significantly higher in tumor samples from patients with the most aggressive form of glioma (grade IV or glioblastoma) as compared to grade III or grade II glioma patients.

[0139] Example 2, Tumor growth and survival in CD93 deficient mice

[0140] The role of CD93 in glioma progression was assessed in CD93 knockout and wild-type mice orthotopically injected with the murine glioma model GL261. Tumor growth and survival analysis was performed.

[0141] Material and Methods:

[0142] Mice:

[0143] C57BL / 6 mice with a global knockout of CD93 (CD93_ / “) and wild-type (WT) littermates were bred in house. C57BL / 6 WT mice were purchased from Taconic M&B.

[0144] Tumor cell culture:

[0145] GL261 murine glioma cells were cultured in DMEM (Life Technologies) supplemented with 10% FBS (Sigma-Aldrich) at 37°C and 5% CO2 / 95% air in a humidified chamber.

[0146] In vivo tumor model:

[0147] GL261 cells (20 000 cells in 2pl of Dulbecco’s PBS) were orthotopically injected in the brain of CD93_ / “ and wild-type mice. Tumor growth was assessed at the endpoint of the study (23 days after tumor injection) by analyzing the tumor area visualized by nuclear staining on at least 6 tissue sections (80um thick section) across the entire brain. The p value was determined using unpaired t-test. p < 0.05 was considered statistically significant.

[0148] For survival studies, mice were sacrificed when they lost more than 10% body weight or showed symptoms. The p value was determined using Gehan-Breslow-Wilcoxon test, p < 0.05 was considered statistically significant.

[0149] Results:

[0150] The graph in Figure 3A represents the quantification of GL261 tumor area in CD93- deficient (CD93_ / ) mice and in wild-type mice showing a significant reduction of tumor growth in CD93- deficient mice compared to the wild-type group. In line with this observation, Figure 3B shows improved survival in CD93 / _mice compared to the wild-type (WT) mice.

[0151] Conclusion:

[0152] These data indicate that CD93 plays an important role in tumor progression in a murine glioma model. Indeed, CD93-deficiency results in reduced tumor growth and improved survival in the GL261 model.

[0153] Example 3, Endothelial activation and T cell recruitment in glioma-bearing CD93-deficient mice, and survival in glioma-bearing CD93 deficient mice after aPD1 -blockade immunotherapy.

[0154] The role of CD93 in endothelial activation and T cell infiltration as well as the response to aPD1 -blockade immunotherapy during glioma progression was assessed in CD93 knockout and wild-type mice orthotopically injected with the murine glioma model GL261. Survival analysis was performed.

[0155] Material and Methods:

[0156] Mice:

[0157] C57BL / 6 mice with a global knockout of CD93 (CD93_ / “) and wild-type (WT) littermates were bred in house. C57BL / 6 WT mice were purchased from Taconic M&B.

[0158] Tumor cell culture:

[0159] GL261 murine glioma cells were cultured in DMEM (Life Technologies) supplemented with 10% FBS (Sigma-Aldrich) at 37°C and 5% CO2 / 95% air in a humidified chamber.

[0160] In vivo tumor model:

[0161] GL261 cells (20 000 cells in 2pl of Dulbecco’s PBS) were orthotopically injected in the brain of CD93_ / “ and wild-type mice. Mice were treated with aPD1 antibody (clone:RMP1-14, Cat# BE0146, BioXCell, USA) three times (200pg / dose every 3 days starting at day 10 after tumor injection). Mice were sacrificed when they lost egual or more than 20% body weight or showed symptoms. The survival analysis was performed using Gehan-Breslow-Wilcoxon test. P < 0.05 was considered statistically significant.

[0162] Immunofluorescent staining:

[0163] Tumor tissue were stained with anti-CD31 (2H8, Thermo Fisher Scientific, MA3105), anti- VCAM1 (R&D System, AF643) and anti-CD8 (Abeam, ab237723). Sections were washed and incubated with specific Alexa Fluor-conjugated secondary antibodies (Invitrogen). Nuclei were visualized by Hoechst (Life Technologies) and images were taken using confocal microscope (Leica SP8).

[0164] Results: Figure 4A and Figure 4B shows increased expression of VCAM1 in tumor vessels in CD93-deficient mice. Figure 4C and Figure 4D shows increased infiltration of T cell (CD8 positive cells) in CD93 / _tumors. Figure 4E shows improved survival in CD93-deficient mice after aPD1 -blockade immunotherapy as compared to the wild-type mice.

[0165] Conclusion:

[0166] This data indicates that CD93 expression limits endothelial activation and reduces infiltration of T cells in glioma, and is associated with worse response to immunotherapy. Indeed, CD93 deficiency results in increased endothelial activation and an increased abundance of cytotoxic CD8 T cells in the GL261 tumors. In addition, survival after aPD1- antibody therapy in the GL261 model is improved in CD93-deficient mice.

[0167] Example 4, Role of CD93 in the organization of extracellular matrix

[0168] Extracellular matrix plays an important role in tumor progression and spreading. Here, we have analyzed the role of CD93 in controlling the organization of the extracellular matrix protein fibronectin into fibrillar structures both in vitro on cultured human endothelial cells as well as in vivo using the GL261 glioma model.

[0169] Material and Methods:

[0170] Mice:

[0171] C57BL / 6 mice with a global knockout of CD93 (CD93_ / “) and wild-type (WT) littermates were bred in house. C57BL / 6 WT mice were purchased from Taconic M&B.

[0172] Cell culture:

[0173] Human dermal blood endothelial cells (HDBECs) (PromoCell) were cultured in gelatin- coated culture dishes in Endothelial Cell Basal Medium with full supplements (PromoCell, EBM- MV2). GL261 murine glioma cells were cultured in DMEM (Life Technologies) supplemented with 10% FBS (Sigma-Aldrich). Both cell types were maintained at 37°C and 5% CO2 / 95% air in a humidified chamber.

[0174] In vivo tumor model:

[0175] GL261 cells (20 000 cells in 2pl of Dulbecco’s PBS) were orthotopically injected in the brain of CD93_ / “ and wildtype mice. Tumor tissue was harvested 23 days after tumor implantation and vibratome sectioned for further analysis. siRNA transfections:

[0176] HDBECs were incubated with scrambled control siRNA or siRNA specifically targeting CD93 (Hs_CD93_1 ; FlexiTube, Qiagen) at a concentration of 2nM in a mixture of 20% Opti- MEM (Life Technologies) in endothelial cell medium supplemented with 30 pl / ml Lipofectamine RNAiMAX (Life Technologies) for 4-6 hours, after which the medium was replaced with fresh medium. Experiments were performed at day 2 after siRNA transfection.

[0177] Immunofluorescent staining: siRNA-transfected HDBECs growth in 8-well chamber slides were fixed with 4% PFA and incubated with anti- fibronectin (Abeam, ab2413). Cells were washed and incubated with Alexa Fluor-conjugated secondary antibodies (Invitrogen). Nuclei and cytoskeleton were visualized by Hoechst and Alexa Fluor-647 labeled phalloidin respectively (all from Life Technologies). Immunofluorescent staining of tumor tissue was performed in vibratome sections of murine GL261 tumors. Tumor tissue was stained with anti-CD31 (2H8, Thermo Fisher Scientific, MA3105) and anti-fibronectin (Abeam, ab2413). Sections were washed and incubated with specific Alexa Fluor-conjugated secondary antibodies (Invitrogen). Nuclei were visualized by Hoechst (Life Technologies). Cells and tumor tissue were analyzed under a confocal microscope (Leica SP8).

[0178] Results:

[0179] CD93 downregulation in human endothelial cells impaired fibronectin fibrillogenesis in vitro. As showed in Figure 5A, control endothelial cells (siCtrl) form a dense fibronectin matrix organized in a fibrillary network. However, downregulation of CD93 (siCD93) was associated with a disruption in the fibronectin fibers as compared to control condition.

[0180] Similarly, immunostaining of GL261 tumors revealed a significant reduction of fibronectin deposition in CD93_ / _mice as compared with tumors from wild-type mice (Figure 5B and quantification in Figure 5C). In addition, as shown by the high magnification image in Figure 5D, fibronectin deposition in wild-type tumors was predominantly associated with the vasculature (CD31 positive) suggesting that the fibronectin matrix is mainly deposited and organized in fibrillary structures by endothelial cells in GL261 tumors.

[0181] Conclusion:

[0182] Here we showed that CD93 plays a crucial role in the deposition and organization of the extracellular matrix protein fibronectin both in human endothelial cells in vitro as well as in the GL261 vasculature in vivo.

[0183] Example 5, CD93 is expressed in vessels co-opted by cancer cells at the tumor invasive front

[0184] Malignant gliomas are highly invasive cancers. Migration along blood vessels is one of the strategies used by glioma cells to invade the brain parenchyma. We have previously described that CD93 is highly expressed in the glioma vessels at the tumor core (Langenkamp E. et al Cancer Res. 2015 Nov 1;75(21):4504-16) and that CD93 controls endothelial-associated fibronectin deposition and organization (Lugano R. et al J Clin Invest. 2018 Aug 1;128(8):3280-3297), important for tumor cell invasion.

[0185] Here we analyzed the expression of CD93 in vessels localized at the tumor invasive front both in vivo as well as in the ex vivo model of mouse brain slices.

[0186] Material and Methods:

[0187] Mice:

[0188] C57BL / 6 wild-type mice were purchased from Taconic M&B.

[0189] Tumor cell culture:

[0190] GL261-GFP murine glioma cells were cultured in DMEM (Life Technologies) supplemented with 10% FBS (Sigma-Aldrich) at 37°C and 5% CO2 / 95% air in a humidified chamber.

[0191] In vivo tumor model:

[0192] GL261 cells expressing GFP (20 000 cells in 2pl of Dulbecco’s PBS) were orthotopically injected in the brain of wildtype mice. Tumor tissue was harvested 23 days after tumor implantation and vibratome sectioned for further immunofluorescent staining.

[0193] Brain slice model:

[0194] Brains from wild-type mice were freshly dissected, embedded in low melting point agarose (Invitrogen) and vibratome sectioned into 300um thick slices in HBSS complete buffer (IxHBSS, 2.5mM HEPES, 30mM D-Glucose, 1 mM CaCI2, 1 mM MgSO4, 3mM NaHCO3, all from Sigma-Aldrich). Slices were subsequently cultured on 12-well culture inserts 8pm membrane (vWR; 7342736P) in DMEM medium supplemented with L-Glutamine, 5% FBS and 25%HBSS complete buffer in presence of antibiotics. GL261-GFP cells (10 000 cells resuspended in 2ul of DMEM medium) were injected using 10pl hamilton syringe in the brain slices and co-cultured during 48h at 37°C and 5% CO2 / 95% air in a humidified chamber. After this time period, brain slices were fixed with 4%PFA during 1 .5 hours at room temperature and stained as described below.

[0195] Immunofluorescent staining:

[0196] Tumor tissue and brain slices were stained with anti-CD31 (2H8, Thermo Fisher Scientific, MA3105) and anti-CD93 (R&D System, AF1696). Sections were washed and incubated with specific Alexa Fluor-conjugated secondary antibodies (Invitrogen). Nuclei were visualized by Hoechst (Life Technologies) and GL261-GFP signal was detected at 488nm wavelength with a confocal microscope (Leica SP8). Results:

[0197] To investigate the role of CD93 in the tumor invasion, we have analyzed the expression of CD93 in the vessels co-opted by tumor cells at the tumor invading area by immunofluorescent staining. Interestingly, CD93 was found to be highly expressed in the vessels at the tumor border in GL261 tumor bearing mice. As shown in Figure 6A (highlighted by arrowheads), a strong CD93 signal was detected in the vessels (CD31 positive) co-opted by tumor cells (GL261) at the tumor edge indicated by a dotted line. Similarly, immunofluorescent staining in mouse brain slices co-cultured with GL261 cells showed a high expression of CD93 in the vessels associated with tumor cells at the tumor invasive front (arrowheads in Figure 6B).

[0198] Conclusions:

[0199] These data indicate that CD93 expression is enhanced not only in the vessels at the tumor core as previously shown (Langenkamp E. et a! Cancer Res. 2015 Nov 1;75(21):4504-16) but also in the vessels co-opted by tumor cells localized at the tumor invading border.

[0200] Example 6, Glioma cells migrate along the fibronectin fibers

[0201] Fibronectin is a major component of the extracellular matrix and is essential for tumor progression and tumor cell invasion. Here we assessed whether the GL261 cells migrate along the endothelial-associated fibronectin fibers on in vitro, ex-vivo and in vivo models.

[0202] Material and Methods:

[0203] Tumor cell culture:

[0204] GL261-GFP cells were cultured as described in Example 5.

[0205] GL261 spheroids:

[0206] For spheroid formation, GL261-GFP cells were enzymatically detached, counted and plated in 24-well low adhesion plates (Sarstedt) at a density of 80 000 cells per well. Cells were cultured in serum-free DMEM in presence of B-27 supplement (Gibco) and growth factors (20ng / ml FGF and 20ng / ml EGF, all from Peprotech). After 24h in culture, spheroids of similar sizes were seeded on a monolayer formed by wild-type mouse brain endothelial cell and cocultured for 48h at 37°C and 5% CO2 / 95% air in a humidified chamber.

