Heteromultimeric proteins

Abstract

Described herein are heteromultimeric proteins comprising at least two different polypeptide chains. The polypeptides may comprise Cε3 domain and / or Cε4 domain derived from an immunoglobulin E, and a Cγ2 and / or a Cγ3 domain derived from an immunoglobulin G. The two polypeptide chains may preferentially heterodimerize to promote formation of the heteromultimer. The two polypeptide chains may bind to two different target molecules, e.g. a cancer antigen and a T cell antigen.

Description

[0001] HETEROMULTIMERIC PROTEINS FIELD OF THE INVENTION The present invention lies in the design of synthetic (non-naturally occurring) heteromultimeric proteins, including bispecific antibodies, together with their therapeutic use. 5 BACKGROUND TO THE INVENTION Immunoglobulin E (IgE) is a class (or “isotype”) of immunoglobulin (Ig) that has only been found in mammals. IgE is synthesised by plasma cells. As with all antibody classes, monomers of IgE consist of two larger, identical heavy chains (ε chain) and two identical light chains (which are common to all antibody classes), with the ε chain containing four Ig-like constant 10 domains (Cε1-Cε4). It is the nature of the heavy chains that differentiates the different immunoglobulin classes, with those of the IgE class being larger and more heavily glycosylated than the heavy chains of the more common IgG class. Each immunoglobulin chain is comprised of a series of tandemly arranged immunoglobulin domains. The N-terminal domains (one each on the light 15 and heavy chains) contain regions of highly variable sequence (the variable domains) that enable binding to a huge range of antigens. The remaining domains consist of highly conserved so-called constant (Fc) domains. The structural differences confer different biological activities among the classes of immunoglobulin due to the panoply of effector cells and factors that bind to the different 20 constant domains of each Ig class. The gamma chain of IgG binds to a broad family of receptors that include classical membrane-bound surface receptors, as well as atypical intracellular receptors and cytoplasmic glycoproteins. The membrane-bound surface receptors include FcγRI (CD64), FcγRIIa, FcγRIIb, FcγRIIIa (CD16) and FcγRIIIb. Similarly, the epsilon chain of IgE binds to a high affinity receptor, FcεRI and a lower affinity receptor FcεRII. The 25 differential expression of these various receptors on differing immune effector cells determines the type of immune response that can be generated by IgG and IgE. IgE is mostly known for its detrimental role in allergy, but several studies have long pointed towards a natural tumour surveillance function of this antibody isotype (Jensen-Jarolim E. et al (2008) Allergy 63: 1255-1266; Jensen-Jarolim E., Pawelec G. (2012) Cancer Immunol. 1

[0002] Immunother.61: 1355-1357). Pioneer studies with IgG and IgE antibodies of the same epitope specificity tested head-to-head revealed a higher potential of the IgE in terms of cytotoxicity (Gould H.J. et al (1999) Eur. J. Immunol.29: 3527-3537). IgE has evolved to kill tissue-dwelling multicellular parasites, endowing it with several key 5 features that make it ideal for use in the treatment of solid tumours, which also mostly reside in tissue. The epsilon constant region of IgE has a uniquely high affinity for its cognate receptor (FcεRI) on the surfaces of immune effector cells including macrophages, monocytes, basophils and eosinophils (Ka~ 1010 / M for FcεRI and Ka~ 108-109 / M for the CD23 trimer complex; Gould H.J., Sutton B.J. (2008) Nat. Rev. Immunol.8: 205-217). This interaction is up to 10,000- 10 fold greater than the affinity that the gamma chain of IgG has for its cognate receptors and this results in the majority of IgE molecules being permanently attached to the surface of immune effector cells (Fridman W.H. (1991) FASEB J.5: 2684-2690). Therefore, the latter are primed and ready to destroy cells expressing the antigen recognised by the IgE. As a result, IgE is able to permeate tissues more effectively than IgG and stimulate significantly greater levels of both15 antibody-dependent cell-mediated phagocytosis (ADCP) and antibody dependent cell- mediated cytotoxicity (ADCC), the two main mechanisms by which immune effector cells can kill tumour cells. Despite the potential advantages of IgE compared to IgG, all currently approved therapeutic antibodies are of the IgG class. Recently, bispecific IgG antibodies have become of particular 20 interest. Bispecific antibodies are capable of binding simultaneously to two different antigens, i.e. to a different target molecule via each of the two arms of the antibody. This means that, unlike in naturally-occuring antibodies, the heavy and light chain sequences in each arm of the antibody are not identical to one another (although sometimes engineered common light chains are used). In particular, the variable domains (especially the complementarity determining 25 regions or CDRs thereof) of one heavy chain / light chain pair direct specific binding to one target antigen, and the variable domains (CDRs) of a second heavy chain / light chain pair in antibody direct specific binding to a different target antigen. Bispecific IgG antibodies have been indicated for use in treating various diseases, including cancer. In such bispecific IgG antibodies, typically at least one arm of the antibody binds to a 30 tumour antigen. In some cases, the second arm of the antibody binds to a different antigen on the tumour cell. In this way, bispecific antibodies may have increased specificity for tumour cells, a reduced risk of off-target side-effects and may prevent the development of treatment 2

[0003] resistance (see e.g. Mazor, et al. (2017) Sci.Rep.7, 40098; Moores, et al. (2016) Cancer Res. 76, 3942–3953). Alternatively, the second arm of the antibody may bind to a molecule present on the surface of immune cells, e.g. a T cell antigen. In this way, the antibody can direct the immune cell to the tumour cell, and / or facilitate crosslinking of receptors on the immune cell, 5 leading to cytotoxic activity and target cell killing. In order to produce bispecific antibodies, it is essential that two different heavy chains (i.e. each responsible for binding, in combination with a respective light chain, to a different target antigen) heterodimerize to form a multimer. A number of technologies have been developed to faciliate heterodimerization of IgG heavy chains, including knobs-into-holes (KiH), 10 electrostatic engineering and leucine zippers (see e.g. Ridgway JB, et al. Protein Eng (1996) 9(7):617–21; Atwell S, et al. J Mol Biol (1997) 270(1):26–35; Merchant AM, et al. Nat. Biotechnol. (1998) 16(7):677–81; Gunasekaran K, et al. J Biol Chem (2010) 285(25):19637– 46). In the knobs-into-holes approach, specific amino acids of an IgG Cγ3 domain are engineered to form a knob in one heavy chain represented by a bulky amino acid, and a hole in 15 an IgG Cγ3 domain of an opposite heavy chain represented by a small amino acid. These mutations force the heterodimerization and formation of bispecific IgG antibodies. However, very little is known about how such technologies might be applied to IgE antibodies. Vukovic et al., J. Biol. Chem. (2022) 298(8) 102153 suggested increasing the target selectivity of IgE antibodies by using a bispecific approach. The authors paired a Fab fragment that bound 20 to mouse prostate-specific membrane antigen (mPSMA) with a single domain antibody that bound to mEGFR either by introducing KiH mutations into Cε2 domains or by a leucine zipper. Vukovic considered that like Cγ3 in IgG, the Cε2 domains of IgE are responsible for dimerization of the heavy chains and are, therefore, a target for KiH engineering (see also Cooke et al. MABS 2018, 10(8): 1248–1259). 25 WO2021 / 064152 discloses hybrid antibodies having binding capabilities for Fcε receptors and Fcγ receptors, which may be achieved e.g. by grafting heavy chain constant domain sequences (e.g. Cγ2 and Cγ3 domains) derived from IgG to IgE. These hybrid antibodies, termed IGEGs, may comprise e.g. IgE-IgG Fc fusions comprising Cε2, Cε3 and Cε4 domains followed by Cγ2 and Cγ3 domains. 30 However there is still a need for alternative heteromultimeric proteins having improved properties compared to those of the prior art. In particular, there is a need for heteromultimers 3

[0004] having one or more properties of an IgE and an IgG (e.g. comprising one or more IgE constant domains and one or more IgG constant domains) that are easy to produce, stable and capable of binding simultaneously to two or more targets. SUMMARY OF THE INVENTION 5 The present inventors surprisingly found that in embodiments of the present invention, stable and effective heteromultimers comprising IgE and IgG constant domains could be produced by selecting Cε3 and / or a Cε4 domains (preferably Cε4 domains) that heterodimerize with one another, optionally in combination with Cγ2 and / or Cγ3 domains (preferably Cγ3 domains) that heterodimerize with one another. In embodiments of the present invention, these 10 heteromultimers are capable of simultaneous specific binding to two different antigens. Accordingly, in one aspect the present invention provides a heteromultimer comprising at least a first polypeptide and a second polypeptide. Typically the first polypeptide and the second polypeptide each comprise a Cε3 domain and / or a Cε4 domain (preferably at least a Cε4 domain), or variants or functional fragments thereof. The Cε3 and / or Cε4 domains (preferably 15 the Cε4 domains) of the first and second polypeptides differ in amino acid sequence from one another. The Cε3 and / or Cε4 domains (preferably the Cε4 domains) of the first and second polypeptides may heterodimerize with one another, e.g. such that the first and second polypeptides form the heteromultimer. The first polypeptide and the second polypeptide each further comprise a Cγ2 domain and / or a 20 Cγ3 domain (preferably at least a Cγ3 domain), or variants or functional fragments thereof. In some embodiments, the Cγ2 and / or Cγ3 domains (preferably the Cγ3 domains) of the first and second polypeptides differ in amino acid sequence from one another. The Cγ2 and / or Cγ3 domains (preferably the Cγ3 domains) of the first and second polypeptides may heterodimerize with one another, e.g. such that the first and second polypeptides form the heteromultimer. 25 In some embodiments, the Cε3 and / or Cε4 domains (preferably the Cε4 domains) of the first and second polypeptides preferentially heterodimerize. For instance, heterodimerization of the Cε3 and / or Cε4 domains (preferably the Cε4 domains) may be energetically favourable compared to homodimerization of the Cε3 and / or Cε4 domains. By this it is meant that, for example, the Cε4 domain of the first polypeptide may preferentially heterodimerize with the 30 Cε4 domain of the second polypeptide (which comprises a different amino acid sequence), rather than homodimerizing with a Cε4 domain of another molecule of the first polypeptide 4

[0005] (which comprises the same amino acid sequence). For instance, the Cε4 domain of the first polypeptide may bind with higher affinity to the Cε4 domain of the second polypeptide than to a Cε4 domain of another molecule of the first polypeptide. Heterodimerization of the Cε4 domains may also be favoured by kinetic or other factors. 5 Similarly the Cε3 domain of the first polypeptide may preferentially heterodimerize with the Cε3 domain of the second polypeptide, rather than homodimerizing with a Cε3 domain of another molecule of the first polypeptide. Thus it will be understood that “heterodimerization of the Cε3 and / or Cε4 domains” refers to heterodimerization of one Cε3 domain with a different Cε3 domain or heterodimerization of one Cε4 domain with a different Cε4 domain, and not to 10 an association of a Cε3 domain with a Cε4 domain. In some embodiments, the Cγ2 and / or Cγ3 domains (preferably the Cγ3 domains) of the first and second polypeptides preferentially heterodimerize. For instance, heterodimerization of the Cγ2 and / or Cγ3 domains (preferably the Cγ3 domains) may be energetically favourable compared to homodimerization of the Cγ2 and / or Cγ3 domains. For instance, the Cγ3 domain 15 of the first polypeptide may bind with higher affinity to the Cγ3 domain of the second polypeptide than to a Cγ3 domain of another molecule of the first polypeptide. Heterodimerization of the Cγ3 domains may also be favoured by kinetic or other factors. Thus when the first and second polypeptides are produced together, typically heteromultimers comprising at least the first and second polypeptides are predominantly formed (rather than 20 multimers comprising e.g. at least two molecules of the first polypeptide, or at least two molecules of the second polypeptide). Preferably at least 50%, 60%, 70%, 80%, 90% or 95% of the formed proteins are heteromultimers, i.e. comprise at least the first and second polypeptides. Thus by “heteromultimer” it is meant a protein comprising at least two different polypeptide 25 chains, e.g. two, four or more polypeptides. For instance, in some embodiments, the heteromultimer may comprise the first polypeptide and second polypeptide, and optionally one or more immunoglobulin light chains or fragments thereof. Preferably the heteromultimer comprises the first polypeptide, the second polypeptide and two immunoglobulin light chains, which may be the same or different. Preferably the heteromultimer comprises two 30 immunoglobulin light chains which differ in amino acid sequence. Thus the first polypeptide and second polypeptide may heterodimerize via the Cε3 and / or Cε4 domains (preferably the 5