[0207] Murine brain endothelial cell isolation:

[0208] Endothelial cells were freshly isolated from 12-weeks old wild-type mice as previously described (Lugano R. et al J Clin Invest. 2018 Aug 1;128(8):3280-3297) and cultured on 8-well chamber slide until confluency. GL261-GFP spheroids were seeded on the endothelial cell monolayer and co-cultured for 24h at 37°C and 5% CO2 / 95% air in a humidified chamber.

[0209] Brain slice model: Mouse brain slice culture and GL261-GFP co-culture were performed as described in Example 5.

[0210] In vivo tumor model:

[0211] Tumor tissue was obtained from wild-type mice bearing GL261-GFP tumors as described in Example 5.

[0212] Immunofluorescent staining:

[0213] Immunofluorescent staining for fibronectin in cultured endothelial cells, brain slices and tumor tissue was performed as described in Example 5.

[0214] Results:

[0215] To investigate if GL261 glioma cells interact with fibronectin matrix produced by endothelial cells we first co-cultured GL261-GFP spheroids on a monolayer of mouse brain endothelial cells isolated from wild-type mice. As shown in Figure 7A, endothelial cells grown until confluency produce a thick fibronectin matrix organized as fibrillar structures (see Fibronectin fibers image in Figure 7A). Interestingly, when GL261 spheroids were seeded on endothelial cells, they migrated out from the spheroids and invaded the endothelial monolayer along fibronectin fibers produced by the endothelial cells (arrowheads in Figure 7A). In line with this observation, GL261 cells co-cultured on mouse brain slices derived from wild-type mice interact with and migrate along the CD31 positive vessels expressing fibronectin (arrowheads in Figure 7B). Similarly, GL261 tumor cells grown orthotopically in the brain of wild-type mice invade the brain parenchyma migrating along the CD31 positive vessels expressing fibronectin (arrowheads in Figure 7C).

[0216] Conclusions:

[0217] These data indicate that GL261 glioma cells migrate along the endothelial-associated fibronectin matrix.

[0218] Example 7, CD93 deficiency is associated with reduced tumor cell invasion

[0219] The impact of tumor vessel CD93 expression on glioma cell invasion was assessed using the CD93 knockout mouse model. CD93-dependent tumor invasion was analyzed through in vitro, ex-vivo and in vivo tumor models.

[0220] Material and methods:

[0221] Tumor cell culture:

[0222] GL261-GFP cells were cultured as described in Example 5.

[0223] Murine brain endothelial cell isolation: Mouse brain endothelial cells were isolated from wild-type and CD93 / _mice as described in Example 6.

[0224] Brain slice model:

[0225] Mouse brain slices derived from wild-type and CD93Amice were obtained as described in the Example 5.

[0226] GL261 spheroids:

[0227] GL261-GFP spheroids were obtained as described in Example 6. GL261 spheroids of similar sizes were seeded on wild-type and CD93 / _brain slices and co-cultured for 48h.

[0228] In vivo tumor model:

[0229] Tumor tissue was obtained from wild-type and CD93 / _mice bearing GL261 tumors as described in Example 5. GL261 tumor sections were counterstained with Hematoxylin and Eosin (H&E) to visualize the tumor border and quantify the tumor invasion area.

[0230] Imaging analysis:

[0231] Image analysis was performed using the Imaged software. GL261cell invasion on the mouse brain endothelial monolayer was quantified by measuring the area covered by tumor cell normalized by the area occupied by the endothelial cell monolayer.

[0232] The ability of GL261 cells to invade the mouse brain slices was determined by quantifying the number of vessel-associated GL261 sprouts for each spheroid and the invasion distance calculated from the spheroid body.

[0233] GL261 invasion in vivo was assessed by quantifying the area of the brain parenchyma invaded by tumor cells in close proximity to the tumor bulk identified by staining as a dense and packed cellular area. Tumor invasive area was normalized by the tumor bulk perimeter.

[0234] Results:

[0235] To evaluate the role of CD93 in glioma invasion, we have first analyzed the ability of GL261 cells to migrate on a monolayer of mouse brain endothelial cells isolated from wild-type or CD93 / _mice. As shown in Figure 8A and in the quantification graph in Figure 8B, GL261 cells co-cultured with CD93 / _endothelial cells show a significant reduction of the invasion capacity compared to the GL261 cells co-cultured with endothelial cells derived from wild-type mice. Similar result was obtained when GL261 spheroids were cultured on mouse brain slices obtained from CD93 / _mice. Indeed, as shown in Figure 8C and by the quantification graphs Figure 8D and Figure 8E, tumor cell invasion along the vessels was significantly reduced on CD93 / _brain slices compared to the tumor cell migration observed in the brain slices derived from wild-type mice. This observation was further confirmed in vivo by analyzing the tumor invasive area of GL261 tumors grown in wild-type and CD93 / _mice. Interestingly, while tumors in the wild-type mice showed an irregular and invasive tumor border, tumors in CD93 / _mice were characterized by a linear tumor edge with significantly reduced tumor invading area (Figure 8F and quantification graph in Figure 8G).

[0236] Conclusions:

[0237] These results highlight the important role of CD93 in promoting perivascular invasion of glioma cells. Indeed, CD93-deficiency significantly reduces the invasion of GL261 cells along the blood vessels in the brain parenchyma. This observation suggests CD93 as a potential therapeutic target to inhibit glioma cell spreading.

[0238] Example 8, CD93-deficiency reduces tumor cell proliferation

[0239] Proliferation of human glioma cell line U87 and the murine GL261 glioma cell line was assessed in presence or absence of CD93 through in vitro and ex-vivo co-culture experiments.

[0240] Material and methods:

[0241] Cell culture:

[0242] Human dermal blood endothelial cells (HDBECs) and GL261 cells were cultured as described in Example 4.

[0243] U87 human glioma cells were cultured in MEM (Gibco) supplemented with 10% of FBS. siRNA transfections:

[0244] CD93 downregulation in HDBECs was performed as described in Example 4.

[0245] Proliferation assay:

[0246] In vitro proliferation assay was performed in U87 cells. Briefly, U87 cells were plated in 24-well plate at a density of 30 000 cells each well. After adhesion, cells were treated with conditioned medium derived from control (Mock and siCtrl) or CD93 siRNA (siCD93_1 and siCD93_5) transfected HDBECs. Conditioned medium (EBM medium supplemented with 1 %FBS) was collected after 24h of incubation, centrifuged at 1200rpm to remove cell debris and transferred to the U87 cells. Tumor cells were incubated with the conditioned medium during 4 days. Cells proliferation was assessed at day 2, 3 and 4 after seeding by manual cell counting.

[0247] Mouse brain slices and GL261 spheroids:

[0248] Mouse brain slices derived from wild-type and CD93Amice were obtained as described in the Example 5. Co-culture with GL261 spheroids was performed as described in Example 6.

[0249] Immunofluorescent staining: Brain slices were stained with anti-Ki67 (Abeam, ab16667), washed and incubated with specific Alexa Fluor-568 conjugated secondary antibody (Invitrogen). Nuclei were visualized by Hoechst (Life Technologies). Ki67 expression analyzed under a confocal microscope (Leica SP8).

[0250] Results:

[0251] Membrane bound CD93 can be cleaved by metalloproteases and released as a soluble form. To investigate if the soluble CD93 released from endothelial cells regulates tumor cell proliferation, human glioma U87 cells were incubated with the conditioned medium derived from human endothelial cells control, containing soluble CD93, as well as from CD93 siRNA silenced endothelial cells, containing significantly reduced amount of soluble CD93 (Figure 9A). Surprisingly, the analysis of U83 proliferation monitored during 4 days, revealed that U87 cells incubated with the conditioned medium derived from CD93 silenced endothelial cells (siCD93_1 and siCD93_5) proliferate significantly less compared to the U87 treated with the conditioned medium derived from control endothelial cells (Mock and siCtrl). In line with this, the immunofluorescence staining with the proliferation marker Ki67 in the GL261 spheroids cultured on CD93 / _brain slices showed a significant reduction of the Ki67 positive cells, compared with the Ki67 signal detected in spheroids co-cultured with wild-type brain slices (Figure 9B and Figure 9C).

[0252] Conclusion:

[0253] This data shows that CD93 deficiency inhibits tumor cell proliferation both in vitro and in the ex vivo brain slice model.

[0254] Example 9, Phage display selection on human and / or mouse CD93 using scFv and Fab antibody libraries

[0255] Phage display selections were performed to enable isolation of scFv and Fab fragments with specificity for human and / or mouse CD93.

[0256] Material and Methods

[0257] Phage display selection

[0258] Biopanning was performed using four selection rounds of enrichment employing the two in-house constructed human synthetic scFv phage library, SciLifeLib 1 , and Fab phage library, SciLifeLib 3 (SciLifeLab, Stockholm, Sweden). SciLifeLib 1 is a naive human synthetic scFv libraries, similar in design and construction to previously reported (Sall, et al., Protein Eng Des Sei (2016) 29: 427-437). Briefly, human germline genes IGHV3-23 and IGKV1-39 were used as library scaffold and Kunkel mutagenesis was used to introduce diversity into four of the six complementarity determining regions (CDR); namely CDR-H1 , CDR-H2, CDR-H3 and CDR-L3. SciLifeLib 3 was constructed in a similar way but in a Fab format. The selection was performed using streptavidin-coated magnetic beads (Dynabeads M-280, ThermoFisher Scientific, #11206D) and biotinylated antigen. Three Avi-tagged antigen constructs were used: human or mouse CD93 extracellular domain (amino acids 1-580), referred to as hCD93-avi and mCD93- avi, respectively, and avi-tagged human CD93 lectin domain (amino acids 1-258), referred to as hCD93lectin-avi. The Avi-tag allows site-specific incorporation of biotin. For each of the two phage libraries, a straight selection track using either hCD93-avi, mCD93-avi or hCD93lectin-avi, respectively, as bait was included. In an additional track, in order to preferentially select for cross-species reactive scFv and Fab clones, the antigen was alternated between hCD93-avi and mCD93-avi between the different rounds. The selection pressure was increased by gradually decreasing the antigen amount (round 1 : 200 nM, round 2: 50 nM, round 3: 10 nM, round 4: 2 nM) and by increasing the number and intensity of washes between the different rounds. In order to remove non-specific or streptavidin binders, pre-selection was performed by incubation of the phage stocks against empty streptavidin coated magnetic beads prior round 1 and 2. Also, 1 % bovine serum albumin (BSA) was included as blocking agent throughout the selection procedure. Elution of antigen-bound phages was performed using a trypsin-aprotinin approach. The entire selection process, except the phage-target protein incubation step, was automated and performed with a Kingfisher Flex robot. Recovered phages were propagated in Top10F’ E. coli, either on agar plates at 37°C overnight (Round 1 and 2) or in solution at 30°C overnight (Rounds 3 and 4). Phage stocks were made by infecting with an excess of M13K07 helper phage (New England Biolabs, # N0315S) and scFv expression induced by the addition of IPTG. The overnight cultures were PEG / NaCI-precipitated, resuspended in selection buffer and used for the next round of selection.

[0259] Re-cloning and expression of scFv and Fab

[0260] To allow production of soluble scFv and Fab, phagemid DNA from the third and fourth round of each selection track was isolated. In pools, the genes encoding the scFv fragments were restriction enzyme digested and sub-cloned into screening vector, providing a signal for secretion of the scFv along with a triple-Flag tag and a hexahistidine (His) tag at the C-terminus. Analogously, the genes encoding the Fab fragments were restriction enzyme digested and subcloned into a vector containing a signal for Fab secretion with a C-terminal a hexahistidine (His) tag. The constructs were subsequently transformed into TOP10 E. coli. Single colonies were picked, cultivated and IPTG-induced for soluble scFv expression in 96-well format. In total, 752 clones, 376 scFv and 376 Fab, present in bacterial supernatant were prepared for a primary ELISA screen.

[0261] ELISA screen

[0262] Antigens hm CD93, ms CD93 and hm lectin domain, were coated through streptavidin at 1 pg / ml into a 384-ELISA well plate. ScFv and Fab clones, present in bacterial supernatant diluted 1 :5 in block buffer (PBS supplemented with 0.5% BSA + 0.05% Tween20) were added and allowed to bind to the coated proteins. Detection of binding was enabled through an HRP- conjugated aFlag M2 antibody (Sigma-Aldrich #A8592) followed by incubation with TMB ELISA substrate (ThermoFisher Scientific #34029). The colorimetric signal development was stopped by adding 1 M sulfuric acid and the plate was read at 450 nm. All samples were assayed in duplicates from which a mean Abs 450 nm was calculated and subtracted from the mean Abs 450 nm from a blank well (block buffer added instead of a scFv or Fab clone).

[0263] DNA sequencing

[0264] 537 positive scFv or Fab clones showing binding to CD93 were sent for Sanger DNA sequencing to GATC Biotech (Ebersberg, Germany).