[0006] Cε4 domains), and optionally also via the Cγ2 and / or Cγ3 domains (preferably the Cγ3 domains), but may additionally homo- or heterodimerize with further polypeptides such as immunoglobulin light chains to form the heteromultimer. In some embodiments, the first polypeptide binds specifically to a first binding site or target 5 molecule, e.g. a first epitope or antigen. The second polypeptide may bind specifically to a second binding site or target molecule different from the first binding site or target molecule, e.g. a second epitope or antigen. For instance the first and second polypeptides may bind to two different epitopes on the same antigen, or the first and second polypeptides may bind to two different antigens. The first and second polypeptides may bind to the target molecules via 10 one or more immunoglobulin variable domains, or variants or fragments thereof (e.g. a chimeric, humanized or human variable domain). Preferably the first and / or second polypeptide further comprises an immunoglobulin heavy chain variable domain (i.e. a VH domain), or a variant or functional fragment thereof. The first and / or second polypeptide may bind to the target molecules alone (e.g. via a single-chain variable domain) or combination with 15 a light chain variable domain present in the heteromultimer. Thus in some embodiments the heteromultimer is a bispecific antibody. Typically the variable domain(s) of the heteromultimer bind to one or more target antigens useful in the treatment of cancer, e.g. to a cancer antigen (i.e. an antigen expressed selectively on cancer cells or overexpressed on cancer cells) or to an antigen that inhibits or suppresses 20 immune-mediated tumour cell killing. The first or second polypeptide may bind specifically to a cancer antigen such as HER2 / neu or folate receptor alpha. Alternatively, the first or second polypeptide may bind specifically to a T cell antigen, preferably CD3. In one embodiment, the first polypeptide binds specifically to a cancer antigen (e.g. HER2 or folate receptor alpha) and the second polypeptide binds specifically to a T cell antigen (e.g. CD3). 25 The first and second polypeptides differ in amino acid sequence at least in the Cε3 and / or Cε4 domains thereof, preferably in the Cε4 domains. In some embodiments first and second polypeptides differ additionally in amino acid sequence of the Cγ2 and / or Cγ3 domains (preferably the Cγ3 domains). Preferably the first and second polypeptides further differ in amino acid sequence in terms of a target-binding sequence, e.g. in terms of immunoglobulin 30 variable domains that bind to two different epitopes and / or antigens. The Cε3 and / or Cε4 domains, and / or the Cγ2 and / or Cγ3 domains, may comprise (at least) one, two, three, four, five or more differences in amino acid sequence with respect to one another, i.e. the two 6

[0007] sequences may differ in terms of (at least) one, two, three, four, five or more amino acid residues. In some embodiments, first and second polypeptides each comprise (i) a modified or engineered Cε3 and / or Cε4 domain, preferably a modified or engineered Cε4 domain; and / or 5 (ii) a modified or engineered Cγ2 and / or Cγ3 domain, preferably a modified or engineered Cγ3 domain. By “modified” or “engineered” it is typically meant that the modified or engineered Cε3 and / or Cε4 domain differs in amino acid sequence compared to a native or wild type Cε3 and / or Cε4 domain sequence, e.g. a native human Cε3 and / or Cε4 domain. Similarly the modified or engineered Cγ2 and / or Cγ3 domain differs in amino acid sequence compared to a 10 native or wild type Cγ2 and / or Cγ3 sequence, e.g. a native human Cγ2 and / or Cγ3 domain. The modified Cε3 and / or Cε4 domain of the first polypeptide may associate with or bind to the modified Cε3 and / or Cε4 domain of the second polypeptide. The modified Cγ2 and / or Cγ3 domain of the first polypeptide may associate with or bind to the modified Cγ2 and / or Cγ3 domain of the second polypeptide. For instance, a modified Cε4 (and / or Cγ3) domain of the 15 first polypeptide may associate with or bind to a modified Cε4 (and / or Cγ3) domain of the second polypeptide. The modified Cε3 and / or Cε4 domains (and / or Cγ2 and / or Cγ3 domains) typically associate via a modified or engineered interface to form the heteromultimer. In particular, amino acid mutations in the modified Cε3 and / or Cε4 domains (and / or Cγ2 and / or Cγ3 domains) may form part of the modified or enginered interface between the domains. 20 Preferably the Cε3 and / or Cε4 domains (and / or Cγ2 and / or Cγ3 domains) are each based on the same native or wild type Cε3 and / or Cε4 domain (and / or Cγ2 and / or Cγ3 domain) sequence, but comprise one or more pairs of mutations with respect thereto. For instance, the Cε4 (and / or Cγ3) domain of the first polypeptide may comprise a first mutation at a first residue, and the Cε4 (and / or Cγ3) domain of the second polypeptide may comprise a second mutation at a 25 second residue that binds to the first residue when the two Cε4 (and / or Cγ3) domains dimerize. In further embodiments, the first polypeptide comprises at least one engineered protuberance in the Cε3 and / or the Cε4 domain, and / or the Cγ2 and / or Cγ3 domains. The protuberance may be, for instance, a “knob” or large or bulky amino acid residue, which extends away from an outer surface of Cε3 and / or the Cε4 domain (preferably the Cε4 domain) and / or Cγ2 and / or 30 Cγ3 domain (preferably the Cγ3 domain). Typically the protuberance is engineered in that it is not present in a native or wild type Cε3 and / or Cε4 domain (and / or Cγ2 and / or Cγ3) sequence, e.g. a native human Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domain, at a 7

[0008] corresponding position. The protuberance (e.g. a larger or bulkier amino acid residue) may extend away from an outer surface of the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domain more than a corresponding structure (e.g. a smaller amino acid residue) present in a native or wild type Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domain sequence, e.g. a native human Cε3 5 and / or Cε4 (and / or Cγ2 and / or Cγ3) domain, at the same position. In further embodiments, the second polypeptide comprises at least one engineered cavity in the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domain, preferably the Cε4 (and / or Cγ3) domain. Typically the cavity is engineered in that it is not present in a native or wild type Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domain sequence, e.g. a native human Cε3 and / or Cε4 (and / or 10 Cγ2 and / or Cγ3) domain, at a corresponding position. The cavity may be, for instance, a “hole” formed by the presence of a smaller or less bulky amino acid residue at a region on the surface of the Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domain. Typically the cavity extends towards the interior of the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domain, from an outer surface thereof. 15 In some embodiments, the engineered protuberance and cavity preferentially associate at an engineered interface to promote heterodimerization of the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domains. For instance, the engineered protuberance and cavity may form a “knob- into-hole” binding pair at an interface between two Cε3 (and / or Cγ2) domains or an interface between two Cε4 (and / or Cγ3) domains. Thus the first and second polypeptides may comprise 20 at least one pair of knobs-into-holes mutations in the Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domains. Knobs-into-holes mutations are described with respect to IgG in e.g. US 8,216,805, Ridgway JB, et al. Protein Eng (1996) 9(7):617–21 and Atwell S, et al. J Mol Biol (1997) 270(1):26–35. In one embodiment the first polypeptide comprises at least one larger amino acid residue 25 substituted in the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domain, the second polypeptide comprises at least one smaller amino acid residue substituted in the Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domain; and the larger and smaller amino acid residues preferentially associate to promote heterodimerization of the Cε3 and / or Cε4 (and / or Cγ2 and / or Cγ3) domains. By “larger” and “smaller” it is typically meant that a sidechain of an amino acid 30 substituted in the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domain of the first polypeptide is larger or bulkier than a a sidechain of an amino acid substituted in the Cε3 and / or the Cε4 (and / or Cγ2 and / or Cγ3) domain of the second polypeptide. Amino acids suitable for forming 8

[0009] protuberances or “knobs” in the present invention preferably have large and hydrophobic side chains and include, for example, phenylalanine, tryptophan, tyrosine, isoleucine, leucine or valine (preferably phenylalanine or tryptophan). Amino acids suitable for forming cavities or “holes” preferably have smaller and hydrophobic or neutral side chains, e.g. glycine, serine, 5 alanine or threonine (preferably glycine). In embodiments of the present invention, it has surprisingly been found that modification of Cε4 domains, optionally in combination with modified Cγ3 domains, is particularly useful in promoting efficient and stable heterodimerization of the first and second polypeptide. Thus in one embodiment (i) the first polypeptide and the second polypeptide each comprise a Cε4 10 domain and optionally a Cγ3 domain; (ii) the Cε4 domains of the first and second polypeptides differ in amino acid sequence, and optionally the Cγ3 domains of the first and second polypeptides differ in amino acid sequence; and (iii) the Cε4 domains of the first and second polypeptides preferentially heterodimerize; and optionally the Cγ3 domains of the first and second polypeptides preferentially heterodimerize; such that the first and second polypeptides 15 form the heteromultimer. Thus in some embodiments, the first and second polypeptides may comprise at least two pairs of knobs-into-holes mutations. Preferably one pair of knobs-into-holes mutations is in the Cε4 domain and one pair of knobs-into-holes mutations is in the Cγ3 domain. In this way, preferential heterodimerization of both the IgE domains and the IgG domains within the first 20 and second polypeptides may promote efficient formation of the heteromultimer. In such embodiments, the “knobs” and “holes” in the heteromultimer may be in the “trans” or “cis” conformation. By “cis” it is meant that both “knobs” (e.g. engineered protuberances comprising a larger amino acid residue) are present on the same polypeptide, and both “holes” (e.g. engineered cavities comprising smaller amino acid residues) are present on the other 25 polypeptide. For instance, in one “cis” embodiment the first polypeptide comprises a “knob” in the Cε4 domain and a knob in the Cγ3 domain, and the second polypeptide comprises a “hole” in Cε4 domain and a hole in the Cγ3 domain. In the “trans” conformation, each polypeptide comprises one “knob” and one “hole”. For instance, in one “trans” embodiment the first polypeptide comprises a “knob” in the Cε4 domain and a “hole” in the Cγ3 domain, 30 and the second polypeptide comprises a “hole” in Cε4 domain and a “knob” in the Cγ3 domain. A diagrammatic representation of “cis” and “trans” conformation of bispecific IGEGs is shown in 33. 9

[0010] In some embodiments, the first and / or second polypeptides may comprise one or more further domains derived from an immunoglobulin, or variants or functional fragments thereof. Preferably the first and / or second polypeptides further comprise one or more IgE constant domains, or variants or functional fragments thereof. In one embodiment, the first and / or 5 second polypeptide comprises one or more heavy chain constant domains derived from an IgE antibody (e.g. derived from an ɛ heavy chain). Preferably the first and / or second polypeptide comprises at least Cε2, Cε3 and Cε4 domains. For instance, the first and / or second polypeptides may each further comprise (i) a Cε1 domain, (ii) a Cε2 domain or (iii) a Cε1 domain and a Cε2 domain, or variants or functional fragments 10 thereof. Preferably the first and / or second polypeptides each comprise Cε1, Cε2, Cε3 domain and Cε4 domains. The first and / or second polypeptides may further comprise one or more immunoglobulin variable domains, or variants or functional fragments thereof. In one embodiment, the first and / or second polypeptides each comprise an immunoglobulin E heavy chain. 15 In one embodiment, the first and / or second polypeptides may together form a tetrameric IgE having an Fc region comprising CH2, CH3 and CH4 domains derived from IgE (i.e. Cε2, Cε3 and Cε4 domains) in which one or more of the constant domains may include one or more amino acid substitutions (preferably in the Cε4 domains), preferably fused to a Cγ2 and / or a Cγ3 domain. The fragment crystallisable / constant region (Fc region) is the tail region of an 20 antibody that interacts with cell surface Fc receptors. This property allows antibodies to activate the immune system. In some embodiments, the first and / or second polypeptides may together form a hybrid antibody as described in WO2021 / 064152. Such hybrid antibodies may be referred to as “IGEGs”. For instance, the heteromultimer may be capable of binding to an Fcε receptor and 25 an Fcγ receptor. In this context, “binding” typically refers to binding of the hybrid antibody via one or more constant domains thereof, i.e. “binding” does not refer to specificity of the heteromultimer binding to target antigen via its variable domains. Typically the heteromultimer is capable of binding to and activating both an Fcε receptor and an Fcγ receptor, thereby transducing receptor signalling and effector functions in cells of immune system in 30 which these receptors are expressed. In one embodiment, the first and second polypeptides each comprise Cε2, Cε3, Cε4, Cγ2 and Cγ3 domains, or variants or functional fragments thereof. 10

[0011] The domains in the first and / or second polypeptide may be joined by any suitable attachment, link, graft, fixation or fusion. For example, the constructs may include all or part of the hinge region derived from IgG. It will be appreciated that all or part of the constant domain sequence may be used, as well as variants thereof. The immunoglobulin domains described herein may 5 be derived from any species, preferably a mammalian species, more preferably from human. The first and / or second polypeptides may further comprise a variable domain sequence that determines specific binding to one or more target antigen(s). Such variable domain sequences may be derived from any immunoglobulin isotype (e.g. IgA, IgD, IgE, IgG or IgM). In one embodiment, the variable domain sequence may be derived from IgE. In another embodiment, 10 the variable domain sequence may be derived from IgG, e.g. IgG1. Alternatively, the variable domains may comprise sequences derived from two or more different isotypes, e.g. the variable domain may comprise a partial sequence derived from IgE and a partial sequence derived from IgG1. In one embodiment, the first and / or second polypeptides comprise one or more complementarity-determining regions (CDRs) derived from an immunoglobulin isotype other 15 than IgE (e.g. IgA, IgD, IgG or IgM, for example IgG1), and one or more framework regions and / or constant domains derived from an immunoglobulin of the isotype IgE. The variable domains or portions thereof (e.g. the complementarity-determining regions (CDRs) or framework regions) may also be derived from the same or a different mammalian species to the constant domains present in the first and / or second polypeptide sequences. Thus, 20 the heteromultimer may be chimaeric, humanised or fully human. In some embodiments, the Cε4 domain of the first or second polypeptide comprises an amino acid sequence as defined in any of SEQ ID NO:185 or SEQ ID NO:195, or a variant or fragment thereof. Variants and fragments of SEQ ID NO:185 or SEQ ID NO:195 include sequences having at least 85%, 90%, 95% or 99% sequence identity with the sequence of SEQ ID NO:185 25 or SEQ ID NO:195, e.g. over at least 30, 50 or 100 amino acid residues of, or over the the full length of SEQ ID NO:185 or SEQ ID NO:195 and fragments of a similar length, provided that the sequence retains the functional properties of the Cε4 domain of SEQ ID NO:185 or SEQ ID NO:195. Typically such variants and fragments comprise an amino acid substitution at position 26 and / or 69 of the Cε4 domain (i.e. a substitution with respect to the wild type Cε4 30 sequence as defined in SEQ ID NO:189) as described below. 11

[0012] Preferably the first or second polypeptide comprises an amino acid substitution at position 26 and / or 69 of the Cε4 domain (e.g. as defined in SEQ ID NO:189). For instance, the first polypeptide may comprise a larger amino acid residue (e.g. F or W) at position 26 of the Cε4 domain, and / or the second polypeptide comprises a smaller amino acid residue (e.g. G) at 5 position 69 of the Cε4 domain. In one embodiment, the first polypeptide comprises an amino acid substitution A26W or A26F in the Cε4 domain (e.g. as defined in SEQ ID NO:189), and / or the second polypeptide comprises an amino acid substitution F69G in the Cε4 domain (e.g. as defined in SEQ ID NO:189); including variant thereof having at least 85%, at least 90%, at least 95% or at least 10 99% sequence identity thereto. In further embodiments, the first or second polypeptide comprises an amino acid substitution as described in Table 4 below. For instance, the first or second polypeptide may comprise a Cε4 domain comprising an amino acid substitution at one or more of the following positions: 9, 11, 13, 24, 26, 28, 30, 54, 56, 61, 67, 69, 71 or 73 (e.g. in SEQ ID NO:7). In some 15 embodiments, the a Cε4 domain comprises one or more of the following amino acid substitutions: Y9L, Y9T, F11V, T13Y, T13W, T24K, A26Y, L28A, L28T, Q30D, S54Y, T56Y, T61D, F67A, F67T, F69L, R71F, R71Y or E73K (e.g. in SEQ ID NO:7). In further embodiments, the first or second polypeptide comprises an amino acid substitution as described in Table 5 below. For instance, the first or second polypeptide may comprise a 20 Cε4 domain comprising an amino acid substitution at one or more of the following positions: 9, 10, 11, 13, 24, 26, 28, 30, 54, 56, 61, 67, 69, 71, 73, (e.g. in SEQ ID NO:7). In some embodiments, the a Cε4 domain comprises one or more of the following amino acid substitutions: Y9C, Y9L, Y9T, A10V, F11D, F11K, F11V, F11Y, T13C, T13Y, T13W, T24H, T24K, A26F, A26L, A26K, A26S, A26W, A26Y, L28A, L28E, L28T, Q30D, S54L, S54Y, 25 T56F, T56Y, T56W, T61D, F67A, F67T, F69G, F69L, F69V, R71F, R71Y or E73K (e.g. in SEQ ID NO:7). In other embodiments, the first or second polypeptide may comprise a Cε3 domain comprising an amino acid substitution at one or more of the following positions: 1, 65, 102, 103 or 105, (e.g. in SEQ ID NO:5). In some embodiments, the a Cε3 domain comprises one or more of the 30 following amino acid substitutions: D1K, N65K, R102W, S103I or T105I (e.g. in SEQ ID NO:5). The first or second polypeptide may also comprise a Cε2 domain comprising an amino 12

[0013] acid substitution, provided that at least one amino acid substitution is present in the Cε3 domain as described above. For instance, the first or second polypeptide may comprise a Cε2 domain comprising an amino acid substitution at one or more of the following positions: 1, 26 or 83, preferably G21D, P26G or L83A (e.g. in SEQ ID NO:4). 5 In other embodiments, the first or second polypeptide may comprise a Cε3 domain comprising an amino acid substitution as described in Table 5, e.g. at one or more of the following positions: 1, 26, 65, 102, 103 or 105, (e.g. in SEQ ID NO:5). In some embodiments, the Cε3 domain comprises one or more of the following amino acid substitutions: D1K, N65K, R102W, S103I or T105I (e.g. in SEQ ID NO:5). 10 In some embodiments, the first or second polypeptide may comprise a Cε2 domain comprising an amino acid substitution as defined in Table 5, in combination with one or more amino acid substitutions in a Cε3 and / or Cε4 domain as described above. Preferably, the first or second polypeptide may comprise a Cε2 domain comprising an amino acid substitution at one or more of the following positions: 21, 26 or 83, more preferably G21D, P26G or L83A (e.g. in SEQ 15 ID NO:4).In further embodiments, the Cγ3 domain of the first or second polypeptide comprises an amino acid sequence as defined in any of SEQ ID NO:188 or SEQ ID NO:196, or a variant or fragment thereof. Variants and fragments of SEQ ID NO:188 or SEQ ID NO:196 include sequences having at least 85%, 90%, 95% or 99% sequence identity with the sequence of SEQ ID NO:188 or SEQ ID NO:196, e.g. over at least 30, 50 or 100 amino acid residues of, or over 20 the the full length of SEQ ID NO:188 or SEQ ID NO:196 and fragments of a similar length, provided that the sequence retains the functional properties of the Cγ3 domain of SEQ ID NO:188 or SEQ ID NO:196. Typically such variants and fragments comprise an amino acid substitution at position 26 and / or 69 of the Cγ3 domain (i.e. a substitution with respect to the wild type Cγ3 sequence as defined in SEQ ID NO:190) as described below. 25 Preferably the first or second polypeptide comprises an amino acid substitution at position 26 and / or 67 of the Cγ3 domain (e.g. as defined in SEQ ID NO:190). For instance, the first polypeptide may comprise a larger amino acid residue (e.g. Y) at position 26 of the Cγ3 domain, and / or the second polypeptide comprises a smaller amino acid residue (e.g. T) at position 67 of the Cγ3 domain. 30 In one embodiment, the first polypeptide comprises an amino acid substitution T26Y in the Cγ3 domain (e.g. as defined in SEQ ID NO:190), and / or the second polypeptide comprises an 13

[0014] amino acid substitution Y67T in the Cγ3 domain (e.g. as defined in SEQ ID NO:190); including variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity thereto. In one embodiment, the first and second polypeptides together comprise a pair (or group) of 5 amino acid substitutions as defined in Table 4, i.e. the first polypeptide comprises a heavy chain 1 with the mutations defined in Table 4, and the second polypeptide comprises the corresponding heavy chain 2 with the mutations described in the same row in Table 4. In another embodiment, the first and second polypeptides together comprise a pair (or group) of amino acid substitutions as defined in Table 5, i.e. the first polypeptide comprises a heavy 10 chain 1 with the mutations defined in Table 5, and the second polypeptide comprises the corresponding heavy chain 2 with the mutations described in the same row in Table 5. Further preferred positions for mutations, as well as specific mutations and combinations of mutations in the first and second polypeptides, are described in Tables 6 and 7. Thus in further embodiments, the first or second polypeptide comprises a Cε4 domain comprising an amino 15 acid substitution at a position as defined in Table 6 or 7 (e.g. in SEQ ID NO:7), or first and second polypeptides together comprise a pair (or group) of amino acid substitutions as defined in any single row of Table 6 or Table 7 (e.g. in SEQ ID NO:7).In preferred embodiments, the first polypeptide comprises one or more of the following mutations, including any combination of two or more of the following mutations, in SEQ ID NO:7: 20 A10V, F11W, T13W, T13Y, A26L, A26W, A26Y, S54L, T56W, R71F, R71Y; and / or the second polypeptide comprises one or or more of the following mutations, including any combination of two or more of the following mutations, in SEQ ID NO:7: Y9T, A10V, F11I, F11Y, L28V, F67A, F67I, F67T, F69G, F69I, F69V. In particularly preferred embodiments, the first and second polypeptides together comprise Cε4 25 domains comprising one of the following groups of mutations in SEQ ID NO:7 (shown below as those present in [first polypeptide] / [second polypeptide]): i) [A10V, A26L, S54L, T56W] / [A10V, F11Y, F67A, F69V]; ii) [T13W] / [Y9T]; 14

[0015] iii) [A26Y] / [F69G]; iv) [R71Y] / [F67T]; or v) [R71F] / [F67T]. In further preferred embodiments, the first and second polypeptides together comprise Cε4 5 domains comprising one of the following groups of mutations in SEQ ID NO:7 (shown below as those present in [first polypeptide] / [second polypeptide]): (vi) [T13Y] / [Y9F, L28V]; (vii) [T13W] / [Y9F, L28V]; (viii) [A26Y] / [L28V, F69I]; 10 (ix) [A26W] / [L28V, F69I]; (x) [R71Y] / [L28V, F67I]; (xi) [R71F] / [L28V, F67I]; or (xii) [F11W, T13W] / [Y9F, F11I]. In further embodiments, the first or second polypeptide comprises an amino acid sequence as 15 defined in any of SEQ ID NO:s 180-189, 192-196, 200, 201, 205-207 or 209, or a variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity to any of SEQ ID NO:s any of SEQ ID NO:s 180-189, 192-196, 200, 201, 205-207 or 209. In preferred embodiments, the first or second polypeptide comprises an amino acid sequence as defined in any of SEQ ID NO:s 180, 185, 188, 192, 193, 194, 195, 196, 200, 201, 205, 206, 20 or 209, or a variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity to any of SEQ ID NO:s any of the above sequences. More preferred combinations of the first and second polypeptide sequences are described in the examples below. The heteromultimer may comprise one or more (preferably two) immunoglobulin light chains, 25 e.g. kappa or lambda light chains, preferably human kappa or lambda light chains. The immunoglobulin light chains may comprise any wild-type or native light chain sequences, or 15

[0016] synthetic or non-native light chain sequences. In some embodiments, the heteromultimer comprises two different immunoglobulin light chains. The two immunoglobulin light chains may each preferentially dimerize with a different immunoglobulin heavy chain. For instance, a first immunoglobulin light chain may preferentially dimerize with the first polypeptide, and 5 a second immunoglobulin light chain may preferentially dimerize with the second polypeptide. In this way, correct pairing of immunoglobulin heavy and light chains in the heteromultimer may be ensured. Methods for ensuring correct heavy and light chain pairing in immunoglobulins (e.g. IgG) are described, for example, in WO 2012 / 123949. In some embodiments, this may be achieved by 10 moving the position of a disulphide bridge that links the heavy and light chains present in one arm of the heteromultimer, typically from the constant domains into the variable domains. For instance, a (native) disulphide bond may link the first immunoglobulin heavy chain (e.g. comprising a “knob” mutation) and the first immunoglobulin light chain between the Cε1 domains and CL domains thereof. A (non-native) disulphide bond may link the second 15 immunoglobulin heavy chain (e.g. comprising a “hole” mutation) and second immunoglobulin light chain between the VH and VL domains. In other words, cysteine residues (that form a disulphide bond between the light and heavy chains) in the constant domains of one binding arm of the heteromultimer may be moved, in the other binding arm only, into the variable domains (to form a non-native disulphide bond between the light and heavy chains). Thus in 20 some embodiments, one of the first and second polypeptides comprises a (non-native) cysteine residue in a variable domain, e.g. which forms a disulphide bridge with a (non-native) cysteine residue in a variable domain of an immunoglobulin light chain in the heteromultimer. In alternative embodiments, the heteromultimer comprises two immunoglobulin light chains having the same amino acid sequence, i.e. the heteromultimer comprises a common light chain. 25 Methods for engineering bispecific heteromultimers with a common light chain are described, for example, in Krah et al. (2017), Protein Engineering, Design and Selection, 30(4):291–301; Shiraiwa et al. (2019), Methods 154: 10-20; Ching et al. (2021), MAbs. 13(1): 1862451; WO2016 / 079081 and US 2021 / 0101997. In some embodiments, the heteromultimer comprises an immunoglobulin light chain sequence 30 as defined in any of SEQ ID NO:s 191, 198, 203 or 208 or a variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity thereto. More preferred 16