[0265] Results

[0266] A total of eight selection tracks were carried out in parallel on human CD93, human lectin domain and mouse CD93 using phage libraries SciLifeLib 1_scFv and SciLifeLib 3_Fab. Following re-cloning of scFv and Fab clones into soluble expression vectors, 94 clones (colonies) were picked from selection rounds 3 and 4 per track, resulting in a total of 752 clones picked and cultivated.

[0267] ELISA screen resulted in the identification of 537 potential hits having four distinct binding characteristics with clones binding either 1) hCD93-avi, 2) hCD93-avi and hCD93lectin- avi, 3) hCD93-avi and mCD93-avi, or 4) mCD93-avi. The majority of hits displayed binding towards hCD93-avi and hCD93lectin-avi.

[0268] DNA sequencing of the 537 clone hits resulted in the identification of 154 sequence unique clones; 117 scFv and 37 Fab clones.

[0269] Conclusion

[0270] ScFv and Fab clones binding to human and / or mouse CD93 have successfully been isolated by phage display selections. Following an initial ELISA screen on 752 clones and DNA sequencing of 537 positive ELISA hits, a total of 154 sequence unique clones (117 scFv clones and 37 Fab clones) were identified.

[0271] Example 10, Assessing the binding characteristics of 154 sequence unique clones The 154 sequence unique scFv and Fab clones from Example 9 were selected for further characterization by a secondary ELISA (Enzyme-linked immunosorbent assay), a HTRF assay (Homogenous Time Resolved Fluorescence) and surface plasmon resonance (SPR) in a kinetic screen-based approach to enable ranking of the different clones. The most promising clones were tested by flow cytometry (FACS) to assess their binding to endothelial cells.

[0272] Material and Methods

[0273] ELISA hCD93, mCD93, hCD93lectin and a non-relevant protein (negative control), were coated through streptavidin at 1 pg / ml into a 384-ELISA well plate. Hm CD93-Fc (RnD Systems #2379- CD) was directly coated at 1 pg / ml in PBS o / n at 4°C. scFv and Fab clones, present in bacterial supernatant, were diluted 1 :10 in block buffer (PBS supplemented with 0.5% BSA + 0.05% Tween20) and allowed to bind to the coated proteins. Detection of binding was enabled through an HRP-conjugated aFLAG M2 antibody for scFv clones (Sigma-Aldrich #A8592) or an HRP- conjugated a-hm KAPPA antibody for Fab clones (Southern Biotech #9230), followed by incubation with TMB ELISA substrate (ThermoFisher Scientific #34029). The colorimetric signal development was stopped by adding 1 M sulfuric acid and the plate were analyzed at Abs 450 nm.

[0274] All samples were assayed in duplicates from which a mean Abs 450 nm was calculated and subtracted from the mead abs 450 nm from a blank well (block buffer added instead of a scFv or Fab clone).

[0275] HTRF

[0276] The scFv and Fab clones, together with a non-relevant scFv and a non-relevant Fab (negative controls) were diluted 1 :5 in assay buffer (PBS supplemented with 0.1 % BSA) and allowed to bind to hCD93, mCD93, non-relevant protein and non-relevant peptide, respectively, diluted to 200 nM in assay buffer. Detection of binding was enabled through donor molecule terbium-conjugated a-FLAG antibody (Cisbio #611 FG2TL) for scFv clones or a europium- conjugated aKAPPA antibody (Cisbio #61 KAPKAA) for Fab clones, in combination with acceptor molecule streptavidin-conjugated XL665 (Cisbio #610SAXL). Plates were incubated dark for 2 h at room temperature before being analyzed on a Envision spectrometer (Perkin Elmer) at 615 nm (background / noise signal) and 665 nm (binding signal). The 665 nm value was divided with the 615 nm value and multiplied with 10 000 to retrieve a delta R value for each sample.

[0277] All samples were assayed in duplicates from which a mean value was retrieved and subtracted with the mean value of a blank well (assay buffer added instead of a clone). Kinetic screen

[0278] The kinetic screen was performed on a BIAcore T200 instrument (Cytiva). An aFLAG M2 antibody (Merck #F1804) or an a-hm KAPPA antibody (Cytiva #28958325) functioning as a capture ligand for scFv or Fab clones, respectively, was immobilized onto all 4 surfaces of a CM5 series S sensor chip using EDC / NHS amine coupling chemistry according to manufacturer’s recommendations. All experiments were performed at 25°C in running buffer (HBS supplemented with 0,05% Tween20, pH 7.5). scFv and Fab clones, present in bacterial supernatant, were each injected and captured onto the chip surface through the antibody capture ligand, followed by a single antigen injection of hCD93, mCD93 and hCD93lectin, respectively, at 50 nM. Following a dissociation phase, surfaces were regenerated with 10 mM glycine-HCI at pH 2.1.

[0279] For each clone, the obtained sensorgram was subtracted from a reference surface and from a blank run, being the same clone captured with running buffer injected as antigen. Double referenced subtracted data was fitted to a 1 :1 Langmuir binding model using BIAeval software (Cytiva), and apparent kinetic constants could be retrieved.

[0280] Flow cytometry:

[0281] Human endothelial cells (HDBECs) were cultured as described in Example 4. U87 cells were cultured as described in Example 8. Mouse endothelial cells (MS1) were cultured in DMEM (Gibco) supplemented with 10%FBS. GL261 cells were cultured as described in Example 5.

[0282] Cells growth in cell culture plates were detached by using accutase-dissociating buffer (ThermoFisher Scientific). Cells were incubated 20 minutes at 4°C with different clones, specifically, HDBECs and the negative control U87 cells were incubated with 150ng / ml of the human specific and human-mouse cross reactive clones, MS1 and the negative control GL261 cells were incubated with 300ng / ml of the mouse specific and human-mouse cross reactive clones. Clone concentrations were determined based on the expression of CD93 on the endothelial cell cultures used in the assay. After this incubation time, cells were washed and stained with the secondary antibody anti-human kappaPE (SouthernBiotech #9230-09) to detect the scFv and Fab clones. Samples were run on CytoFLEX LX (Beckman Coulter) and data were analyzed using FlowJo version 10.5.3 (FlowJo LLC). Percentage of positive cells for each clone was assessed.

[0283] Results

[0284] Binding of 154 scFv and Fab clones to hCD93, mCD93 and hCD93lectin was measured by ELISA, HTRF, SPR kinetic screen and cell binding (FACS). The results of the most promising clones are summarized in Table 1 above. The results of the methods correlated well. The scFv clones could be divided into four main groups based on binding specificity; 1) clones binding to hCD93; 2) clones binding to hCD93 and hCD93lectin; 3) clones binding to hCD93 and mCD93 and 4) clones binding to mCD93. The Fab clones could be divided into three main groups based on binding specificity; 1) clones binding to hCD93 and hCD93lectin; 2) clones binding to hCD93 and mCD93 and 3) clones binding to mCD93. No significant background binding to the non- relevant protein or streptavidin was detected for any of the clones. The most promising clones displayed an apparent affinity, KDapp, in the sub- to low-nanomolar range towards their antigen or antigens (the abbreviation apparent affinity, KDapp, is used since only a single antigen concentration was analyzed). The most promising clones were subsequently tested for their specific binding to CD93 expressed by endothelial cells. For this, human and / or mouse endothelial cells were exposed to each clone and the binding was assessed by flow cytometry. U87 and GL261 cells do not express CD93 and were therefore used as negative controls. As shown in Table 1 , all tested clones bind with high affinity and specificity to endothelial cells. Indeed, the percentage of cells positive for the clones varied between 74.6% to 99.8% and the percentage of positive cells in the negative controls were below 4.7%.

[0285] Based on the results shown in Table 1 , 26 clones (16 scFv and 10 Fab clones) were selected for conversion to full length human lgG1 LALA.

[0286] Conclusions

[0287] A secondary ELISA, a HTRF assay and a kinetic screen by SPR have been performed on 154 sequence unique scFv and Fab clones. Specific binding of the most promising clones has been tested on endothelial cells by flow cytometry showing high affinity and specificity. The combined data facilitated further ranking of clones and enabled the selection of the 27 most promising clones (17 scFv clones and 10 Fab clones) for small scale protein purification and further analysis.

[0288] Example 11 , Conversion of 26 clones to human lqG1 LALA and further assessment of binding to CD93

[0289] Twenty-six of the most promising scFv and Fab clones from Example 9 were successfully converted to human lgG1 LALA. Selection of these clones was based on their performance in biophysical, biochemical and cell binding assays, as well as on reactivity across different antigen constructs.

[0290] Material and Methods Cloning The PCR amplified VH and VL regions of the 26 scFv and Fab clones to be converted were inserted into in-house constructed vector pHAT-hlgG1-LALA using the InFusion HD Plus Cloning Kit (Clontech #638909). In addition, scFv HuF11 (from patent WO 2020 / 180706) and two isotype controls were also converted into hlgG 1 LALA.

[0291] The cloning reactions were transformed into E. coli Stellar cells and a subsequent colony PCR screen was performed from which two clones per construct were sent for DNA sequencing to verify correct DNA sequences of each antibody clone.

[0292] Expression and purification

[0293] Plasmids were isolated for transient transfection. Expi293FTMcells (Thermo Scientific A14527) were cultured at 37°C, 105 rpm, 70% rH, 7% CO2. In Expi293TMexpression medium (Thermo Fisher A1435101). For transfection, cells were diluted to appr 2.5*106cells / ml with medium. Sterile-filtered DNA was mixed with Opti-MEM (Thermo Scientific 31985062) and FectoPRO DNA transfection reagent (Polyplus 116-010) and incubated for 10-60 minutes before adding the mixture to the cells. FectoPRO Booster (Polyplus 116-010) was added to the 125 ml cultures and expression was for five days. At harvest, cultures were centrifuged at 3000 ref. Supernatants were filtered using Steriflip-GP 0.22 pm polyethersulfone gamma irradiated filter units (Sigma SCGP00525) and stored at 4°C until purification.

[0294] Purification was run on an AKTA Pure 25 Protein Purification system (Cytiva) in a cooling cabinet. An autosampler (Teledyne Cetac ASX-280) was used to load the samples. A Fibro HiTrap PrismA column (Cytiva 17549855) was equilibrated with Dulbecco’s phosphate buffered saline buffer (PBS; Sigma D8537-500ML), sample was loaded, and the column was washed with 10 column volumes (CV) of PBS. Bound protein was eluted with 6 CV of 0.1 M glycine-HCI buffer (Polysciences Inc 24074-500). Eluted antibody was collected in a 2 ml loop using the peak collection feature in the Unicorn software. The antibody was then loaded on a HiTrap Desalting column (Cytiva 17140801) equilibrated with PBS from the sample loop; the buffer-exchanged protein was collected, and the column re-equilibrated for the next sample. Antibodies were randomly picked out for endotoxin testing (Endosafe Portable Endotoxin System, Charles river). All purified antibodies were analyzed by SDS-PAGE and size exclusion chromatography (SEC) using a HLPC (Agilent 1200 system) and an Agilent Bio-3 column.

[0295] ELISA

[0296] HCD93, mCD93, hCD93lectin and an isotype control antigen were coated through streptavidin at 1 p.g / ml into a 384-ELISA well plate. Purified antibodies were diluted to 1 p.g / ml block buffer (PBS supplemented with 0.5% BSA + 0.05% Tween20) and allowed to bind to the coated proteins. As positive assay controls, an ot-hm CD93 ab (BD Bioscience #552954) and an a-ms CD93 ab (RnD Systems #MAB1696) were also included. Detection of binding was enabled through an HRP-conjugated a-hm KAPPA antibody (Southern Biotech #9230), or an HRP-a- mouse KAPPA chain ab (Southern Biotech #1050) and an HRP-a-rat IgG ab (Thermo Scientific #31470) for the positive control antibodies, followed by incubation with TMB ELISA substrate (Thermo Scientific #34029). The colorimetric-signal development was stopped by adding 1 M sulfuric acid and the plate were analyzed at Abs 450 nm.

[0297] SPR

[0298] A kinetic screen was performed on a BIAcore T200 instrument (Cytiva) using a capturebased approach. All experiments were performed at 25°C in running buffer (HBS supplemented with 0,05% Tween20, pH 7.5).

[0299] An a-hm KAPPA antibody (Cytiva #28958325) functioning as a capture ligand was immobilized onto all 4 surfaces of a CM5 series S sensor chip using EDC / NHS amine coupling chemistry according to manufacturer’s recommendations. Purified hlgG 1 LALA clones diluted in running buffer were each injected and captured onto the chip surface through the antibody capture ligand, followed by a single antigen injection of hCD93, mCD93 and hCD93lectin, respectively, at 50 nM. Chip surfaces were regenerated with 10 mM glycine-HCI at pH 2.1.

[0300] For each clone, the obtained sensorgram was subtracted from a reference surface and from a blank run, being the same clone captured with running buffer injected as antigen. Double referenced subtracted data was fitted to a 1 :1 Langmuir binding model using BIAeval software (Cytiva), and apparent kinetic constants could be retrieved.