[0017] combinations of immunoglobulin light chains, as well as particular combinations with immunoglobulin heavy chains in the heteromultimer, are described in the Examples below. In further embodiments, the heteromultimer comprises one or more of the following: (a) a variable region as defined in SEQ ID NO:16 comprising an amino acid substitution at 5 position 44, e.g. G44C; (b) a Cε1 domain as defined in SEQ ID NO:3 comprising an amino acid substitution at position 14, e.g. C14G; (c) a kappa light chain as defined in SEQ ID NO:18 comprising an an amino acid substitution and / or deletion at position 100 and / or 214, e.g. G100C and / or C214del; 10 (d) a variable region as defined in SEQ ID NO:21 comprising an amino acid substitution at position 43, e.g. G43C; (e) a kappa light chain as defined in SEQ ID NO:23 comprising an an amino acid substitution and / or deletion at position 99 and / or 213, e.g. A99C and / or C213del; (f) a variable region as defined in SEQ ID NO:26 comprising an amino acid substitution at 15 position 44, e.g. G44C; (g) a kappa light chain as defined in SEQ ID NO:28 comprising an an amino acid substitution and / or deletion at position 105 and / or 219, e.g. Q105C and / or C219del; (h) a variable region as defined in SEQ ID NO:31 comprising an amino acid substitution at position 44, e.g. G44C; 20 (i) a lambda light chain as defined in SEQ ID NO:33 comprising an an amino acid substitution and / or deletion at position 102 and / or 214, e.g. G102C and / or C214A; or a variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity thereto. In further embodiments, the heteromultimer comprises one or more of the following: 25 (a) a variable region as defined in SEQ ID NO:197 comprising an amino acid substitution at position 44, e.g. G44C; 17

[0018] (b) a Cε1 domain as defined in SEQ ID NO:182 comprising an amino acid substitution at position 14, e.g. C14G; (c) a kappa light chain as defined in SEQ ID NO:199 comprising an an amino acid substitution and / or deletion at position 100 and / or 214, e.g. G100C and / or C214del; 5 or a variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity thereto. In a further aspect, the present invention provides a pharmaceutical composition comprising a heteromultimer as defined above and a pharmaceutically acceptable excipient, diluent or carrier. Optionally, the composition may further comprise a therapeutic agent such as another 10 antibody or fragment thereof, aptamer or small molecule. The composition may be in sterile aqueous solution. The heteromultimer or pharmaceutical composition as defined above may be used, for example, in preventing or treating cancer, e.g. benign or malignant tumours. Expressed in another way, the invention encompasses use of a heteromultimer as described hereinabove in 15 the manufacture of a medicament for administration to a human or animal for treating, preventing or delaying cancer, e.g. benign or malignant tumours. In another aspect, the invention encompasses a method of preventing, treating and / or delaying cancer (e.g. benign or malignant tumours) in a mammal suffering therefrom, the method comprising administering to the mammal a therapeutically effective amount of the heteromultimer as described 20 hereinabove. The cancer may be e.g. melanoma, Merkel cell carcinoma, non-small cell lung cancer (squamous and non-squamous), renal cell cancer, bladder cancer, head and neck squamous cell carcinoma, mesothelioma, virally induced cancers (such as cervical cancer and nasopharyngeal cancer), soft tissue sarcomas, haematological malignancies such as Hodgkin's and non- 25 Hodgkin's disease and diffuse large B-cell lymphomas, ovarian cancer and breast cancer. It will be appreciated that the heteromultimer of the invention may be administered in the form of a pharmaceutically acceptable composition or formulation. In a further aspect, the present invention provides a nucleic acid sequence that encodes an amino acid sequence as defined in any of SEQ ID NOs.1-39, or any subset of amino acid 30 sequences described herein. 18

[0019] In a further aspect, the present invention provides an expression vector comprising one or more nucleic acid sequences as defined above, e.g. operably linked to a promoter suitable for expression in eukaryotic (especially mammalian) cells. In one embodiment the vector is a CHO vector (i.e. an expression vector suitable for expression of the first and / or second 5 polypeptide in Chinese Hamster Ovary (CHO) cells). In a further aspect, the present invention provides a host cell comprising one or more nucleic acid sequences or expression vectors as defined above. In one embodiment, the expression vector or host cell comprises a first nucleic acid sequence encoding a first polypeptide as described above, and a second nucleic acid sequence encoding 10 a second polypeptide as described above. Thus in some embodiments, the expression vector or host cell may produce at least the first and second polypeptides comprising Cε3 and / or the Cε4 domain which heterodimerize to form the heteromultimer (optionally in combination with one or more immunoglobulin light chains). In a further aspect, the present invention provides a method of producing a heteromultimer as 15 defined above, comprising culturing a host cell as defined above under conditions for expression of the first and second polypeptide sequences, and recovering the heteromultimer from the host cell culture. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 shows an experimental set up for Bio-Layer Interferometry (BLI) of bispecific 20 antibodies. Figure 2 shows how raw data from BLI experiments is normalised to baseline and analysed. Figure 3 shows simultaneous binding of the bispecific IgE antibodies trastuzumab x runimotamab and trastuzumab x tarlatamab to Her2 and CD3 in BLI biosensors (Her2 as first antigen, CD3 as second antigen). 25 Figure 4 shows simultaneous binding of the bispecific IgE antibodies trastuzumab x runimotamab and trastuzumab x tarlatamab to Her2 and CD3 in BLI biosensors (CD3 as first antigen, Her2 as second antigen). 19

[0020] Figure 5 shows simultaneous binding of the bispecific IgE antibodies V26 x runimotamab and V26 x tarlatamab to Her2 and CD3 in BLI biosensors (CD3 as first antigen, Her2 as second antigen). Figure 6 shows that the bispecific IgE antibodies trastuzumab x runimotamab and trastuzumab 5 x tarlatamab bind to FcεR1 in BLI biosensors. Figure 7 shows a diagrammatic representation of an ELISA for detecting simultaneous binding of anti-Her2 x anti-FRα bispecific antibodies to Her2 and FRα. Figure 8 shows a diagrammatic representation of an ELISA for detecting simultaneous binding of anti-Her2 x anti-CD3 bispecific antibodies to Her2 and CD3. 10 Figure 9 shows the results of an of an ELISA for anti-Her2 x anti-FRα bispecific antibodies. Figure 10 shows the results of an ELISA for anti-Her2 x anti-CD3 bispecific antibodies. Figure 11 shows a diagrammatic representation of a flow sandwich cell binding assay for detecting crosslinking of SKBR3 and Jurkat cells with anti-Her2 x anti-CD3 bispecific antibodies. 15 Figure 12 shows that the results of a a flow sandwich cell binding assay in which Blinatumomab x Trastuzumab (i.e. a bispecific IgE antibody that binds Her2 and CD3) induces high levels of crosslinking of Jurkat and SKBR3 cells. Figure 13 shows the results of flow cytometry analysis of SKBR3 and CD8+ T cells in an assay to detect SKBR3 survival following incubation with monospecific and bispecific 20 antibodies. Figure 14 shows that Blinatumomab x Trastuzumab (i.e. a bispecific IgE antibody that binds Her2 and CD3) reduces survival of SKBR3 cells in the presence of CD8+ T cells, compared to monospecific antibodies alone or in combination. Figure 15 shows that Blinatumomab x Trastuzumab (i.e. a bispecific IgE antibody that binds 25 Her2 and CD3) increases activation of CD8+ T cells in the presence of of SKBR3 cells, compared to monospecific antibodies alone or in combination. 20

[0021] Figure 16 shows that Blinatumomab x Trastuzumab (i.e. a bispecific IgE antibody that binds Her2 and CD3) requires target SKBR3 cells for T cell activation, in contrast to monospecific antibodies alone or in combination. Figure 17 to Figure 20 Show SDS-PAGE gels of harvested supernatants from Expi293F cells 5 containing different expression levels of heterodimeric IgE formed by mutations in both IgE heavy chains. Figure 21 to Figure 30 Show SDS-PAGE gels of harvested supernatants from Expi293F cells containing different expression levels of heterodimeric IgE formed by further groups of mutations in both IgE heavy chains, after a second round of optimization. 10 Figure 31 to Figure 32 Show SDS-PAGE gels of harvested supernatants from Expi293F cells highlighting examples of improvement in IgE heavy chain heterodimer formation after a second round of optimization. Figure 33 shows a diagrammatic representation of bispecific IgEG antibodies comprising two pairs of “knobs-into-holes” mutations, in “cis” and “trans” form. 15 Figure 34 shows the results of BLI experiments using an anti-Her2 x anti-FRα bispecific IGEG with knobs-into-holes mutations in cis form, binding to human FcεR1, FcγR1, and FcRn receptors. Figure 35 shows the results of BLI experiments using a bispecific IGEG with knobs-into-holes mutations in cis form, binding to human FcεR1, FcγR1, and FcRn receptors, compared to an 20 IGEG lacking knobs-into-holes mutations. Figure 36 shows the results of an of a sandwich ELISA for anti-Her2 x anti-FRα bispecific IGEGs, indicating that trastuzumab x MOv18 IGEG trans (construct 3) and V26 x MOv18 IGEG trans (construct 4) bind simultaneously to Her2 and FRα. Figure 37 shows the results of an of a sandwich ELISA for anti-Her2 x anti-FRα bispecific 25 IGEGs, indicating that trastuzumab x MOv18 IGEG cis (construct 1) binds simultaneously to Her2 and FRα. 21

[0022] Figure 38 shows the results of a sandwich ELISA for anti-Her2 x anti-CD3 bispecific IGEGs, indicating that trastuzumab x runimotamab IGEG cis (construct 2) binds simultaneously to Her2 and CD3. DETAILED DESCRIPTION OF THE INVENTION 5 As used herein, the singular forms “a”, “an”, and “the” include both singular and plural referents unless the context clearly dictates otherwise. The terms “comprising”, “comprises” and “comprised of” as used herein are synonymous with “including”, “includes” or “containing”, “contains”, and are inclusive or open-ended and do not exclude additional, non-recited members, elements or method steps. The term also 10 encompasses “consisting of” and “consisting essentially of”. Whereas the term “one or more”, such as one or more members of a group of members, is clear per se, by means of further exemplification, the term encompasses inter alia a reference to any one of said members, or to any two or more of said members, such as, e.g., any ^3, ^4, ^5, ^6 or ^7 etc. of said members, and up to all said members. 15 As used herein, the term “antibody” is used in its broadest sense and generally refers to an immunologic binding agent. The term “antibody” is not only inclusive of antibodies generated by methods comprising immunisation, but also includes any polypeptide, e.g., a recombinantly expressed polypeptide, which is made to encompass at least one complementarity-determining region (CDR) capable of specifically binding to an epitope on an antigen of interest. Hence, 20 the term applies to such molecules regardless whether they are produced in vitro or in vivo. An antibody may be a polyclonal antibody, e.g., an antiserum or immunoglobulins purified there from (e.g., affinity-purified). An antibody may be a monoclonal antibody or a mixture of monoclonal antibodies. Monoclonal antibodies can target a particular antigen or a particular epitope within an antigen with greater selectivity and reproducibility. By means of example 25 and not limitation, monoclonal antibodies may be made by the hybridoma method first described by Kohler et al 1975 (Nature 256: 495) or may be made by recombinant DNA methods (e.g., as in US 4,816,567). Monoclonal antibodies may also be isolated from phage antibody libraries using techniques as described by Clackson et al 19 1 (Nature 352: 624-628) and Marks et al 1991 (J. Mol. Biol.222: 581-597), for example. 22

[0023] The term antibody includes antibodies originating from or comprising one or more portions derived from any animal species, preferably vertebrate species, including, e.g., birds and mammals. Without limitation, the antibodies may be chicken, turkey, goose, duck, guinea fowl, quail or pheasant. Also without limitation, the antibodies may be human, murine (e.g., mouse, 5 rat, etc.), donkey, rabbit, goat, sheep, guinea pig, camel (e.g., Camelus bactrianus and Camelus dromaderius), llama (e.g., Lama paccos, Lama glama or Lama vicugna) or horse. A skilled person will understand that an antibody may include one or more amino acid deletions, additions and / or substitutions (e.g., conservative substitutions), insofar such alterations preserve its binding of the respective antigen. An antibody may also include one or 10 more native or artificial modifications of its constituent amino acid residues (e.g., glycosylation, etc.). Methods of producing polyclonal and monoclonal antibodies as well as fragments thereof are well known in the art, as are methods to produce recombinant antibodies or fragments thereof (see for example, Harlow and Lane, “Antibodies: A Laboratory Manual”, Cold Spring Harbour 15 Laboratory, New York, 1988; Harlow and Lane, “Using Antibodies: A Laboratory Manual”, Cold Spring Harbour Laboratory, New York, 1999, ISBN 0879695447; “Monoclonal Antibodies: A Manual of Techniques”, by Zola, ed., CRC Press 1987, ISBN 0849364760; “Monoclonal Antibodies: A Practical Approach”, by Dean & Shepherd, eds., Oxford University Press 2000, ISBN 0199637229; Methods in Molecular Biology, vol.248: “Antibody 20 Engineering: Methods and Protocols”, Lo, ed., Humana Press 2004, ISBN 1588290921). Hence, also disclosed are methods for immunising animals, e.g., non-human animals such as laboratory or farm, animals using (i.e., using as the immunising antigen) any one or more (isolated) markers, peptides, polypeptides or proteins and fragments thereof as taught herein, optionally attached to a presenting carrier. Immunisation and preparation of antibody reagents 25 from immune sera is well-known per se and described in documents referred to elsewhere in this specification. The animals to be immunised may include any animal species, preferably warm-blooded species, more preferably vertebrate species, including, e.g., birds, fish, and mammals. Without limitation, the antibodies may be chicken, turkey, goose, duck, guinea fowl, shark, quail or pheasant. Also without limitation, the antibodies may be human, murine (e.g., 30 mouse, rat, etc.), donkey, rabbit, goat, sheep, guinea pig, shark, camel, llama or horse. The term "presenting carrier" or "carrier" generally denotes an immunogenic molecule which, when bound to a second molecule, augments immune responses to the latter, usually through the 23

[0024] provision of additional T cell epitopes. The presenting carrier may be a (poly)peptidic structure or a non-peptidic structure, such as inter alia glycans, polyethylene glycols, peptide mimetics, synthetic polymers, etc. Exemplary non-limiting carriers include human Hepatitis B virus core protein, multiple C3d domains, tetanus toxin fragment C or yeast Ty particles. 5 The invention described herein resides in heteromultimers comprising one or more IgE constant domains, especially Cε3 and / or Cε4 domains, and one or more IgG constant domains, especially Cγ2 and / or a Cγ3 domains. Without being bound by theory, it has surprisingly been found that the Cε3 and / or Cε4 domains (especially the Cε4 domains) of an IgE heavy chain can be used to drive heterodimerisation, optionally when fused to Cγ2 and / or Cγ3 domains 10 (especially Cγ3 domains). Thus by introducing “knobs-into-holes” mutations into Cε4 domains of a pair of IgE heavy chains, optionally in combination with additional “knobs-into- holes” mutations in Cγ3 domains fused to the IgE heavy chains, stable and efficient bispecific antibodies can be produced. These multimers are capable of simultaneous binding to two different antigens, as well as to FcεRs and FcγRs. 15 In contrast to what was described previously, the novel heteromultimers of the present invention utilise mutations in the Cε3 and / or Cε4 domains, and optionally in the Cγ2 and / or a Cγ3 domains. Residues in the Cε4 domain were identified that are involved in interaction between the two heavy chains of IgE. Having identified such residues, amino acid substitutions were made that enabled heterodimerisation of IgE. These mutations are tailored to the unique 20 structure of IgE, and are not suggested by the prior art. In the novel heteromultimers of the present invention that comprise domains derive from both IgE and from IgG, such mutations in the Cε4 domain may optionally be used in combination with mutations in Cγ3 domains that enable heterodimerisation of IgG. The heteromultimers described herein are typically capable of binding to Fcɛ receptors, e.g. to 25 the FcɛRI and / or the FcɛRII receptors. Preferably the heteromultimer is at least capable of binding to FcɛRI (i.e. the high affinity Fcɛ receptor) or is at least capable of binding to FcɛRII (CD23, the low affinity Fcɛ receptor). Typically, the heteromultimers are also capable of activating Fcɛ receptors, e.g. expressed on cells of the immune system, in order to initiate effector functions mediated by IgE. For instance, 30 the heteromultimers may be capable of binding to FcɛRI and activating mast cells, basophils, monocytes / macrophages and / or eosinophils. 24

[0025] The sites on IgE responsible for these receptor interactions have been mapped to peptide sequences on the Cɛ chain and are distinct. The FcɛRI site lies in a cleft created by residues between Gln 301 and Arg 376 and includes the junction between the Cɛ2 and Cɛ3 domains (Helm, B. et al. (1988) Nature 331, 180183). The FcɛRII binding site is located within Cɛ3 5 around residue Val 370 (Vercelli, D. et al. (1989) Nature 338, 649-651). A major difference distinguishing the two receptors is that FcɛRI binds monomeric Cɛ, whereas FcɛRII will only bind dimerised Cɛ, i.e. the two Cɛ chains must be associated. Although IgE is glycosylated in vivo, this is not necessary for its binding to FcɛRI and FcɛRII. Binding is in fact marginally stronger in the absence of glycosylation (Vercelli, D. et al (1989) supra). 10 The heteromultimers are also typically capable of binding to Fcγ receptors, e.g. to the FcγRIIIa. In another embodiment, the heteromultimer binds to FcεRI. Preferably the heteromultimer binds to both FcγRIIIa and FcεRI. In some embodiments, the heteromultimer is capable of binding to a neonatal Fc receptor (FcRn), typically in addition to a Fcγ and / or Fcɛ receptor as described above. Typically, the heteromultimers are also capable of activating Fcγ receptors, 15 e.g. expressed on cells of the immune system, in order to initiate effector functions mediated by IgG. For instance the heteromultimers may be capable of binding to and activating Fcγ receptors, and / or activating cells of the immune system expressing such receptors (including e.g. monocytes / macrophages and / or natural killer cells). For instance, the heteromultimers described herein may be capable of binding to (e.g. human) 20 Fcγ receptors such as FcγRI (CD64), FcγRIIa, FcγRIIb, FcγRIIIa (CD16a) and / or FcγRIIIb (CD16b). In one embodiment the heteromultimers bind to FcγRI (CD64) and / or FcγRIIIa (CD16a). In another embodiment, the heteromultimers bind to FcγRI (CD64), FcγRIIIa (CD16a) and FcγRIIIb (CD16b). The heteromultimers may also bind to variants of FcγRIIIa (CD16a), e.g. human CD16a 176Phe and / or human CD16a 176Val. Preferably the 25 heteromultimer is at least capable of binding to FcγRI or is at least capable of binding to FcγRIIIa. In some embodiments, the heteromultimers may further bind to the neonatal Fc receptor (FcRn). The heteromultimers may bind to FcRn in a pH-dependent manner. For instance, the heteromultimer may have a higher affinity for FcRn at pH 6.0 than at pH 7.4. The neonatal Fc 30 receptor (FcRn) belongs to the extensive and functionally divergent family of MHC molecules. Contrary to classical MHC family members, FcRn possesses little diversity and is unable to present antigens. Instead, through its capacity to bind IgG and albumin with high affinity at 25

[0026] low pH, it regulates the serum half-lives of both of these proteins. IgG enjoys a serum half-life that is substantially longer than similarly-sized globular proteins, including IgE which does not bind to FcRn (approximately 21 days for IgG and <2 days for IgE). In addition, FcRn plays important role in immunity at mucosal and systemic sites through both its ability to affect the 5 lifespan of IgG as well as its participation in innate and adaptive immune responses. Thus, binding to Fcɛ receptors and related effector functions are typically mediated by the heavy chain constant domains of the antibody, in particular by domains which together form the Fc region of the antibody. The heteromultimers described herein typically comprise at least a portion of an IgE antibody e.g. one or more constant domains derived from an IgE, preferably 10 a human IgE. In particular embodiments, the antibodies comprise one or more domains (derived from IgE) selected from Cɛl, Cɛ2, Cɛ3 and Cɛ4. In one embodiment, the antibody comprises at least Cɛ3 and / or Cɛ4, more preferably at least Cɛ2, Cɛ3 and Cɛ4, preferably wherein the domains are derived from a human IgE. In one embodiment, the antibody comprises an epsilon (ɛ) heavy chain, preferably a human ɛ heavy chain. 15 Constant domains derived from human IgE, in particular Cɛ1, Cɛ2, Cɛ3 and Cɛ4 domains, are shown in SEQ ID NOs: 3, 4, 5 and 7 respectively. Nucleic acid sequences encoding these acid sequences may be deduced by a skilled person according to the genetic code. The amino acid sequences of other human and mammalian IgEs and domains thereof, including human Cɛl, Cɛ2, Cɛ3 and Cɛ4 domains and human ɛ heavy chain sequences, are known in the art and are 20 available from public-accessible databases. For instance, databases of human immunoglobulin sequences are accessible from the International ImMunoGeneTics Information System (IMGT®) website at http: / / www.imgt.org. As one example, the sequences of various human IgE heavy (ɛ) chain alleles and their individual constant domains (Cɛ1-4) are accessible at http: / / www.imgt.org / IMGT_GENE-DB / GENElect?query=2+IGHE&species=Homo+sapiens. 25 The heteromultimers described herein typically comprise at least a portion of an IgG antibody e.g. one or more constant domains derived from an IgG (e.g. an IgG1), preferably a human IgG. In particular embodiments, the antibodies comprise one or more domains (derived from IgG) selected from Cγl, Cγ2 and Cγ3. In one embodiment, the antibody comprises at least Cγ2, more preferably at least Cγ2 and Cγ3, preferably wherein the domains are derived from a 30 human IgG1 antibody. In one embodiment, the antibody further comprises a hinge region derived from IgG, e.g. IgG1. 26

[0027] Constant domains derived from human IgG, in particular Cγ2 and Cγ3 domains, are shown in SEQ ID NO:s 187 and 190 respectively. Nucleic acid sequences encoding these acid sequences can be deduced by a skilled person according to the genetic code. The amino acid sequences of other human and mammalian IgG constant domains, including human Cγ2 and Cγ3 domains 5 and hinge sequences, are known in the art and are available from public-accessible databases, as described above for IgE constant domains. The amino acid sequences of one or more IgE domain and one or more IgG domains may be linked directly or via a suitable linker. Suitable linkers for joining polypeptide domains are well known in the art, and may comprise e.g.1, 2, 3, 4, 5, 6, 7, 8, 9 or 10 amino acid residues. In 10 some embodiments, the linker sequence may comprise up to 20 amino acid residues. Binding of the heteromultimers to Fcε and Fcγ receptors may be assessed using standard techniques. Binding may be measured e.g. by determining the antigen / antibody dissociation rate, by a competition radioimmunoassay, by enzyme-linked immunosorbent assay (ELISA), or by Surface Plasmon Resonance (e.g. Biacore). Binding affinity may also be calculated using 15 standard methods, e.g. based on the Scatchard method as described by Frankel et al (1979) Mol. Immunol.16:101-106. In general, functional fragments of the sequences defined herein may be used in the present invention. Functional fragments may be of any length (e.g. at least 50, 100, 300 or 500 nucleotides, or at least 50, 100, 200, 300 or 500 amino acids), provided that the fragment retains 20 the required activity when present in the antibody (e.g binding to a Fcɛ and / or a Fcγ receptor). Variants of the amino acid and nucleotide sequences described herein may also be used in the present invention, provided that the resulting antibody binds Fcɛ and / or and Fcγ receptors. Typically such variants have a high degree of sequence identity with one of the sequences specified herein. 25 The similarity between amino acid or nucleotide sequences is expressed in terms of the similarity between the sequences, otherwise referred to as sequence identity. Sequence identity is frequently measured in terms of percentage identity (or similarity or homology); the higher the percentage, the more similar the two sequences are. Homologs or variants of the amino acid or nucleotide sequence will possess a relatively high degree of sequence identity when aligned 30 using standard methods. 27

[0028] Methods of alignment of sequences for comparison are well known in the art. Various programs and alignment algorithms are described in: Smith and Waterman (1981) Adv. Appl. Math. 2:482; Needleman and Wunsch (1970) J. Mol. Biol.48:443; Pearson and Lipman (1988) Proc. Natl. Acad. Sci. U.S.A.85:2444; Higgins and Sharp (1988) Gene 73:237; Higgins and Sharp 5 (1989) CABIOS 5:151; Corpet et al (1988) Nucleic Acids Research 16:10881; and Pearson and Lipman (1988) Proc. Natl. Acad. Sci. U.S.A.85:2444. Altschul et al (1994) Nature Genet. 6:119 presents a detailed consideration of sequence alignment methods and homology calculations. The NCBI Basic Local Alignment Search Tool (BLAST) (Altschul et al (1990) J. Mol. Biol. 10 215:403) is available from several sources, including the National Center for Biotechnology Information (NCBI, Bethesda, Md.) and on the internet, for use in connection with the sequence analysis programs blastp, blastn, blastx, tblastn and tblastx. A description of how to determine sequence identity using this program is available on the NCBI website on the internet. Homologs and variants of the specific polypeptide sequences described herein (e.g. a VL, VH, 15 CL or CH domain) typically have at least about 75%, for example at least about 80%, 90%, 95%, 96%, 97%, 98% or 99% sequence identity with the original sequence (e.g. a sequence defined herein), for example counted over at least 20, 50, 100, 200 or 500 amino acid residues or over the full length alignment with the amino acid sequence of the antibody or domain thereof using the NCBI Blast 2.0, gapped blastp set to default parameters. For comparisons of 20 amino acid sequences of greater than about 30 amino acids, the Blast 2 sequences function is employed using the default BLOSUM62 matrix set to default parameters, (gap existence cost of 11, and a per residue gap cost of 1). When aligning short peptides (fewer than around 30 amino acids), the alignment should be performed using the Blast 2 sequences function, employing the PAM30 matrix set to default parameters (open gap 9, extension gap 1 penalties). 25 Proteins with even greater similarity to the reference sequences will show increasing percentage identities when assessed by this method, such as at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, or at least 99% sequence identity. When less than the entire sequence is being compared for sequence identity, homologs and variants will typically possess at least 80% sequence identity over short windows of 10-20 amino acids, and may possess 30 sequence identities of at least 85% or at least 90% or 95% depending on their similarity to the reference sequence. Methods for determining sequence identity over such short windows are available at the NCBI website on the internet. One of skill in the art will appreciate that these 28