[0301] BVP ELISA

[0302] Baculovirus particles (BVP) were precipitated from the supernatant of baculovirus infected Sf9 cells using a standard PEG / NaCI precipitation protocol. Briefly, a final concentration of 4% PEG8000 and 0.5 M NaCI was added to the supernatants, and after incubation on ice for 30 min, the precipitated BVPs were pelleted by centrifugation. Pellets were re-suspended in PBS, and centrifuged again to remove cell debris.

[0303] Baculovirus particles diluted 1 :20 in PBS were direct immobilized to 384-well microtiter at 4°C overnight. aCD93 antibodies and reference IgGs (fletikumab, panitumab, trastuzumab, bevacizumab, duligotuzumab and lanzilumab; Absolute Antibody) were diluted in blocking buffer (PBS + 0.5% BSA) to a final concentration of 100 nM and 20 nM, and added to the wells. Detection of bound IgGs was performed using a horseradish peroxidase (HRP)-conjugated antihuman IgG kappa antibody. After incubation with the secondary detection antibody, signal development was initiated with the chromogen Ultra TMB-ELISA (Thermo Scientific), and reaction was stopped by the addition of 1 M sulphuric acid. Absorbance was measured at 450 nm.

[0304] Off-target binding by ELISA

[0305] A panel of antigens (cardiolipin from bovine heart, Sigma Aldrich; keyhole limpwt haemocyanin, Sigma Aldrich; LPS from E.coli O111 :B4, InvivoGen; ssDNA from calf thymus, Sigma Aldrich; dsDNA from calf thymus, Sigma Aldrich, human insulin, Sigma Aldrich) were immobilized to 384-well microtiter plate through direct immobilization at 4°C overnight, at different concentrations in PBS. Cognate antigens for the reference and AD169 antibodies (IL20, Sino Biological (no tag); biotinylated avitagged human Her2, Aero Biosystems; biotinylated avitagged human VEGF165, Aero Biosystems; biotinylated human ErbB3, Aero Biosystems; biotinylated human GM-CSF, Abeam; biotinylated avitagged SARS CoV-2 S1 RBD, in house production; biotinylated avitagged hCD93, in house production) were immobilized onto the microtiter plates, either through direct immobilization at 4°C overnight (for all non-biotinylated antigens), or via immobilized streptavidin (for all biotinylated antigens) at 1-2 pg / ml in PBS. Addition of IgG and the following steps was performed as described for the BVP ELISA.

[0306] Results

[0307] Cloning, expression and purification

[0308] The genes encoding VH and VL of 26 clones, reference HuF11 and two isotype controls, were successfully transferred into a vector encoding the human IgG 1 LALA subclass, as confirmed by DNA sequencing.

[0309] Following production of clones in Expi293F cells, Protein A purification and buffer exchange to PBS, clone 39 was excluded from the set due to too low yield.

[0310] ELISA

[0311] Binding pattern of clones correlates with previously performed ELISA of purified clones in scFv and Fab format, except for clones AD169-33, -85, -108, -334, which now displayed some binding towards additional CD93 antigens, though lower binding signal than towards their expected CD93 antigens (Table 2).

[0312] Positive control antibodies were found to only bind to their expected antigens. Isotype controls only displayed binding towards their antigen, SARS-CoV-2 S1 protein. Isotype control MO176-301 did however demonstrate an elevated background binding to mCD93.

[0313] SPR

[0314] Converted hlgG 1 LALA clones were subjected to a kinetic screen and challenged for binding to hCD93, hCD93lectin and mCD93, respectively, using a single antigen concentration at 50 nM. SPR binding data correlated well with ELISA data, except for clone AD169-33 which only displayed binding towards hm CD93 and not, as in ELISA, to mCD93.

[0315] The most promising clones displayed an apparent affinity, KDapp, in the sub- to low- nanomolar range (Table 2). The abbreviation apparent affinity, KDapp, is here used since only a single antigen concentration was analyzed. The SPR sensorgrams of the four most promising antibodies are shown in Figure 10.

[0316] Off-target binding assessment hlgG 1 LALA clones were assayed for unspecific binding by measuring binding to BVPs and a panel of various biomolecules. Low signals were observed in the BVP binding assay for most of the aCD93 antibodies (Table 2). However, clones 41 and 334 showed medium level binding to BVPs. None of the clones were classified as high in terms of BVP binding. The binding pattern of reference antibodies was as expected - lenzilumab and duligotuzumab giving medium to high level signal, others remaining low (data not shown).

[0317] Low intensity signals were obtained for most of the aCD93 antibodies in ELISA measuring the binding to the off-target panel (Table 2). Clone 307 gave a medium level signal, and high binding to various off-target antigens was recorded for clone 334.

[0318] Conclusions

[0319] Following conversion of scFv and Fab fragments to hlgG1 LALA, all binders have retained their binding to the antigen, as was determined by ELISA, SPR and FACS experiments. Clones 41 , 307 and 334 were removed due to unspecific binding behavior.

[0320] Example 12, Selection of aCD93 IqGs based on their performance in functional assays

[0321] Selection of the most promising aCD93 antibodies was performed by testing their functional effects in human and mouse endothelial cells (in vitro assays) and in mouse and human brain slices-tumor cell co-culture (ex vivo assays).

[0322] Result

[0323] The blocking effect of the CD93 antibodies was tested in in vitro and ex vivo functional assays. The results from all the functional assays performed with the selected 26 CD93 antibodies are summarized in the Table 3 above. Experiments showing positive results in functional assays using the four selected antibodies are shown as Examples 12-23, where the materials and methods of each assay are included. Specifically, the CD93 antibodies were tested for their ability to induce disruption of endothelial cell-cell junctions (intercellular gaps formation), inhibit fibronectin fiber formation on endothelial cells and hinder endothelial cell migration. These functional effects have been previously demonstrated by knocking-down CD93 in endothelial cells by using siRNA. In addition, ex vivo mouse and human brain slice models were used to assess the blocking effect of the aCD93 antibodies on tumor cell perivascular invasion and deposition of vessel-associated fibronectin, previously found to be inhibited in CD93 knockout mouse brain slices Example 6 and Example 7.

[0324] Interestingly, as shown in Table 3, different aCD93 antibodies show different magnitude effect on the tested functional assays varying between no-effect and strong-effect (Table 3). These results allowed us to score the 26 aCD93 antibodies based on their performance on blocking CD93 function in a complex environment and select the most promising antibodies. As shown in Table 3, the human-mouse cross-reactive aCD93 antibodies AD169-85, AD169-309, AD169-324 as well as the human specific aCD93 antibody AD169-8 showed effects in all functional assays.

[0325] Conclusions

[0326] Four aCD93 antibodies: AD169-8, AD169-85, AD169-309 and AD169-324, have been selected based on their ability to block CD93 functions.

[0327] Example 13, Assessment of the binding of selected aCD93 IqGs to human endothelial cells

[0328] The specificity of four selected aCD93 IgGs was tested in vitro by comparing the antibody binding signal in human endothelial cells expressing CD93 to control cells that do not express CD93.

[0329] Material and methods:

[0330] Cell culture:

[0331] Human endothelial cells (HDBECs) were cultured as described in Example 4. U87 cells were cultured as described in Example 8.

[0332] Flow cytometry:

[0333] HDBECs cells growth in cell culture plates were detached by using accutase dissociating buffer (ThermoFisher Scientific). Cells were incubated 20 minutes at 4°C with different concentrations of the aCD93 lgG1 or Control lgG1 (0.4pg / ml, 2pg / ml, 10pg / ml and 50pg / ml). After this incubation time, cells were washed and stained with the secondary antibody antihuman kappaPE (SouthernBiotech #9230-09) to detect aCD93 IgGs or Control IgG. Samples were run on CytoFLEX LX (Beckman Coulter) and data were analyzed using FlowJo version 10.5.3 (FlowJo LLC).

[0334] Human specific and human and mouse cross-reactive antibodies AD169-8, AD169-85, AD169-309 and AD169-324 were tested in HDBECs and U87 cells. Results:

[0335] The graph in Figure 11 shows the percentage of cells positive for each of the tested aCD93 antibodies. All antibodies bind specifically to human endothelial cells expressing CD93 showing increased percentage of positive cells in samples incubated with higher antibody concentration. Importantly, undetectable or significantly lower signal was found in the U87 cells that do not express CD93 as well as in the control IgG treated cells.

[0336] Conclusions:

[0337] This data show that all the aCD93 antibodies tested had high specificity for CD93 expressed by human endothelial cells in vitro.

[0338] Example 14, Assessment of the binding of selected aCD93 IqGs to mouse endothelial cells

[0339] In Example 13, the specific binding of selected aCD93 IgG antibodies were tested in human endothelial cells in vitro by flow cytometry. In this example, specific binding of human and mouse cross-reactive aCD93 IgGs was tested in mouse endothelial cells MS1 expressing CD93 and on GL261 cells that do not express CD93.

[0340] Material and methods:

[0341] Cell culture:

[0342] Mouse endothelial cells MS1 were cultured in DMEM (Gibco) supplemented with 10%FBS. GL261 cells were cultured as described in Example 5.

[0343] Flow cytometry:

[0344] Specific binding of human and mouse cross-reactive antibodies AD169-85, AD169-309 and AD169-324 were tested in MS1 cells and GL261 cells as described in Example 13.

[0345] Results:

[0346] The graph in Figure 12 shows the percentage of cells positive for each of the tested aCD93 antibodies. All antibodies bind specifically to mouse endothelial cells expressing CD93, showing high percentage of positive cells also in samples incubated with the lower concentration of aCD93 IgGs (compare graph bars corresponding to 0.2pg / ml with graph bars corresponding to 50pg / ml for each antibody). Importantly, undetectable or significantly lower signal was found in the GL261 cells that do not express CD93 as well as in the control IgG treated cells.

[0347] Conclusions:

[0348] This data show that all aCD93 I gG 1 tested had high specificity for CD93 expressed by mouse endothelial cells in vitro. Example 15, Evaluation of the ability of aCD93 IqGs to block fibronectin fibrilloqenesis on human endothelial cells

[0349] As shown in the Example 4, CD93 deficiency resulted in a disruption of the extracellular matrix fibronectin in endothelial cells. Here we tested the ability of the aCD93 IgGs to block CD93 functions by analysing the fibronectin deposition on human endothelial cells.

[0350] Material and methods

[0351] Cell culture:

[0352] Human endothelial cells were cultured as described in Example 4. Confluent HDBECs were incubated during 24h with 10|jg / ml of aCD93 IgG AD169-8; AD169-85; AD169-309 and AD169-32 or control IgG.

[0353] Immunofluorescent staining:

[0354] HDBECs were immunofluorescent stained for fibronectin and cytoskeleton as described in Example 4. Cells were analysed under fluorescent microscope (DMi8, Leica).

[0355] Results

[0356] Immunofluorescence images in Figure 13A show a dense fibronectin matrix deposition in endothelial cells treated with control IgG. However, treatment with aCD93 IgGs show a reduction in fibronectin deposition on the endothelial cell monolayer. The graph in Figure 13B shows the reduction of fibronectin area in the cells treated with aCD93 antibodies compared to the control IgG.

[0357] Conclusion

[0358] This data indicates that selected CD93-blocking antibodies reduce fibronectin fibrillogenesis in human endothelial cells in vitro.

[0359] Example 16, Evaluation of the blocking effect of aCD93 IqGs on fibronectin fibrilloqenesis on mouse endothelial cells in vitro

[0360] Similar to Example 15, we tested the ability of the aCD93 IgGs to block CD93 functions by analysing the fibronectin deposition on mouse endothelial cells.

[0361] Material and methods

[0362] Cell culture:

[0363] Mouse endothelial cells MS1 were cultured as described in Example 14. Confluent MS1 were incubated during 24h with 10|jg / ml of aCD93 IgG AD169-85; AD169-309 and AD169-324 or control IgG.

[0364] Immunofluorescent staining: MS1 were immunofluorescent stained for fibronectin and cytoskeleton as described in Example 4. Cells were analysed under fluorescent microscope (DMi8, Leica).

[0365] Results

[0366] Similar to human endothelial cells in Example 15, immunofluorescence images of MS1 cells show a dense fibronectin matrix deposition in endothelial cells treated with control IgG (fibronectin signal in control IgG, Figure 14A). Notably, treatment with either of the three aCD93 IgGs resulted in a significant reduction in fibronectin deposition on the endothelial cell monolayer as compared to the control IgG (Figure 14A and Figure 14B).

[0367] Conclusion

[0368] This data indicates that selected CD93-blocking antibodies reduce fibronectin fibrillogenesis in mouse endothelial cells in vitro.

[0369] Example 17, Assessment of the blocking effect of aCD93 IgGs on fibronectin fibrillogenesis on mouse brain slice model

[0370] As shown in the Example 16, treatment with aCD93 IgGs resulted in a disruption of the extracellular matrix fibronectin in mouse endothelial cells in vitro. Here we evaluate the ability of the aCD93 antibodies to inhibit vessel-associated fibronectin in mouse brain slices co-cultured with GL261 cells.