[0029] sequence identity ranges are provided for guidance only; it is entirely possible that strongly significant homologs could be obtained that fall outside of the ranges provided. Typically variants may contain one or more conservative amino acid substitutions compared to the original amino acid or nucleic acid sequence. Conservative substitutions are those 5 substitutions that do not substantially affect or decrease the affinity of an antibody to Fcɛ receptors. For example, a human antibody that binds Fcɛ may include up to 1, up to 2, up to 5, up to 10, or up to 15 conservative substitutions compared to the original sequence (e.g. as defined above) and retain specific binding to the Fcɛ receptor. The term conservative variation also includes the use of a substituted amino acid in place of an unsubstituted parent amino acid, 10 provided that the hetetomultimer binds Fcɛ. Non-conservative substitutions are those that reduce an activity or binding to Fcɛ receptors. Functionally similar amino acids which may be exchanged by way of conservative substitution are well known to one of ordinary skill in the art. The following six groups are examples of amino acids that are considered to be conservative substitutions for one another: 1) Alanine 15 (A), Serine (S), Threonine (T); 2) Aspartic acid (D), Glutamic acid (E); 3) Asparagine (N), Glutamine (Q); 4) Arginine (R), Lysine (K); 5) Isoleucine (I), Leucine (L), Methionine (M), Valine (V); and 6) Phenylalanine (F), Tyrosine (Y), Tryptophan (W). The domains described above (e.g. one or more IgE and / or IgG constant domains) are typically present in a heavy chain in the heteromultimer. The heteromultimer may further comprise one 20 or more light chains in addition to one or more heavy chain sequences as described herein. Antibodies are typically composed of a heavy and a light chain, each of which has a variable region, termed the variable heavy (VH) region and the variable light (VL) region. Together, the VH region and the VL region are responsible for binding the antigen recognized by the antibody. Typically, a naturally occurring immunoglobulin has heavy (H) chains and light (L) 25 chains interconnected by disulfide bonds. There are two types of light chain, lambda (λ) and kappa (k). Thus the heteromultimers typically comprise two IgE heavy chains and two light chains (e.g. joined by disulfide bonds). The heteromultimers described herein may bind specifically (i.e. via their variable domains or the complementarity determining regions (CDRs) thereof) to one or more target antigens useful 30 in treating cancer. For instance, the heteromultimers may bind specifically to one or more cancer antigens (i.e. antigens expressed selectively or overexpressed on cancer cells). The 29

[0030] novel combination of effector functions transduced via capability of binding to two different antigens in a bispecific IgE may enhance cytotoxicity, phagocytosis (e.g. ADCC and / or ADCP) and other cancer cell-killing function of immune system cells (e.g. monocytes / macrophages and natural killer cells). For example, the heteromultimers may bind specifically e.g. to EGF- 5 R (epidermal growth factor receptor), VEGF (vascular endothelial growth factor) or erbB2 receptor (Her2 / neu). One example of an antibody comprising variable domains that bind selectively to Her2 / neu is trastuzumab (Herceptin). In some embodiments, one or more of the variable domains and / or one or more of the CDRs, preferably at least three CDRs, or more preferably all six CDRs of the first and / or second 10 polypeptide may be derived from one or more of the following antibodies: alemtuzumab (SEQ ID NOs:40-45), atezolizumab (SEQ ID NOs:46-51), avelumab (SEQ ID NOs:52-57), bevacizumab (SEQ ID NOs:58-63), blinatumomab, brentuximab, cemiplimab, certolizumab (SEQ ID NOs:64-69), cetuximab (SEQ ID NOs:70-75), denosumab, durvalumab (SEQ ID NOs:76-81), efalizumab (SEQ ID NOs:82-87), iplimumab, nivolumab, obinutuzumab, 15 ofatumumab, omalizumab (SEQ ID NOs:88-93), panitumumab (SEQ ID Nos:94-99), pembrolizumab, pertuzumab (SEQ ID NOs:100-105), rituximab (SEQ ID NOs:106-111), or trastuzumab (SEQ ID NOs:112-117). In such embodiments, the variable domains present in the heteromultimer may comprise one or more of the CDRs, preferably at least three CDRs, or more preferably all six of the CDR 20 sequences from one of the antibodies listed in Table 1. 30

[0031] fsro bsni1 aL-muyparehTrecnaCni desusiedo31 bitnAfoslep maxErofsecneuqeSdic AoinT Y 2 T TF )D.s-TPIP56Y RT)3YN) SBievT T O,TeS.mm RbaIniA HGIE E(TN K E8(IT98T G D YNGlmm Y S GGIK)1GG)7G)3R RT.)9Y.GKG E ...)1S SS(N G G)7DDG.G G..)5S.. )10G N)70G)3D 1I MI -actilusriravietD D A4(Y4(GS5( .C K P P Y5(TVG7(D7(SPKFDYK.S9(N1(..1(N..1(QE ACrunoCDSIYIT N NI.ISDSQ WIKIHIK V QT..S Y P DYINGPTP Sg.e. ctdrnednF.deRIWSW GA Y NMIGN YISA A VYIYI indnetagntnumaabaidNopqsesertminimtYs 1F G N M G M N Y D YN G Y W Y T N YerrRDreuzli BAEN S W S D YDYSTDocCcnozII.H γ1R.T)leD C…0HIF4(WS)6M 4(MI)25.(T)…8Y 5( .D)46.(T)…0M 7(W)6H 7( .G)2Y 8(GS)8G 8( .S.S)49.(T)…00 .T)1(…601.K)21 eeactecF 1thhttrasedbaTFDYS FTTFV L G YSY F RTFT SISVSFTF (…I ( ecof F Ti nYNFraiddedoa2T G Y G G Y Y G G GsG G G Gteerkcpvoobti retrpybabab b b ardepanaHs1,niodom ib utzm ntu uzba a ablimlmuumabaiz umbaba abm maumum bababmu insuddieL- dDtPcuczliimxu luauvzil izlmu itzumxuzu teohtoibi-tdnorA mle ozAte ev aA Ave toBreteC CruDaf auE miOntarPeP it tRsaracTiedmtnn iealaia Ptnsrhtanebtelo vedanibmuaci coakcmoNd innomolbceR

[0032] In alternative embodiments, one or more of the variable domains and / or one or more CDRs, preferably at least three CDRs, or more preferably all six CDRs present in the first and / or second polypeptide, may be derived from one or more of the following antibodies: abciximab, adalimumab (SEQ ID NOs:118-123), aducanumab, aducanumab, alefacept, alirocumab, 5 anifrolumab, balstilimab, basiliximab (SEQ ID NOs:124-129), belimumab (SEQ ID NOs:130- 135), benralizumab, bezlotoxumab, brodalumab, brolucizumab, burosumab, cankinumab, caplacizumab, crizanlizumab, daclizumab (SEQ ID NOs:136-141), daratumumab, dinutuximab, dostarlimab, duplilumab, eclizumab, elotuzumab, emapalumab, emicizumab, epitinezumab, erenumab, etrolizumab, evinacumab, evolocumab, fremanezumab, 10 galcanezumab, golimumab, guselkumab, ibalizumab, idarucizumab, inebilizumab, infliximab (SEQ ID NOs:142-147), isatuximab, ixekizumab, lanadelumab, leronlimab, margetuximab, mepolizumab, mogamulizumab, muromonab, narsoplimab, natalizumab (SEQ ID NOs:148- 153), naxitamab, necitumumab, obiltoxaximab, ocrelizumab, omburtamab, palivizumab (SEQ ID NOs:154-159), ramucirumab, ranibizumab (SEQ ID NOs:160-165), reslizumab, 15 risankizumab, romosozumab, sarilumab, satralizumab, secukinumab, spartalizumab, sutimlimab, tafasitamab, tanezumab, teplizumab, teprotumumab, tildrakizumab, toclizumab, toropalimab, ustekinumab, vedolizumab or zalifrelimab. In such embodiments, the variable domains of the first and / or second polypeptide present in the heteromultimer may comprise one or more of the CDRs, preferably at least three CDRs, or20 more preferably all six of the CDR sequences from one of the antibodies listed in Table 2. 32

[0033] sesrtoA1 2B3 2 4 2 2 2 te gtn yeiseuverhutinne A0 ...6P2.8G4G0 6 2 cW:u RqeD T2C SL1(…2Y1(F3L1(G... 31(YY41(Y51(NT51(TG61(euqS- -A2 )3(933SdYiSG L N Y L …G … G.I… YeYs1.2.18731cV D D RSG N YteYGMSYPaYctia 51-71,secA Edo ibnaK83 / 5110neicinG Gm E G D GS stK Ro– :)DB 1(27OSWralA VSEFQ K FKFQFL K S F.s,TG,sb - uR D Q N K3ceQ Q HKD2A TPP D O D N A MAloAN I M.7C HY)N 9 YS)S 5 YK)N 1 YE)7AS)Y 3 K)Y 9 DK)5YT)DI -.1 QA. yal 62Mf-oldaeRDItD H1aCGS1(T.. 2DS1(A3T1(T.. 31(NI 41(T.. 41(…51(P.. 61(ES ecpis452nrG Y SK YG K EGg nimN N F GT S R N D Tin eudt :uotsWGA D Ydnqesaey)3(1JlE.TI PMPAYIIISPP W TopsRd ,s ano A GNIYDIWINIser2R D WrDCnas bAitale oc e ieM. nrebaH G Ne th rasetnMNIH MHIVT1 SMthe tic rbli idIo.isHH R M)W 8Y.. )A 4 NN)R 0.Y)6WH)Y T.. )MG)G Y.. )rasderp ginbtinitrhD AY11(R21(NF 31(.S 31(NS241(D8K41(ST451(H061(tekc ot nnaaltrA C D TFSTTFTFI IY G YNS F LTFarde cs an dioG G G GSFD G Yb sGinud denildocotatye oh tiis nomubdaob b b bm baba aba ama aammctiedm hlamehirlaicRfibu imtnmliiximlumu iimumzil ixizlAadsauziuv iziniehbsrttoa reotmneAaBleBcalfDnIta liNaPnaeRb temuac niibmo tmaNd inmoccfo erT

[0034] In other embodiments, one or more of the variable domains and / or one or more of the CDR sequences, preferably at least three CDRs, or more preferably all six CDRs present in the first and / or second polypeptide, may be derived from an anti-HMW-MAA antibody. In one embodiment, one or more of the variable domains and / or one or more of the CDR sequences, 5 preferably at least three CDRs, or more preferably all six CDRs may be derived from the anti- HMW-MAA antibody described in WO 2013 / 050725 (SEQ ID NOs:172 and 173 for the variable domain and SEQ ID NOs:166-171 for CDRs). HMW-MAA refers to high molecular weight-melanoma associated antigen, also known as chondroitin sulfate proteoglycan 4 (CSPG4) or melanoma chondroitin sulfate proteoglycan (MCSP) – see e.g. Uniprot Q6UVK1. 10 In such embodiments, the variable domains of the heteromultimer may comprise one or more of the CDR sequences, preferably at least three CDRs, or more preferably all six of the CDR sequences defined in Table 3. In other embodiments, one or more of the variable domains of the antibody comprises one or more of the variable domain sequences listed in Table 3. 15 Table 3. Estimated Variable Domains and CDR Sequences of an Anti-HMW-MAA Antibody 34

[0035] Compositions are provided herein that include a carrier and one or more heteromultimers, or functional fragments thereof. The compositions may be prepared in unit dosage forms for administration to a subject. The amount and timing of administration are at the discretion of 5 the treating physician to achieve the desired purposes. The antibody may be formulated for systemic or local (such as intra-tumour) administration. In one example, the antibody may formulated for parenteral administration, such as intravenous administration. The compositions for administration may include a solution of the heteromultimers, or a functional fragment thereof, dissolved in a pharmaceutically acceptable carrier, such as an 10 aqueous carrier. A variety of aqueous carriers may be used, for example, buffered saline and the like. These solutions are sterile and generally free of undesirable matter. These compositions may be sterilised by conventional, well known sterilisation techniques. The compositions may contain pharmaceutically acceptable auxiliary substances as required to approximate physiological conditions such as pH adjusting and buffering agents, toxicity 15 adjusting agents and the like, for example, sodium acetate, sodium chloride, potassium chloride, calcium chloride, sodium lactate and the like. The concentration of antibody in these formulations can vary widely, and will be selected primarily based on fluid volumes, viscosities, body weight and the like in accordance with the particular mode of administration selected and the subject’s needs. 35