[0371] Material and methods

[0372] Cell culture:

[0373] GL261 cells were cultured as indicated in Example 2.

[0374] Brain slice model:

[0375] Brain slices were obtained from wild-type mice as described in Example 5. 10pg / ml of aCD93 antibodies AD169-85, AD169-309 and AD169-324 were added to the brain slice medium 2 hours before to seed the GL261 cells on the tissue. After 24h, brain slice medium containing aCD93 antibodies was replaced with a fresh medium supplemented with 10pg / ml of aCD93 antibodies. Brain slices were cultured for additional 24h for a total of 48h.

[0376] Immunofluorescent staining:

[0377] Immunofluorescent staining for fibronectin and the vessel marker CD31 was performed as described in Example 4.

[0378] Results

[0379] To investigate whether aCD93 IgGs block CD93 function and inhibit fibronectin fibrillogenesis, GL261 cells were co-cultured on brain slice tissues obtained from wild-type mice and fibronectin deposition was assessed by immunofluorescence staining after 48h treatment with aCD93 antibodies or control IgG. As shown in Figure 15, control IgG treated brain slices show a dense fibronectin matrix in the GL261 tumor area (defined by a dotted line) as well as in the vessels at the tumor invasive front (Figure 15, arrowheads in Control IgG image). In contrast, brain slices treated with aCD93 antibodies AD169-85, AD169-309 and AD169-324 display a significant reduction of fibronectin signal both in the area covered by GL261 cells as well as in the vessels at the invasive front (Figure 15, arrowheads in AD169-85, AD169-309 and AD169- 324 images).

[0380] Conclusions

[0381] Human / Mouse cross-reactive aCD93 IgGs AD169-85, AD169-309 and AD169-324 neutralize CD93 function and inhibit vessel-associated fibronectin in mouse brain slices.

[0382] Example 18, Evaluation of the aCD93 IgGs effect on human endothelial cell-to-cell junction integrity in vitro.

[0383] CD93 deficiency has been associated to disrupted endothelial cell-to-cell junctions in vitro (Langenkamp E. et al Cancer Res. 2015 Nov 1;75(21):4504-16). Here we investigate whether treatment with aCD93 IgGs recapitulates the inhibition of the cell junctions observed in CD93-deficient endothelial cells.

[0384] Material and Methods

[0385] Cell culture:

[0386] Human endothelial cells (HDBECs) were cultured as described in Example 3. Confluent HDBECs were incubated during 24h with 10pg / ml of aCD93 IgGs AD169-8; AD169-85; AD169- 309 and AD169-324 or control IgG.

[0387] Immunofluorescent staining:

[0388] HDBECs were immunofluorescent stained with phalloidin-555 to detect the actin cytoskeleton as described in Example 3. Cells were analysed under fluorescent microscope (DMi8, Leica).

[0389] Endothelial cell-to-cell junction analysis:

[0390] Disruption of the cell junctions was assessed by measuring the formation of gaps between adjacent endothelial cells after aCD93 IgG or control IgG treatment. Gaps quantification was analysed by thresholding the cell free area based on the actin staining. Gap areas were normalized by the number of cells in the field of view.

[0391] Results

[0392] To investigate whether aCD93 antibodies affect the endothelial cell junctions, HDBECs were treated with AD169-8, AD169-85, AD169-309 and AD169-324 aCD93 antibodies or with the control IgG and the cell monolayer was analysed under fluorescent microscope. As showed in Figure 16A, treatment with aCD93 antibodies induces a disruption of the endothelial cell-to- cell junctions shown by the appearance of gaps between adjacent cells. In contrast, HDBECs treated with control IgG showed an intact endothelial monolayer with minimal gap formation (see the thresholded gap areas visualized in white, “Gaps” panels in Figure 16A). Increased gap formation in cells treated with aCD93 antibodies is represented by the quantification graph in Figure 16B.

[0393] Conclusions

[0394] These data shows that CD93-blocking antibodies disrupt human endothelial cell-cell junctions in vitro recapitulating the effect observed in silenced CD93 expression as previously described (Langenkamp E. et al Cancer Res. 2015 Nov 1;75(21):4504-16). In addition, this data indicates that the selected aCD93 antibodies exert a blocking function on human endothelial CD93.

[0395] Example 19, Evaluation of the aCD93 IqGs effect on mouse endothelial cell-to-cell junction integrity in vitro.

[0396] Similar to Example 18, here we investigate whether treatment with aCD93 IgGs disrupt the endothelial cell junctions in mouse endothelial cells in vitro.

[0397] Material and Methods

[0398] Cell culture:

[0399] Mouse endothelial cells (MS1) were cultured as described in Example 14. Confluent MS1 were incubated during 24h with 10pg / ml of human / mouse cross-reactive aCD93 IgGs AD169- 85; AD169-309 and AD169-324 or control IgG.

[0400] Immunofluorescent staining:

[0401] MS1 were immunofluorescent stained with phalloidin-555 to detect the actin cytoskeleton as described in Example 4. Cells were analysed under fluorescent microscope (DMi8, Leica).

[0402] Endothelial cell-to-cell junction analysis:

[0403] Disruption of the cell junctions in MS1 cells was assessed as described in Example 18. Results

[0404] Similar to the results observed in HDBEC (Example 18), MS1 cells treated with aCD93 antibodies AD169-85, AD169-309 and AD169-324 showed a compromised endothelial monolayer with formation of gaps between adjacent cells. In contrast, intact monolayers without gaps were observed in cells treated with the control IgG (Figure 17A). The increased gap formation in cell treated with aCD93 antibodies is shown by quantification of the gap area in comparison with the control IgG treated cells in Figure 17B.

[0405] Conclusions

[0406] These data show that CD93-blocking antibodies disrupt mouse endothelial cell-cell junctions. In addition, these results indicate that the selected human / mouse cross-reactive aCD93 antibodies had similar effects in human and mouse endothelial cells.

[0407] Example 20, Assessment of the aCD93 IqGs effect on human endothelial cell migration Downregulation of CD93 inhibits human endothelial cell migration in vitro (Langenkamp E. et al Cancer Res. 2015 Nov 1;75(21):4504-16). Here we investigate whether blocking CD93 functions on human endothelial cell by using aCD93 IgGs recapitulates the inhibition of cell migration observed in CD93 downregulated cells.

[0408] Material and Methods

[0409] Cell culture:

[0410] Human endothelial cells (HDBECs) were cultured as described in Example 4.

[0411] Wound healing assay:

[0412] Cell migration was investigated in a wound healing assay. HDBECs were cultured in IncuCyte ImageLock 96-well Plates (Essen Bioscience) until confluency. Cells were treated with 10pg / ml of aCD93 IgGs AD169-8; AD169-85; AD169-309 and AD169-324 or control IgG 2h prior to perform a linear scratch using the IncuCyte WoundMaker (Essen Bioscience). Images of the scratch were captured automatically every hour by IncuCyte scan (IncuCyte ZOOM live-cell analysis system, Essen Bioscience). Thereafter, the images were processed by defining a scratching mask and a cell confluence mask using IncuCyte Cell Migration Analysis Software. Cell migration was determined by the ability of the cells to close the scratch during 24 hours.

[0413] Results

[0414] Treatment with aCD93 antibodies significantly inhibit the migration of human endothelial cells. Indeed, as shown in Figure 18, at 24h after the scratch was made, the endothelial cells treated with control IgG showed about 90% of wound closure, in contrast only 45% to 60% wound closure was observed in the endothelial cells treated with aCD93 antibodies.

[0415] Conclusions

[0416] This data shows that aCD93 antibodies AD169-8; AD169-85; AD169-309 and AD169- 324 inhibit human endothelial cell migration in vitro.

[0417] Example 21 , Evaluation of the potential cytotoxic effect of the aCD93 IqGs Potential cytotoxic effect on endothelial cells upon treatment with aCD93 antibodies was assessed by analyzing apoptosis.

[0418] Material and Methods

[0419] Cell culture:

[0420] Human endothelial cells (HDBECs) were cultured as described in Example 4. Mouse endothelial cells (MS1) were cultured as described in Example 14.

[0421] Apoptosis assay:

[0422] Apoptosis was analyzed in endothelial cells by detecting the apoptotic marker cleaved- Caspase 3 (cCasp3) by immunofluorescent staining. HDBEC and MS1 cells were treated with 10pg / ml aCD93 IgGLALA antibodies AD169-8; AD169-85; AD169-309 and AD169-324 or control IgG during 24 hours. Thereafter, cells were fixed with 4%PFA and stained for cleaved Caspase-3 (Cell Signaling Technology, #9664) followed by specific AlexaFluor-568 secondary antibody. Actin and nuclei were detected by phalloidin-647 and Hoechst staining. Images were taken under fluorescent microscope (DMi8, Leica). Treatment with 500ng / ml TNFa was used as positive control for cell apoptosis. Cells positive for cleaved Caspase-3 were normalized by the total cell number in each field of view.

[0423] Results

[0424] As shown in Figure 19, aCD93 IgGs treatment do not induce apoptosis in human or mouse endothelial cells after 24h of treatment (Figure 19A and Figure 19B respectively). Indeed, less than 1 % of the analysed human and mouse endothelial cells were positive for the apoptotic marker cleaved Caspase-3 (cCasp3).

[0425] Conclusion

[0426] This data shows that aCD93 antibodies AD169-8; AD169-85; AD169-309 and AD169- 324 do not induce apoptosis in human or mouse endothelial cells in vitro.

[0427] Example 22, Evaluation of the effect of aCD93 antibodies on perivascular invasion of tumor cell on mouse brain slices

[0428] In Example 7 we showed that CD93-deficiency inhibits glioma cell perivascular invasion in the mouse brain slices model. Here we assessed whether targeting CD93 by using the blocking antibodies AD169-85, AD169-309 and AD169-324 inhibit GL261 perivascular invasion, recapitulating the CD93-knockout phenotype.

[0429] Material and Methods Cell culture:

[0430] GL261 cells were cultured as described in Example 5. GL261 spheroids:

[0431] GL261 spheroids were obtained as described in Example 6.

[0432] Brain slice model and antibody treatment:

[0433] Mouse brain slices derived from wild-type mice were obtained as described in the Example 5. AD169-85, AD169-309 and AD169-324 treatment was performed on brain slices as described in Example 17.

[0434] Immunofluorescent staining:

[0435] Immunofluorescent staining of CD31 -positive vessels and detection of GFP signal from GL261 cells were performed as described in Example 5.

[0436] Results

[0437] Treatment with CD93 blocking antibodies showed a significant inhibition of the perivascular invasion of the GL261 cells on mouse brain slices. As shown by the immunofluorescence images in Figure 20A, GL261 cells seeded on control IgG treated brain slices invade the brain tissue by co-opting the vessels (arrowheads in Control IgG image, Figure 20A). In contrast, significantly less GL261 vessel-associated sprouts were observed in the brain slices treated with the aCD93 antibodies AD169-85, AD169-309 and AD169-324 (arrowheads in AD169-85, AD169-309 and AD169-324 images, Figure 20A). Quantification of GL261 co-opted vessels (indicated by arrowheads) in close proximity to the GL261 spheroid body (indicated by dotted line) showed a statistically significant inhibition of tumor cell invasion in the brain slices treated with AD169-85, AD169-309 and AD169-324 antibodies compared to the control IgG treated samples (Figure 20B).

[0438] Conclusion

[0439] This data indicates that CD93-blocking antibodies inhibit migration of GL261 cells along the vessels in mouse brain slices, recapitulating the inhibitory effect of the CD93-deficiency shown in Example 7.

[0440] Example 23, Assessment of the effect of aCD93 IqGs on perivascular invasion of tumor cell on human brain slices

[0441] In Example 22 we have showed that aCD93 antibodies treatment inhibits perivascular invasion of murine glioma cells in mouse brain slice model. Here we evaluate whether aCD93 antibodies blocks perivascular invasion of human glioma cells co-cultured on human brain slices derived from surgically resected brain tissue of glioblastoma patients.

[0442] Material and Methods Cell culture: U3013 human glioma cells were obtained from the human glioblastoma cell culture (HGCC) biobank (https: / / www.hqcc.se / ). U3013 cells were cultured in Primaria 60mm dishes (Corning, 353802) coated with laminin (Sigma, L2020) in Neurobasal medium and DMEM / F12 Glutamax 1 :1 ratio (Life technologies) supplemented with B27 1X (Life technologies 12587-010), N2 1X (Life technologies 17502-048), 10|jg / ml of human FGFBasic and human EGF (Peprotech). When confluent, U3013 cells were stained with 6uM fluorescent dye CellTracker Orange CMRA (ThermoFisher Scientific, C34551) following the manufacture instruction.

[0443] Human brain slices:

[0444] Freshly resected tissue from the tumor periphery, from glioblastoma patients undergoing brain surgery, was vibratome sectioned into 300pm thick slices in Hibernate-A Medium (A1247501 ; Gibco) supplemented with 13mM D-Glucose (Sigma-Aldrich), 30mM N-methyl-D- GLucamin (Sigma-Aldrich) and 1 mM GlutaMAX (Gibco). Brain slices were cultured on 12-well culture inserts 8um membrane (vWR; 7342736P) in Neurobasal A-Medium (10888022, Gibco) supplemented with 2% serum-free B-27 (50X) (17504001 , Gibco), 13mM D-glucose (Sigma- Aldrich), 1 mM MgSO4(Sigma-Aldrich), 15mM Hepes (Sigma-Aldrich) and 2mM GlutaMAX (Lot No. 1978435; Gibco) in presence of antibiotics. 10|jg / ml of aCD93 antibodies AD169-8, AD169- 85, AD169-309 and AD169-324 or Control IgG were added to the brain slice medium. After 2 hours of incubation with antibodies, U3013 cells (10 000 cells resuspended in 1 l of complete medium) were injected using 10 l Hamilton syringe in the brain slices. Brain slice medium containing aCD93 antibodies was replaced after 24h with a fresh medium supplemented with lOpg / ml of aCD93 antibodies. Brain slices were cultured for an additional 24h (for a total of 48h) at 37°C and 5% CO2 / 95% air in a humidified chamber. After this time period, brain slices were fixed with 4%PFA during 1.5 hours at room temperature and stained as described below.

[0445] Immunofluorescent staining:

[0446] Human brain slices were stained with monoclonal mouse ahuman CD31 (MO823 Dako). Sections were washed and incubated with specific mouse Alexa Fluor 488-conjugated secondary antibody (Invitrogen). Nuclei were visualized by Hoechst (Life Technologies) and the U3013 fluorescent signal was detected by the 555nm laser. Brain slices were analyzed under a confocal microscope (Leica SP8).

[0447] Results

[0448] U3013 glioma cells co-cultured on human brain slices showed perivascular invasion in the presence of control IgG treatment. Indeed as shown by the florescence image (Figure 21 , Control IgG), U3013 cells lined up along the CD31 positive vessels, exhibiting an elongated shape (arrowheads in U3013 Control IgG image) indicating a migratory phenotype. In contrast, U3013 cells co-cultured on human brain slices treated with aCD93 antibodies AD169-8, AD169- 85, AD169-309 and AD169-324, showed a shift in the cell morphology from the elongated and migratory shape observed in the Control IgG treated brain slices towards a round and less invasive phenotype. Indeed, U3013 perivascular invasion was significantly inhibited in the presence of all four aCD93 antibodies as indicated by a scarcity of cells elongating along the CD31-positive vessels. Treatment with AD169-8 and AD169-309 antibodies induce an aggregation of tumor cells (arrowheads in AD169-8 and AD169-309 images, Figure 21), inhibiting the interaction with the vasculature. AD169-85 and AD169-324 treatment instead induced a change of the U3013 morphology towards a round single cell shape, indicating a less invasive phenotype (arrowheads in AD169-85 and AD169-309 images, Figure 21).

[0449] Conclusion

[0450] This data shows that all four CD93-blocking antibodies largely inhibit tumor cell perivascular invasion during co-culture in human brain slices derived from glioblastoma patients.

[0451] Example 24, In vivo distribution of aCD93 antibodies

[0452] Biodistribution of selected human-mouse cross-reactive aCD93 IgGs (AD169-85, AD169- 309 and AD169-324) was assessed in healthy and tumor bearing mice after intravenous administration.

[0453] Material and Methods

[0454] In vivo study:

[0455] GL261 cells (20 000 cells in 2pl of Dulbecco’s PBS) were orthotopically injected in the brain of C57BL / 6 mice. Tumor tissue was harvested 23 days after tumor implantation and brain, kidneys, liver and spleen were harvested and cryosectioned for immunofluorescent staining. 5mg / Kg of aCD93 antibodies (AD169-85, AD169-309 and AD169-324) as well as control IgG were injected into the tail vein 24 hours before sacrificing the mice and collecting organs.

[0456] Immunofluorescent staining and imaging:

[0457] Cryosectioned tissues were immunofluorescent stained for anti-kappa-AF488 (SouthemBiotech, ref 2060-30) to detect aCD93 antibodies and control IgG and for CD31 (2H8; ThermoFisher Scientific, MA3105) to visualize blood vessels. Images were taken under confocal microscope (Leica SP8).

[0458] Results

[0459] Biodistribution analysis of the intravenously injected crCD93 antibodies AD169-85, AD169-309 and AD169-324 showed that all three antibodies were able to reach the tumor site and bind with high affinity to the tumor associated vessels after 24 hours circulation (Figure 22). Indeed, as showed by Figure 22A, positive signal of the aCD93 antibodies (anti-kappa) was found in the GL261 tumor core colocalizing with the vessel marker CD31. In contrast, negligible anti-kappa signal was detected in the tumor core of mice injected with control IgG (Figure 22A). Similarly, analysis of the tumor invasive front (dotted line in Figure 22B) showed a strong positive signal of the aCD93 antibodies in the vessels at the tumor border (arrowheads in Figure 22B). Importantly, minimal aCD93 antibody signal was detected in the vessels at the tumor contralateral hemisphere (Tumor CLH, Figure 22C) or in brain tissue from healthy mice (Figure 22D), suggesting that aCD93 antibodies target tumor vessels with high affinity. Quantification of the percentage of anti-kappa signal relative to the vessel marker CD31 in brain as well as in kidneys, liver and spleen is showed in the graph in Figure 22E. Interestingly, signal from all three tested antibodies was covering between 43-85% of the total vessel area in the tumor core as well as in the tumor border. In contrast, signal from all three tested antibodies was covering equal or less than 12.8% of the vessels in the tumor contralateral hemisphere or in healthy mice. In addition, minimal signal of the aCD93 antibodies was detected in kidneys, liver and spleen of GL261 bearing mice (graph in Figure 22E).

[0460] Conclusion

[0461] This data shows that all three human-mouse cross reactive antibodies against CD93 are able to cross the impaired blood-barrier at the tumor site and bind with high affinity and specificity to tumor vessels. In addition, these results showed a minimal binding of the antibodies to brain tissue far from the tumor site as well as minimal binding to tissue including brain, kidney, liver and spleen of healthy individuals.

[0462] Example 25, In vivo assessment of the effect of aCD93 antibodies on perivascular invasion of tumor cells

[0463] In Example 24 it was demonstrated that aCD93 antibodies administered intravenously in tumor bearing mice bind specifically to the tumor vessels. In addition, in Example 22 and in Example 23 it was showed that aCD93 antibodies inhibit perivascular invasion of glioma cells in the ex vivo model of mouse and human brain slices.

[0464] In this example, it is evaluated whether intravenous administration of aCD93 antibodies converted to a murine lgG1 backbone blocks perivascular invasion of glioma cells in tumor bearing mice.

[0465] Materials and Methods In vivo study: GL261-GFP cells (20 000 cells in 2pl of Dulbecco’s PBS) were orthotopically injected in the brain of C57BL / 6 mice (10 mice / group). At day 7 after tumor injection 20mg / Kg of aCD93 antibodies AD169-85 and AD169-309 converted to a murine IgG 1 backbone or control IgGs were injected into the tail vein. The antibody administration was repeated at day 11 , day 15 and day 19 after tumor injection for a total of 4 intravenous injections. Tumor tissue was harvested at day 23 after tumor implantation and brains were collected and vibratome sectioned for immunofluorescent staining and tumor border analysis (see experiment timeline in Figure 23A).

[0466] Immunofluorescent staining and imaging:

[0467] Brain tissues were immunofluorescence stained for CD31 (2H8; ThermoFisher Scientific, MA3105) to visualize blood vessels. Nuclei were visualized by Hoechst (Life Technologies) and the GL261-GFP fluorescent signal was detected by the 488nm laser. Images were taken under confocal microscope (Leica SP8).

[0468] Results

[0469] Repeated intravenous administration of aCD93 antibodies AD169-85 or AD169-309 in GL261 bearing mice (Figure 23A) resulted in significant reduction of perivascular invasion of GL261 cells. As shown in Figure 23B, the tumor invasion area, indicated by the dotted line, was reduced in mice treated with aCD93 antibodies AD169-85 and AD169-309 as compared to the invasion area observed in the tumor border of the control IgG group. A quantification of the invasion area analysed in the tumor border is shown by the graph in Figure 23C.

[0470] Conclusion

[0471] This data shows that systemic administration of AD169-85 or AD169-309 aCD93 antibodies significantly inhibits perivascular invasion of tumor cells in GL261 glioma bearing mice.

[0472] Example 26, In vivo evaluation of the inhibitory effect of aCD93 antibodies on fibronectin fibrillogenesis in GL261 bearing mice

[0473] In Example 17 it was shown that aCD93 antibodies inhibit vessel-associated fibronectin fibrillogenesis in mouse brain slices co-cultured with GL261 cells.

[0474] In this example it is assessed whether in vivo administration of aCD93 antibodies converted to a mouse backbone affect fibronectin fibrillogenesis in GL261 bearing mice.

[0475] Materials and Methods

[0476] In vivo study:

[0477] The in vivo study was designed as described in Example 25.

[0478] Immunofluorescent staining and imaging: Tumor tissue was stained with anti-CD31 (2H8, Thermo Fisher Scientific, MA3105) to visualize blood vessels, anti-fibronectin (Abeam, ab2413) and nuclei were visualized by Hoechst (Life Technologies). Images were taken under confocal microscope (Leica SP8).

[0479] Results

[0480] Intravenous administration of aCD93 antibodies AD169-85 or AD169-309 in GL261 bearing mice showed that both antibodies significantly reduce the vessel-associated deposition of fibronectin in the tumor area compared with the levels observed in the control IgG treated mice (Figure 24A and quantification graph in Figure 24B).

[0481] Conclusion

[0482] This data shows that systemic administration of AD169-85 or AD169-309 aCD93 antibodies significantly inhibits fibronectin fibrillogenesis in GL261 glioma bearing mice.

[0483] Example 27, Assessment of the effects of aCD93 antibody treatment on endothelial activation, T-cell recruitment and response to aPD1 -blockade immunotherapy in GL261 bearing mice.

[0484] In Example 3 it was shown that CD93 deficiency induces endothelial activation and T cell recruitment in glioma-bearing mice. In this example it is evaluated whether systemic administration of aCD93 antibodies in GL261 -bearing mice recapitulates this phenotype.

[0485] Materials and Methods

[0486] In vivo study:

[0487] The in vivo study, aimed to analyze the levels of VCAM1 and the abundance of immune cells (CD8 and CD3 positive T-cells), was designed as described in Example 25.

[0488] For the immunotherapy treatment, a survival study was performed. Mice treated with Control IgG or AD169-309, as described in Example 25, were additionally treated with intraperitoneal injections of aPD1 antibody (clone:RMP1-14,Cat# BE0146, BioXCell, USA) or rat-lgG2 (used as an isotype control) three times (200pg / dose at day 9, day 13 and day 17 after tumor injection). Mice were sacrificed when they lost equal or more than 20% body weight or showed symptoms.

[0489] Immunofluorescent staining and imaging:

[0490] Tumor tissue was stained with anti-CD31 (2H8, Thermo Fisher Scientific, MA3105) to visualize blood vessels, anti-VCAM1 (R&D System, AF643), anti-CD8 (BD BioSciences, 550281) and anti-CD3 (BD BioSciences, 557869). Nuclei were visualized by Hoechst (Life Technologies). Images were taken under confocal microscope (Leica SP8).

[0491] Results Analysis of tumor tissue after systemic delivery of aCD93 antibodies in GL261 -bearing mice showed that mice receiving AD169-309 antibodies displayed increased expression of the vascular adhesion molecule VCAM1 in tumor vessels as compared to the VCAM1 levels observed in the group treated with AD169-85 or with the control IgG (Figure 25A and quantification graph in Figure 25B). The increase in endothelial activation observed in GL261 tumors treated with AD169-309 aCD93 antibodies was accompanied by an increased abundance of T-cells. Indeed, as shown in Figure 25C-D and Figure 25E-F the presence of CD8 positive T-cells as well as CD3 positive T-cells were increased in the AD169-309-treated group as compared to the AD169-85-treated or the control IgG groups. Moreover, as shown by the survival graph in Figure 25G, combination therapy with the CD93 blocking antibody AD169-309 and aPD1 -blockade immunotherapy shows a significant improved survival as compared with the Control IgG mice receiving aPD1 immunotherapy (Control lgG-aPD1)

[0492] Conclusion

[0493] This data shows that systemic treatment of the aCD93 antibody AD169-309 in GL261 tumor bearing mice is associated with a significant induction of tumor endothelial activation, as indicated by increased VCAM1 expression in the tumor vessels, and promotes cytotoxic T cell infiltration. In addition, the CD93-blocking antibody AD169-309 prolonged survival of mice when administered in a combination with aPD1 immunotherapy, indicating that AD169-309 improves the efficacy of immunotherapy.

[0494] Example 28, Evaluation of the effect of aCD93 treatment on vessel permeability in GL261 -tumor bearing mice.