[0036] A typical dose of the pharmaceutical composition for intravenous administration includes about 0.1 to 15 mg of heteromultimer per kg body weight of the subject per day. Dosages from 0.1 up to about 100 mg per kg per day may be used, particularly if the agent is administered to a secluded site and not into the circulatory or lymph system, such as into a body cavity or into a 5 lumen of an organ. Actual methods for preparing administrable compositions will be known or apparent to those skilled in the art and are described in more detail in such publications as Remington’s Pharmaceutical Science, 19th ed., Mack Publishing Company, Easton, Pa. (1995). Heteromultimers may be provided in lyophilised form and rehydrated with sterile water before 10 administration, although they are also provided in sterile solutions of known concentration. The antibody solution may be then added to an infusion bag containing 0.9% sodium chloride, USP, and typically administered at a dosage of from 0.5 to 15 mg / kg of body weight. Heteromultimers may be administered by slow infusion, rather than in an intravenous push or bolus. In one example, a higher loading dose may be administered, with subsequent, 15 maintenance doses being administered at a lower level. For example, an initial loading dose of 4 mg / kg may be infused over a period of some 90 minutes, followed by weekly maintenance doses for 4-8 weeks of 2 mg / kg infused over a 30 minute period if the previous dose was well tolerated. The heteromultimers described herein (or functional fragment thereof) may be administered to 20 slow or inhibit the growth of cells, such as cancer cells. In these applications, a therapeutically effective amount of a heteromultimer may be administered to a subject in an amount sufficient to inhibit growth, replication or metastasis of cancer cells, or to inhibit a sign or a symptom of the cancer. In some embodiments, the heteromultimers may be administered to a subject to inhibit or prevent the development of metastasis, or to decrease the size or number of 25 metasases, such as micrometastases, for example micrometastases to the regional lymph nodes (Goto et al (2008) Clin. Cancer Res.14(11):3401-3407). A therapeutically effective amount of the heteromultimer will depend upon the severity of the disease and the general state of the patient’s health. A therapeutically effective amount of the heteromultimer is that which provides either subjective relief of a symptom(s) or an objectively 30 identifiable improvement as noted by the clinician or other qualified observer. These compositions may be administered in conjunction with another chemotherapeutic agent, either simultaneously or sequentially. 36

[0037] Many chemotherapeutic agents are presently known in the art. In one embodiment, the chemotherapeutic agents may be selected from the group consisting of mitotic inhibitors, alkylating agents, anti-metabolites, intercalating antibiotics, growth factor inhibitors, cell cycle inhibitors, enzymes, topoisomerase inhibitors, anti-survival agents, biological response 5 modifiers, anti-hormones, e.g. anti-androgens, and anti-angiogenesis agents. All documents cited in the present specification are hereby incorporated by reference in their entirety. The invention will now be described in more detail by way of the following non- limiting examples. EXAMPLES 10 Example 1 - Design and cloning of bispecific IgE constructs In this example, human IgE bispecific antibodies were constructed. Each DNA encoding the antibody chain was synthesised and subsequently cloned into an expression vector pTWIST CMV Beta Globin (Twist Inc.) together with secretion signal peptide. For the heavy chain sequence, the secretion signal is as follows: 15 MEWSWVFLFFLSVTTGVHS (SEQ ID NO:176) For the light chain sequence, the secretion signal is as follows: MSVPTQVLGLLLLWLTDARC (SEQ ID NO:177) Each plasmid encoding the antibody chain is propagated in bacterial host strain DH5α. A single colony from each transformation reaction was picked and grown in LB media containing 20 ampicillin at 100 μg / mL. Each plasmid was purified using QIAGEN Plasmid Plus kit (Qiagen), following the manufacturer’s recommendation. IgE heavy chain Anti-Her2 arm (Trastuzumab) with “knob” mutation The following amino acid sequence (SEQ ID NO:1) describes an IgE heavy chain with anti- Her2 (Trastuzumab) variable domain and “knob” mutation in the Cε4 domain. The knob 25 mutation is A344W (numbering from the first residue of Cε1) or A26W (numbering from the first residue of Cε4) and is highlighted in bold and underlining. SEQ ID NO:1: 37

[0038] EVQLVESGGGLVQPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNGYTRYADSVKGRFTISADTS KNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDYWGQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLGCL ATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVC SRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTL 5 SQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTW SRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFA TPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVTRAEWEQKDEFIC RAVHEAASPSQTVQRAVSVNPGK 10 SEQ ID NO:1 comprises the following domains: Trastuzumab heavy chain variable region (SEQ ID NO:2): EVQLVESGGGLVQPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNGYTRYADSVKGRFTISADTS KNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDYWGQGTLVTVSS Cε1 (SEQ ID NO:3): 15 ASTQSPSVFPLTRCCKNIPSNATSVTLGCLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVS GAWAKQMFTCRVAHTPSSTDWVDNKTFS Cε2 (SEQ ID NO:4): VCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSEL TLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCA 20 Cε3 (SEQ ID NO:5): DSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWSRASGKPVNHSTRKEEKQRNGTLTVTSTLP VGTRDWIEGETYQCRVTHPHLPRALMRSTTKTS and Cε4 comprising mutation A26W (SEQ ID NO:6): GPRAAPEVYAFATPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVT 25 RAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK The original (wild type) Cε4 sequence is shown in SEQ ID NO:7: GPRAAPEVYAFATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVT RAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK Kappa light chain of anti-Her2 arm (Trastuzumab) 30 The following amino acid sequence (SEQ ID NO:8) describes the kappa light chain with anti- Her2 (Trastuzumab) variable domain. 38

[0039] DIQMTQSPSSLSASVGDRVTITCRASQDVNTAVAWYQQKPGKAPKLLIYSASFLYSGVPSRFSGSRSGTDFTLTI SSLQPEDFATYYCQQHYTTPPTFGQGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKV DNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC 5 IgE heavy chain of anti-Her2 arm (EPS226) with “knob” mutation The following amino acid sequence (SEQ ID NO:9) describes an IgE heavy chain with anti- Her2 (EPS226, also referred to herein as V26) variable domain and “knob” mutation in the Cε4 domain. The knob mutation is A344W (numbering from the first residue of Cε1) or A26W (numbering from the first residue of Cε4) and is highlighted in bold and underlining. 10 QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYAMHWVRQAPGQRLEWIGWINAGNGNTKYSQKFQGRVTITRDTS ASTAYMELSSLRSEDTAVYYCARDFSSQVATAAVDYWGQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLG CLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFS VCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSEL TLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNL 15 TWSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYA FATPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVTRAEWEQKDEF ICRAVHEAASPSQTVQRAVSVNPGK SEQ ID NO:9 comprises the following domains: 20 EPS226 heavy chain variable region (SEQ ID NO:10): QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYAMHWVRQAPGQRLEWIGWINAGNGNTKYSQKFQGRVTITRDTS ASTAYMELSSLRSEDTAVYYCARDFSSQVATAAVDYWGQGTLVTVSS Cε1 (SEQ ID NO:3) Cε2 (SEQ ID NO:4) 25 Cε3 (SEQ ID NO:5) and Cε4 comprising mutation A26W (SEQ ID NO:6). Lambda light chain of anti-Her2 arm (EPS226) The following amino acid sequence (SEQ ID NO:11) describes the lambda light chain with anti-Her2 (EPS226, also referred to herein as V26) variable domain. 30 QSVLTQPASVSGSPGQSITISCTGTSSDVGSYNLVSWYQQHPGKAPKLMIYEVSNRPSGVSNRFSGSKSGNTASL TISGLQAEDEADYYCSSYTSSSTLVFGGGTKLTVLGQPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAVTV AWKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWKSHRSYSCQVTHEGSTVEKTVAPTECS IgE heavy chain of anti-FRα arm (MOv18) with “hole” mutation 39

[0040] The following amino acid sequence (SEQ ID NO:12) describes an IgE heavy chain with anti- FRα (MOv18) variable domain and “hole” mutation in the Cε4 domain. The hole mutation is F387G (numbering from the first residue of Cε1) or F69G (numbering from the first residue of Cε4) and the G is highlighted in bold and underlining below. 5 SEQ ID NO:12 further comprises mutations at positions 44 and 135, which aid correct pairing of heavy chain SEQ ID NO:12 with light-chain SEQ ID NO:17. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of SEQ ID NO:12 to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted 10 in bold and underlining). SEQ ID NO:12 QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQCLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSSASTQSPSVFPLTRGCKNIPSNATSVTLGC LATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSV 15 CSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELT LSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLT WSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAF ATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFI CRAVHEAASPSQTVQRAVSVNPGK 20 SEQ ID NO:12 comprises the following domains: MOv18 heavy chain variable region, comprising mutation G44C (SEQ ID NO:13): QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQCLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSS 25 Cε1 comprising mutation C14G (SEQ ID NO:14): ASTQSPSVFPLTRGCKNIPSNATSVTLGCLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVS GAWAKQMFTCRVAHTPSSTDWVDNKTFS Cε2 (SEQ ID NO:4) Cε3 (SEQ ID NO:5) 30 and Cε4 comprising mutation F69G (SEQ ID NO:15). 40

[0041] GPRAAPEVYAFATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVT RAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK The original MOv18 heavy chain variable domain sequence is shown in SEQ ID NO:16: QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQGLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS 5 SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSS Kappa light of chain anti-FRα arm (MOv18) The following amino acid sequence (SEQ ID NO:17) describes a kappa light chain with anti- FRα (MOv18) variable domain. SEQ ID NO:17 further comprises mutations at positions 100 and 214, which aid correct pairing of light-chain SEQ ID NO:17 with heavy chain SEQ ID 10 NO:12. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the constant domain of SEQ ID NO:17 to the variable region (G100C and C214del, numbering from first residue of variable domain; mutated residues highlighted in bold and underlining). SEQ ID NO:17: 15 DIQMTQTTSSLSASLGDRVTISCRASQDINNFLNWYQQKPDGTVKLLIYYTSRLHSGVPSRFSGSGSGTDYSLTI INLEQEDIAIYFCQQSSTIPRTFGCGTKLEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKV DNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGE* (* the C at the C-terminus from the original sequence is deleted) The original MOv18 kappa light chain sequence is shown in SEQ ID NO:18: 20 DIQMTQTTSSLSASLGDRVTISCRASQDINNFLNWYQQKPDGTVKLLIYYTSRLHSGVPSRFSGSGSGTDYSLTI INLEQEDIAIYFCQQSSTIPRTFGGGTKLEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKV DNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC IgE heavy chain of anti-CD3 arm (Blinatumomab) with “hole” mutation 25 The following amino acid sequence (SEQ ID NO:19) describes an IgE heavy chain with anti- CD3 (blinatumomab) variable domain and “hole” mutation in the Cε4 domain (F387G, numbering from the first residue of Cε1, or F69G, numbering from the first residue of Cε4, and highlighted in bold and underlined). SEQ ID NO:19 further comprises mutations at positions 43 and 132, which aid correct pairing 30 of heavy chain SEQ ID NO:19 with light-chain SEQ ID NO:22. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of SEQ ID NO:19 to the variable region (G43C, numbering from first residue of 41

[0042] variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted in bold and underlining). SEQ ID NO:19: DIKLQSGAELARPGASVKMSCKTSGYTFTRYTMHWVKQRPGQCLEWIGYINPSRGYTNYNQKFKDKATLTTDKSS 5 STAYMQLSSLTSEDSAVYYCARYYDDHYCLDYWGQGTTLTVSSASTQSPSVFPLTRGCKNIPSNATSVTLGCLAT GYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVCSR DFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTLSQ KHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWSR ASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFATP 10 EWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFICRA VHEAASPSQTVQRAVSVNPGK SEQ ID NO:19 comprises the following domains: Blinatumomab anti-CD3 heavy chain variable region, comprising mutation G43C (SEQ ID NO:20): 15 DIKLQSGAELARPGASVKMSCKTSGYTFTRYTMHWVKQRPGQCLEWIGYINPSRGYTNYNQKFKDKATLTTDKSS STAYMQLSSLTSEDSAVYYCARYYDDHYCLDYWGQGTTLTVSS Cε1 comprising mutation C14G (SEQ ID NO:14): Cε2 (SEQ ID NO:4) Cε3 (SEQ ID NO:5) 20 and Cε4 comprising mutation F69G (SEQ ID NO:15). The original blinatumomab anti-CD3 heavy chain variable region sequence is shown in SEQ ID NO:21: DIKLQSGAELARPGASVKMSCKTSGYTFTRYTMHWVKQRPGQGLEWIGYINPSRGYTNYNQKFKDKATLTTDKSS STAYMQLSSLTSEDSAVYYCARYYDDHYCLDYWGQGTTLTVSS 25 Kappa light chain of anti-CD3 arm (Blinatumomab) The following amino acid sequence (SEQ ID NO:22) describes a kappa light chain with anti- CD3 (blinatumomab) variable domain. SEQ ID NO:22 further comprises mutations at positions 99 and 213, which aid correct pairing of light-chain SEQ ID NO:22 with heavy chain SEQ ID 42