[0495] In Examples 18 and 19 it was shown that in vitro treatment of human and mouse endothelial cells with aCD93 antibodies disrupts endothelial cell-to-cell junctions suggesting a potential effect in the vessel permeability. Therefore, here we assess if systemic administration of aCD93 IgGs in GL261 -tumor bearing mice increase vessel permeability by loosening the endothelial cell-to cell junctions.

[0496] Materials and Methods

[0497] In vivo study:

[0498] The in vivo study was designed as described in Example 25. Immunofluorescent staining and imaging:

[0499] Tumor tissue was stained with anti-CD31 (2H8, Thermo Fisher Scientific, MA3105) to visualize blood vessels, anti-VE-Cadherin (BD BioSciences, 555289) and anti-fibrinogen (Dako, A0080). Nuclei were visualized by Hoechst (Life Technologies). Images were taken under confocal microscope (Leica SP8).

[0500] Results

[0501] The effects of in vivo treatment with aCD93 antibodies on endothelial junctional integrity was assessed in GL261 tumors by immunofluorescent staining to visualize the endothelial junction molecule VE-cadherin. As showed in Figure 26A and by the quantification graph in Figure 26B, AD169-85 treatment significantly reduced the VE-Cadherin signal in the tumor vessels. In contrast, the tumors in the AD169-309-treated group did not show any effect on VE- cadherin levels as compared to the control IgG group (Figure 26A-B).

[0502] In line with this result, increased extravasation of endogenous fibrinogen was observed in mice treated with AD169-85 as compared to the control IgG group (Figure 26C-D). In contrast, AD169-309 treated mice did not show any increase in fibrinogen leakage as compared to the control group (Figure 26C-D).

[0503] Conclusion

[0504] These data show that treatment with the aCD93 antibody AD169-85 affects the endothelial junctional integrity and promotes vessel permeability in GL261 -bearing mice. This observation suggest that AD169-85 treatment may facilitate the delivery of some anti-tumor drugs by increasing vessel permeability.

[0505] Example 29, Epitope binning of hlqG1 LALA antibodies and characterization of their interaction with CD93 and interaction partners

[0506] CD93 blocking antibodies AD169-8, AD169-85, AD169-309 and AD169-324 were binned against each other and against previously identified interaction partners of CD93. Binding was also tested against different truncation constructs of CD93 to determine their approximate binding site.

[0507] Materials and Methods

[0508] Truncation and production of hCD93 constructs

[0509] Constructs used in this example are shown in Figure 27. hCD93 and hCD93 lectin constructs were ordered as synthetic genes with a from GeneArt (Thermo Scientific). hCD93 (aa 1-580) had a C-terminal aviHis tag, whereas hCD93lectin (aa 1-258) had an N-terminal aviHis tag followed by a TEV protease site. Constructs hCD93_470, hCD93_344, hCD93EGF and hEGF1-2 were produced by PCR from hCD93; all have a C-terminal aviHis tag. hCD93 truncation constructs all had a gp67 signal sequence and were expressed as secreted protein in baculovirus-infected insect cells. Virus was generated with the flashBac system (Oxford Expression Technologies). After ca 5 days, recombinant virus was harvested. Virus was then amplified by infecting Sf9 cells with first generation virus. After 3-4 days, the virus was harvested, and the amplification was repeated. The third-generation virus was used for expression. For expression of hCD93, Sf9 cells were expanded to 1 -2 L of 1 .5-2 million cells / ml. Virus and biotin were added to the medium and the medium containing secreted hCD93 was harvested after 92-96h. The protein was captured from the medium by immobilized metal affinity chromatography (IMAC) in 50 mM Tris pH 7.8, 300 mM NaCI, 20 mM imidazole and eluted with 250 mM imidazole. The protein was further purified by either size exclusion chromatography in TBS + 5% glycerol or capture to Streptavidin mutein matrix (Roche) and elution with biotin. The mass and identity of the constructs was confirmed by mass spectrometry.

[0510] ELISA

[0511] To test binding of IgGs to different biotinylated CD93 constructs (Figure 27) by ELISA, streptavidin (SA; 1 mg / ml diluted in PBS) was directly immobilized into 384-well microtiter plates at 4°C overnight. Plates were washed four times between each incubation step with PBS + 0.05% Tween20 using automated plate washer. Biotinylated CD93 constructs (1 pg / ml, Table 1) were captured to SA coating for one hour at RT. Microtiter plate was blocked with blocking buffer (PBS + 0.5% BSA + 0.05% Tween20) for one hour at RT. IgG antibodies diluted to 1 pg / ml in blocking buffer were captured to immobilized antigens for one hour at RT. Detection of bound IgG was performed using a horseradish peroxidase (HRP)-conjugated anti-human IgG kappa antibody (Southern Biotech, #9230-05), followed by incubation with the chromogen Ultra TMB- ELISA (Thermo Scientific). The signal development was stopped by addition of 1 M sulphuric acid and the absorbance was measured at 450 nm.

[0512] SPR

[0513] The following reference antibodies were used: R3 (BD Bioscience), R139 (BD Bioscience), MM01 (Sino Biological), MM02 (Sino Biological). The reported CD93 interaction partner IGFBP7 was purchased from Aero Biosystems. A fragment of MMRN2 which has been reported to interact with CD93 (aa 495-674) was produced in house, expressed in E.coli with a C-terminal His tag and purified by IMAC in 50 mM Tris pH 7.8, 300 mM NaCI, 20 mM imidazole (elution with 250 mM imidazole). It was further purified by size exclusion chromatography on a Superdex 200 column (Cytiva) in 50 mM Tris pH 7.8, 150 mM NaCI. Binding to hCD93 was confirmed by SPR and ELISA and mass and identity were measured by mass spectrometry.

[0514] The sensor chip was prepared according to the manufacturer’s recommendations. An a- human Fab antibody mix was diluted to 50 pg / ml (1 :20) in immobilisation buffer and immobilised onto all 8 surfaces, including all 8 reference surfaces, of a CM5 series S chip. Approximately 7 000 RU was immobilized per surface.

[0515] Each AD169 antibody was diluted to 5 pg / ml in running buffer to obtain capture RU levels of approximately 400-500 RU. Once captured onto an a-human Fab surface, the remaining unbound a-human Fab antibodies were blocked by injection of Herceptin at 150 pg / ml. A single injection of hCD93 at 250 nM pre-incubated with an AD169 antibody at 50 nM was performed, followed by regeneration with 10 mM glycine-HCI, pH 2.1. Binding or no binding of the captured AD169 binders to CD93 pre-incubated with a second AD169 binder was compared with a cycle where only CD93 was injected. Reference antibodies R3, R139, MM01 and MM02 were only assayed as pre-incubations with CD93, not captured onto the chip surfaces (not compatible with immobilised capture antibody).

[0516] Results

[0517] ELISA to analyse binding of antibodies to various truncated CD93 protein variants

[0518] Binding of 15 aCD93 antibodies to truncations of CD93 was determined by ELISA. The antibodies could be sorted in three major groups: EGF1-2 binders, lectin-EGF1-2 binders, and lectin domain binders (Table 8). The EGF1-2 and lectin-EGF1-2 binders also bind to mCD93, whereas the lectin domain binders do not. AD169-33 binds only to hCD93, AD169-108 and 334 mainly bind to mCD93. The ELISA results were confirmed by SPR (data not shown).

[0519] Table 8

[0520] Epitope binning of antibodies and interaction partners

[0521] The epitope binning results were evaluated and grouped into bins of clones that shared similar binding patterns. The data suggested that clones AD169-8 does not compete with other binders for binding to CD93, whereas AD169-85, -309 and -324 compete with one another (Table 9). Reference antibodies R139 and MM02 did not compete with any of the AD169 clones for binding towards CD93. AD169-8 competed with reference R3 for binding and did not compete with MM01 .

[0522] Table 9

[0523] No blocking

[0524] Uncertain, may block

[0525] Blocking Conclusions

[0526] The binning experiment shows that clones AD169-85, -309 and -324 compete for binding to CD93, suggesting they bind to overlapping epitopes. AD169-8 does not compete with the other binders. Reference antibodies R139 and MM02 did not compete with any of the AD169 clones for binding towards CD93. AD169-8 competed with reference antibody R3 for binding and was the only clone not competing with MM01 .

[0527] ITEMIZED EMBODIMENTS

[0528] 1 . An agent comprising a binding moiety that specifically binds to vascular Cluster of Differentiation 93, CD93.

[0529] 2. The agent according to item 1 , wherein binding of said agent to CD93 results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation.

[0530] 3. The agent according to items 1 or 2, wherein the binding of the agent to CD93 provide an antagonistic effect by preventing functional activation of CD93 by its natural ligands.

[0531] 4. The agent according to items 1-3, wherein the agent or binding moiety thereof is a binding protein.

[0532] 5. A binding protein that specifically binds CD93 and comprises a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (VH) and a light chain variable domain (VL), each comprising three complementarity determining regions (CDRs), wherein the amino acid sequences of the CDRs are selected from the group comprising:

[0533] VHCDR1 as defined by SEQ ID NO: 1 ;

[0534] VHCDR2 as defined by SEQ ID NO: 2;

[0535] VHCDR3 as defined by SEQ ID NO: 3;

[0536] VLCDR1 as defined by SEQ ID NO: 4;

[0537] VLCDR2 as defined by AAS;

[0538] VLCDR3 as defined by SEQ ID NO: 5; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto, wherein said binding protein has a KD< 21 nM. The binding protein of item 5, wherein VHCDR1 , VHCDR2 and VLCDR2 are present next to specific framework amino acids, wherein the CDR and framework amino acid (faa) sequences are selected from the group comprising:

[0539] VHCDR1 and faa defined by SEQ ID NO: 6;

[0540] VHCDR2 and faa as defined by SEQ ID NO: 7; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto. The binding protein of item 1 , wherein the CDRs are individually selected from the groups comprising:

[0541] VHCDR1 is selected from SEQ ID NO: 8-10;

[0542] VHCDR2 is selected from SEQ ID NO: 11-14;

[0543] VHCDR3 is selected from SEQ ID NO: 15-18;

[0544] VLCDR1 as defined by SEQ ID NO: 4;

[0545] VLCDR2 as defined by AAS;

[0546] VLCDR3 is selected from SEQ ID NO : 19-21 ; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto. The binding protein of items 5 or 7, wherein the amino acid sequences of the CDRs are selected from the group comprising: i) a binding protein having

[0547] VHCDR1 as defined by SEQ ID NO: 8;

[0548] VHCDR2 as defined by SEQ I D NO: 11 ;

[0549] VHCDR3 as defined by SEQ ID NO: 15;

[0550] VLCDR1 as defined by SEQ ID NO: 4;

[0551] VLCDR2 as defined by AAS; and

[0552] VLCDR3 as defined by SEQ ID NO: 19; ii) a binding protein having

[0553] VHCDR1 as defined by SEQ ID NO: 8;

[0554] VHCDR2 as defined by SEQ ID NO: 12; VHCDR3 as defined by SEQ ID NO: 16;

[0555] VLCDR1 as defined by SEQ ID NO: 4;

[0556] VLCDR2 as defined by AAS; and

[0557] VLCDR3 as defined by SEQ ID NO: 20; iii) a binding protein having

[0558] VHCDR1 as defined by SEQ ID NO: 9;

[0559] VHCDR2 as defined by SEQ ID NO: 13;

[0560] VHCDR3 as defined by SEQ ID NO: 17;

[0561] VLCDR1 as defined by SEQ ID NO: 4;

[0562] VLCDR2 as defined by AAS; and

[0563] VLCDR3 as defined by SEQ ID NO: 20; iv) a binding protein having

[0564] VHCDR1 as defined by SEQ ID NO: 10;

[0565] VHCDR2 as defined by SEQ ID NO: 14;

[0566] VHCDR3 as defined by SEQ ID NO: 18;

[0567] VLCDR1 as defined by SEQ ID NO: 4;

[0568] VLCDR2 as defined by AAS; and

[0569] VLCDR3 as defined by SEQ ID NO: 21 ; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto. The binding protein of item 6, wherein the sequences of the CDRs including the framework amino acids (faa) are selected from the group comprising: i) a binding protein having

[0570] VHCDR1 and faa defined by SEQ ID NO: 43;

[0571] VHCDR2 and faa as defined by SEQ ID NO: 46; and ii) a binding protein having

[0572] VHCDR1 and faa defined by SEQ ID NO: 43;

[0573] VHCDR2 and faa as defined by SEQ ID NO: 47; and iii) a binding protein having

[0574] VHCDR1 and faa defined by SEQ ID NO: 44;

[0575] VHCDR2 and faa as defined by SEQ ID NO: 48; and iv) a binding protein having

[0576] VHCDR1 and faa defined by SEQ ID NO: 45; VHCDR2 and faa as defined by SEQ ID NO: 49; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto.

[0577] 10. The binding protein of items 5 and 7-8 wherein the VH sequence comprises an amino acid sequence selected from the group consisting of SEQ ID NO: 22-25 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, and wherein the VL sequence comprises an amino acid sequence selected from the group consisting of SEQ ID NO: 26-29 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto.

[0578] 11 . The binding protein of item 10, wherein the CDR sequences comprise no variations in the amino acid sequence, or wherein the sequence variation of the CDR amino acid sequences is at most 5%, such as 4 %, 3 %, 2 %, 1 % or less.

[0579] 12. The binding protein of items 10-11 , wherein the amino acid sequences of the VH and VL are selected from the group comprising: i) a binding protein having

[0580] VH as defined by SEQ ID NO: 22; and

[0581] VL as defined by SEQ ID NO: 26; ii) a binding protein having

[0582] VH as defined by SEQ ID NO: 23; and

[0583] VL as defined by SEQ ID NO: 27; iii) a binding protein having

[0584] VH as defined by SEQ ID NO: 24; and

[0585] VL as defined by SEQ ID NO: 28; iv) a binding protein having

[0586] VH as defined by SEQ ID NO: 25; and

[0587] VL as defined by SEQ ID NO: 29; and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto. 13. The binding protein of items 5-12, wherein the binding protein is a monoclonal antibody, or an antigen-binding fragment selected from the group consisting of Fv fragments, Fab- like fragments, disulphide-bonded fragments and domain antibodies.

[0588] 14. The binding protein of item 13, wherein the Fv fragment is an scFv fragment.

[0589] 15. The binding protein of item 13, wherein the Fab-like fragment is a Fab or F(ab’)2fragment.

[0590] 16. The binding protein of items 5-13 and 15, wherein the binding protein is a monoclonal antibody of the lgG1 isotype, such as an lgG1 LALA antibody.

[0591] 17. The binding protein of items 13-16, wherein the binding protein comprises a heavy chain and a light chain, the heavy chain comprising the VH domain and three constant domains, CH1 , CH2, and CH3, wherein CH1 and CH2 are joined via a hinge region, and the light comprising the VL domain and constant domain, CL, wherein the amino acid sequences of the domains and hinge region are defined as:

[0592] CH1 as defined by SEQ ID NO: 30;

[0593] CH2 as defined by SEQ ID NO: 31 :

[0594] CH3 as defined by SEQ ID NO: 32:

[0595] CL as defined by SEQ ID NO: 33;

[0596] Hinge region as defined by SEQ ID NO: 34, and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto.

[0597] 18. A cell engineered to express a chimeric antigen receptor (CAR), wherein the CAR comprises an antigen-binding domain, a transmembrane domain connected to the antigen binding domain by a hinge region, and an intracellular domain optionally connected to one or more co-stimulatory domains, wherein the antigen binding domain comprises the binding protein of any one of items 5 to 17.

[0598] 19. The cell of item 18, wherein the cell is a human cell. The cell of items 18 or 19, wherein the cell is an immune effector cell, such as a T cell, an NK cell or a macrophage. A nucleic acid molecule encoding an agent or binding protein according to items 1-17. The nucleic acid molecule according to item 21 , wherein the nucleic acid molecule encodes a binding protein comprising a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (VH) and a light chain variable domain (VL). The nucleic acid molecule according to item 22, wherein the VH encoding sequence comprises a nucleotide sequence selected from the group consisting of SEQ ID NO: 35- 38 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, and wherein the VL encoding sequence comprises a nucleotide sequence selected from the group consisting of SEQ ID NO: 39-42 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto. The nucleic acid molecule according to item 23, wherein nucleotide sequence of the VH and VL are selected from the group comprising: i) a binding protein encoded by

[0599] VH as defined by SEQ ID NO: 35; and

[0600] VL as defined by SEQ ID NO: 39; ii) a binding protein encoded by

[0601] VH as defined by SEQ ID NO: 36; and

[0602] VL as defined by SEQ ID NO: 40; iii) a binding protein encoded by

[0603] VH as defined by SEQ ID NO: 37; and

[0604] VL as defined by SEQ ID NO: 41 ; iv) a binding protein encoded by

[0605] VH as defined by SEQ ID NO: 38; and

[0606] VL as defined by SEQ ID NO: 42; and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto. 25. The nucleic acid molecule according to any one of items 21-24, wherein the nucleic acid molecule comprises a nucleotide sequence in the form of mRNA to be expressed in mammalian cells.

[0607] 26. The nucleic acid molecule according to any one of items 21-25, wherein a nucleotide sequence of the nucleic acid molecule is inserted into a mammalian cell using the CRISPR / Cas9 method.

[0608] 27. A vector, such as an expression vector, comprising the nucleic acid molecule of items 21-24.

[0609] 28. The vector according to item 27, wherein the vector is a plasmid or a viral vector.

[0610] 29. An isolated host cell for the production of a binding protein comprising the vector of items 27-28.

[0611] 30. A pharmaceutical composition containing an agent of any one of items 1-4, a binding protein of any one of items 5-17, a cell of any one of items 18-20 or a nucleic acid molecule of any one of items 21-26, and a pharmaceutically acceptable carrier or excipient.

[0612] 31 . An agent of any one of items 1-4, a binding protein of any one of items 5-17, a cell of any one of items 18-20 or a nucleic acid molecule of any one of items 21-26, or a pharmaceutical composition of item 30, for use in therapy.

[0613] 32. The agent, binding protein, cell, nucleic acid molecule or pharmaceutical composition according to item 31 , for use in cancer therapy.

[0614] 33. The agent, binding protein, cell, nucleic acid molecule, or pharmaceutical composition according to item 32, for use in brain cancer therapy, such as glioma, including glioblastoma, therapy. 34. The agent, binding protein, cell, or pharmaceutical composition according to any one of items 31-33, for use in inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and / or iii) inhibition of tumor cell proliferation.

[0615] 35. The agent, binding protein, cell, nucleic acid molecule, or pharmaceutical composition according to any one of items 31-35, for use in combination with one or more additional treatment regime, wherein the additional treatment regime is selected from i) antineoplastic agents, ii) chemotherapy, iii) immunotherapy, including tumor targeting antibodies, iv) irradiation and v) pathway blockers for treating cancer, such as tyrosine kinase inhibitors or neutralizing antibodies towards growth factors / cytokines or their receptors.

[0616] 36. A method of treating a subject in need thereof, comprising administering a therapeutically effective amount of an agent of any one of items 1-4, a binding protein of any one of items 5-17, a cell of any one of items 18-20 or a nucleic acid molecule of any one of items 21-26, or a pharmaceutical composition of item 30.

[0617] 37. Use of an agent of any one of items 1 -4, a binding protein of any one of items 5-17, a cell of any one of items 18-20 or a nucleic acid molecule of any one of items 21 -26, or a pharmaceutical composition of item 30, for the treatment or prevention of cancer, preferably glioblastoma.

Claims

CLAIMS1 . An agent comprising a binding moiety that specifically binds to vascular Cluster of Differentiation 93, CD93, wherein binding of said agent to CD93 results in of one or more of i) inhibition of perivascular tumor cell migration, ii) inhibition of tumor cell invasion and iii) inhibition of tumor cell proliferation.

2. The agent according to claim 1 , wherein the agent or binding moiety thereof is a binding protein that comprises a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (VH) and a light chain variable domain (VL), each comprising three complementarity determining regions (CDRs), wherein the amino acid sequences of the CDRs are selected from the group consisting of:VHCDR1 as defined by SEQ ID NO: 1 ;VHCDR2 as defined by SEQ ID NO: 2;VHCDR3 as defined by SEQ ID NO: 3;VLCDR1 as defined by SEQ ID NO: 4;VLCDR2 as defined by AAS;VLCDR3 as defined by SEQ ID NO: 5; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto, wherein said binding protein has a KD< 21 nM.

3. The agent according to claim 2, wherein the CDRs are individually selected from the groups comprising:VHCDR1 is selected from SEQ ID NO: 8-10;VHCDR2 is selected from SEQ ID NO: 11-14;VHCDR3 is selected from SEQ ID NO: 15-18;VLCDR1 as defined by SEQ ID NO: 4;VLCDR2 as defined by AAS;VLCDR3 is selected from SEQ ID NO : 19-21 ; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto.

4. The agent according to any one of claims 2 or 3, wherein the amino acid sequences of the CDRs are selected from the group comprising:i) a binding protein havingVHCDR1 as defined by SEQ ID NO: 8;VHCDR2 as defined by SEQ I D NO: 11 ;VHCDR3 as defined by SEQ ID NO: 15;VLCDR1 as defined by SEQ ID NO: 4;VLCDR2 as defined by AAS; andVLCDR3 as defined by SEQ ID NO: 19; ii) a binding protein havingVHCDR1 as defined by SEQ ID NO: 8;VHCDR2 as defined by SEQ ID NO: 12;VHCDR3 as defined by SEQ ID NO: 16;VLCDR1 as defined by SEQ ID NO: 4;VLCDR2 as defined by AAS; andVLCDR3 as defined by SEQ ID NO: 20; iii) a binding protein havingVHCDR1 as defined by SEQ ID NO: 9;VHCDR2 as defined by SEQ ID NO: 13;VHCDR3 as defined by SEQ ID NO: 17;VLCDR1 as defined by SEQ ID NO: 4;VLCDR2 as defined by AAS; andVLCDR3 as defined by SEQ ID NO: 20; iv) a binding protein havingVHCDR1 as defined by SEQ ID NO: 10;VHCDR2 as defined by SEQ ID NO: 14;VHCDR3 as defined by SEQ ID NO: 18;VLCDR1 as defined by SEQ ID NO: 4;VLCDR2 as defined by AAS; andVLCDR3 as defined by SEQ ID NO: 21 ; and CDR sequences having 95 % or more, such as 96 %, 97 %, 98 %, 99 % or more, identity thereto.

5. The agent according to any one of claims 2-4, wherein the VH sequence comprises an amino acid sequence selected from the group consisting of SEQ ID NO: 22-25 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto,and wherein the VL sequence comprises an amino acid sequence selected from the group consisting of SEQ ID NO: 26-29 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, and wherein the CDR sequences comprise no variations in the amino acid sequence, or wherein the sequence variation of the CDR amino acid sequences is at most 5%, such as 4 %, 3 %, 2 %, 1 % or less.

6. The agent according to any one of claims 2-5, wherein the amino acid sequences of the VH and VL are selected from the group comprising: i) a binding protein havingVH as defined by SEQ ID NO: 22; andVL as defined by SEQ ID NO: 26; ii) a binding protein havingVH as defined by SEQ ID NO: 23; andVL as defined by SEQ ID NO: 27; iii) a binding protein havingVH as defined by SEQ ID NO: 24; andVL as defined by SEQ ID NO: 28; iv) a binding protein havingVH as defined by SEQ ID NO: 25; andVL as defined by SEQ ID NO: 29; and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto.

7. The agent according to any one of claims 2-6, wherein the binding protein is a monoclonal antibody, such as an lgG1 LALA antibody, or an antigen-binding fragment selected from the group consisting of Fv fragments, such as scFv fragments, Fab-like fragments, such as Fab or F(ab’)2fragments, disulphide-bonded fragments and domain antibodies.

8. A cell engineered to express a chimeric antigen receptor (CAR), wherein the CAR comprises an antigen-binding domain, a transmembrane domain connected to the antigen binding domain by a hinge region, and an intracellular domain optionally connected to one or more co-stimulatory domains, wherein the antigen binding domaincomprises the agent of any one of claims 2 to 7, or a cell engineered to express a CAR, wherein the antigen binding domains are produced in the cell to direct it towards tumor vessels, wherein the antigen binding domain comprises the agent of any one of claims 2 to 7.

9. A nucleic acid molecule encoding an agent according to any one of claims 1 -7.

10. The nucleic acid molecule according to claim 9, wherein the nucleic acid molecule encodes a binding protein comprising a binding domain of an antibody, the binding domain comprising a heavy chain variable domain (VH) and a light chain variable domain (VL), wherein the VH encoding sequence comprises a nucleotide sequence selected from the group consisting of SEQ ID NO: 35-38 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto, and wherein the VL encoding sequence comprises a nucleotide sequence selected from the group consisting of SEQ ID NO: 39-42 and sequences having 80 % or more, such as 85 %, 90 %, 95 % or more, identity thereto.11 . A vector, such as an expression vector, comprising the nucleic acid molecule according to any one of claims 9 or 10.

12. A pharmaceutical composition containing an agent of any one of claims 1 -7, a cell according to claim 8, or a nucleic acid molecule of any one of claims 9-10, and a pharmaceutically acceptable carrier or excipient.

13. An agent according to any one of claims 1-7, a cell according to claim 8, a nucleic acid molecule according to any one of claims 9-10, or a pharmaceutical composition according to claim 12, for use in therapy.

14. The agent, cell, nucleic acid molecule or pharmaceutical composition according to claim 13, for use in cancer therapy, such as brain cancer therapy, including glioma and glioblastoma therapy.

15. The agent, cell, nucleic acid molecule, or pharmaceutical composition according to any one of claims 13-14, for use in combination with one or more additional treatment regime,wherein the additional treatment regime is selected from i) treatment with antineoplastic agents, ii) chemotherapy, iii) immunotherapy, including tumor targeting antibodies, iv) irradiation and v) use of pathway blockers for treating cancer.

Images