[0043] NO:19. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the constant domain of SEQ ID NO:22 to the variable region (A99C and C213del, numbering from first residue of variable domain; mutated residues highlighted in bold and underlining). 5 SEQ ID NO:22 DIQLTQSPAIMSASPGEKVTMTCRASSSVSYMNWYQQKSGTSPKRWIYDTSKVASGVPYRFSGSGSGTSYSLTIS SMEAEDAATYYCQQWSSNPLTFGCGTKLELKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVD NALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGE* (* the C at the C-terminus from the original sequence is deleted) 10 The original blinatumomab anti-CD3 kappa light chain sequence is shown in SEQ ID NO:23: DIQLTQSPAIMSASPGEKVTMTCRASSSVSYMNWYQQKSGTSPKRWIYDTSKVASGVPYRFSGSGSGTSYSLTIS SMEAEDAATYYCQQWSSNPLTFGAGTKLELKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVD NALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC IgE heavy chain of anti-CD3 arm (Runimotamab) with “hole” mutation 15 The following amino acid sequence (SEQ ID NO:24) describes an IgE heavy chain with anti- CD3 (runimotamab) variable domain and “hole” mutation in the Cε4 domain (F387G, numbering from the first residue of Cε1, or F69G, numbering from the first residue of Cε4, and highlighted in bold and underlined). SEQ ID NO:24 further comprises mutations at positions 44 and 133, which aid correct pairing 20 of heavy chain SEQ ID NO:24 with light-chain SEQ ID NO:27. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of SEQ ID NO:24 to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted in bold and underlining). 25 SEQ ID NO:24: EVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYIHWVRQAPGQCLEWIGWIYPGDGNTKYNEKFKGRATLTADTS TSTAYLELSSLRSEDTAVYYCARDSYSNYYFDYWGQGTLVTVSSASTQSPSVFPLTRGCKNIPSNATSVTLGCLA TGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVCS RDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTLS 30 QKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWS RASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFAT 43

[0044] PEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFICR AVHEAASPSQTVQRAVSVNPGK SEQ ID NO:24 comprises the following domains: Runimotamab anti-CD3 heavy chain variable region, comprising mutation G44C (SEQ ID 5 NO:25): EVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYIHWVRQAPGQCLEWIGWIYPGDGNTKYNEKFKGRATLTADTS TSTAYLELSSLRSEDTAVYYCARDSYSNYYFDYWGQGTLVTVSS Cε1 comprising mutation C14G (SEQ ID NO:14): Cε2 (SEQ ID NO:4) 10 Cε3 (SEQ ID NO:5) and Cε4 comprising mutation F69G (SEQ ID NO:15). The original runimotamab anti-CD3 heavy chain variable region sequence is shown in SEQ ID NO:26: EVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYIHWVRQAPGQGLEWIGWIYPGDGNTKYNEKFKGRATLTADTS 15 TSTAYLELSSLRSEDTAVYYCARDSYSNYYFDYWGQGTLVTVSS Kappa light chain of anti-CD3 arm (runimotamab) The following amino acid sequence (SEQ ID NO:27) describes a kappa light chain with anti- CD3 (runimotamab) variable domain. SEQ ID NO:27 further comprises mutations at positions 105 and 219, which aid correct pairing of light-chain SEQ ID NO:27 with heavy chain SEQ20 ID NO:24. This is achieved by moving a cysteine residue involved in formation of an inter- chain disulphide bond from the constant domain of SEQ ID NO:27 to the variable region (Q105C and C219del, numbering from first residue of variable domain; mutated residues highlighted in bold and underlining). SEQ ID NO:27: 25 DIVMTQSPDSLAVSLGERATINCKSSQSLLNSRTRKNYLAWYQQKPGQPPKLLIYWASTRESGVPDRFSGSGSGT DFTLTISSLQAEDVAVYYCTQSFILRTFGCGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAK VQWKVDNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGE* (* the C at the C-terminus from the original sequence is deleted) 44

[0045] The original runimotamab anti-CD3 kappa light chain sequence is shown in SEQ ID NO:28: DIVMTQSPDSLAVSLGERATINCKSSQSLLNSRTRKNYLAWYQQKPGQPPKLLIYWASTRESGVPDRFSGSGSGT DFTLTISSLQAEDVAVYYCTQSFILRTFGQGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAK VQWKVDNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC 5 IgE heavy chain of anti-CD3 arm (Tarlatamab) with “hole” mutation The following amino acid sequence (SEQ ID NO:29) describes an IgE heavy chain with anti- CD3 (tarlatamab) variable domain and “hole” mutation in the Cε4 domain (F387G, numbering from the first residue of Cε1, or F69G, numbering from the first residue of Cε4, and highlighted in bold and underlined). 10 SEQ ID NO:29 further comprises mutations at positions 44 and 139, which aid correct pairing of heavy chain SEQ ID NO:29 with light-chain SEQ ID NO:32. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of SEQ ID NO:29 to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted 15 in bold and underlining). SEQ ID NO:29: EVQLVESGGGLVQPGGSLKLSCAASGFTFNKYAMNWVRQAPGKCLEWVARIRSKYNNYATYYADSVKDRFTISRD DSKNTAYLQMNNLKTEDTAVYYCVRHGNFGNSYISYWAYWGQGTLVTVSSASTQSPSVFPLTRGCKNIPSNATSV TLGCLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNK 20 TFSVCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQ SELTLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGT VNLTWSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPE VYAFATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQK DEFICRAVHEAASPSQTVQRAVSVNPGK 25 SEQ ID NO:29 comprises the following domains: Tarlatamab anti-CD3 heavy chain variable region, comprising mutation G44C (SEQ ID NO:30): EVQLVESGGGLVQPGGSLKLSCAASGFTFNKYAMNWVRQAPGKCLEWVARIRSKYNNYATYYADSVKDRFTISRD DSKNTAYLQMNNLKTEDTAVYYCVRHGNFGNSYISYWAYWGQGTLVTVSS 30 Cε1 comprising mutation C14G (SEQ ID NO:14): 45

[0046] Cε2 (SEQ ID NO:4) Cε3 (SEQ ID NO:5) and Cε4 comprising mutation F69G (SEQ ID NO:15). The original tarlatamab anti-CD3 heavy chain variable region sequence is shown in SEQ ID 5 NO:31: EVQLVESGGGLVQPGGSLKLSCAASGFTFNKYAMNWVRQAPGKGLEWVARIRSKYNNYATYYADSVKDRFTISRD DSKNTAYLQMNNLKTEDTAVYYCVRHGNFGNSYISYWAYWGQGTLVTVSSLambda light chain of anti-CD3 arm (tarlatamab) The following amino acid sequence (SEQ ID NO:32) describes a lambda light chain with anti- 10 CD3 (tarlatamab) variable domain. SEQ ID NO:32 further comprises mutations at positions 102 and 214, which aid correct pairing of light-chain SEQ ID NO:32 with heavy chain SEQ ID NO:29. This is achieved by moving a cysteine residue involved in formation of an inter- chain disulphide bond from the constant domain of SEQ ID NO:32 to the variable region (G102C and C214A, numbering from first residue of variable domain; mutated residues 15 highlighted in bold and underlining). SEQ ID NO:32: QTVVTQEPSLTVSPGGTVTLTCGSSTGAVTSGNYPNWVQQKPGQAPRGLIGGTKFLAPGTPARFSGSLLGGKAAL TLSGVQPEDEAEYYCVLWYSNRWVFGCGTKLTVLGQPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAVTVA WKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWKSHRSYSCQVTHEGSTVEKTVAPTEAS 20 The original tarlatamab anti-CD3 lambda light chain sequence is shown in SEQ ID NO:33: QTVVTQEPSLTVSPGGTVTLTCGSSTGAVTSGNYPNWVQQKPGQAPRGLIGGTKFLAPGTPARFSGSLLGGKAAL TLSGVQPEDEAEYYCVLWYSNRWVFGGGTKLTVLGQPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAVTVA WKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWKSHRSYSCQVTHEGSTVEKTVAPTECS 25 IgE heavy chain of anti-NIP arm with “hole” mutation The following amino acid sequence (SEQ ID NO:34) describes an IgE heavy chain with anti- NIP variable domain and “hole” mutation in the Cε4 domain (F387G, numbering from the first residue of Cε1, or F69G, numbering from the first residue of Cε4, and highlighted in bold and underlined). NIP refers to 4-hydroxy-3-iodo-5-nitrophenylacetic acid, i.e. a small molecule 46

[0047] substrate not found in humans useful as an isotype control (see e.g. Bax et al. Allergy. 2020;00:1–5). SEQ ID NO:34 further comprises mutations at positions 44 and 134, which aid correct pairing of heavy chain SEQ ID NO:34 with light-chain SEQ ID NO:39. This is achieved by moving a 5 cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of SEQ ID NO:34 to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted in bold and underlining). SEQ ID NO:34: 10 QVQLQQPGAELVKPGASVKLSCKASGYTFTSYWMHWVKQRPGRCLEWIGRIDPNSGGTKYNEKFKSKATLTVDKP SSTAYMQLSSLTSEDSAVYYCARYDYYGSSYFDYWGQGTTLTVSSASTQSPSVFPLTRGCKNIPSNATSVTLGCL ATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVC SRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTL SQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTW 15 SRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFA TPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFIC RAVHEAASPSQTVQRAVSVNPGK SEQ ID NO:34 comprises the following domains: Anti-NIP heavy chain variable region, comprising mutation G44C (SEQ ID NO:35): 20 QVQLQQPGAELVKPGASVKLSCKASGYTFTSYWMHWVKQRPGRCLEWIGRIDPNSGGTKYNEKFKSKATLTVDKP SSTAYMQLSSLTSEDSAVYYCARYDYYGSSYFDYWGQGTTLTVSS Cε1 comprising mutation C14G (SEQ ID NO:14): Cε2 (SEQ ID NO:4) Cε3 (SEQ ID NO:5) 25 and Cε4 comprising mutation F69G (SEQ ID NO:15). IgE heavy chain of anti-NIP arm with “knob” mutation The following amino acid sequence (SEQ ID NO:36) describes an IgE heavy chain with anti- NIP variable domain and “knob” mutation in the Cε4 domain. The knob mutation is A344W 47

[0048] (numbering from the first residue of Cε1) or A26W (numbering from the first residue of Cε4) and is highlighted in bold and underlining. SEQ ID NO:36: QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYAMHWVRQAPGQRLEWIGWINAGNGNTKYSQKFQGRVTITRDTS 5 ASTAYMELSSLRSEDTAVYYCARDFSSQVATAAVDYWGQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLG CLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFS VCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSEL TLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNL TWSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYA 10 FATPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVTRAEWEQKDEF ICRAVHEAASPSQTVQRAVSVNPGK SEQ ID NO:36 comprises the following domains: Anti-NIP heavy chain variable region (SEQ ID NO:37): 15 QVQLQQPGAELVKPGASVKLSCKASGYTFTSYWMHWVKQRPGRGLEWIGRIDPNSGGTKYNEKFKSKATLTVDKP SSTAYMQLSSLTSEDSAVYYCARYDYYGSSYFDYWGQGTTLTVSS Cε1 (SEQ ID NO:3) Cε2 (SEQ ID NO:4) Cε3 (SEQ ID NO:5) 20 and Cε4 comprising mutation A26W (SEQ ID NO:6). Lambda light chain of anti-NIP The following amino acid sequence (SEQ ID NO:38) describes the lambda light chain with anti-NIP variable domain, which pairs with the heavy chain of SEQ ID NO:36. SEQ ID NO:38 25 QAVVTQESALTTSPGETVTLTCRSSTGAVTTSNYANWVQEKPDHLFTGLIGGTNNRAPGVPARFSGSLIGDKAAL TITGAQTEDEAIYFCALWYSNHWVFGGGTKLTVLGQPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAVTVA WKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWKSHRSYSCQVTHEGSTVEKTVAPTECS Alternative lambda light chain of anti-NIP The following amino acid sequence (SEQ ID NO:39) describes a lambda light chain with anti- 30 NIP variable domain. SEQ ID NO:39 further comprises mutations at positions 102 and 214, 48

[0049] which aid correct pairing of light-chain SEQ ID NO:39 with heavy chain SEQ ID NO:34. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the constant domain of SEQ ID NO:39 to the variable region (G102C and C214A, numbering from first residue of variable domain; mutated residues highlighted in bold and 5 underlining). SEQ ID NO:39: QAVVTQESALTTSPGETVTLTCRSSTGAVTTSNYANWVQEKPDHLFTGLIGGTNNRAPGVPARFSGSLIGDKAAL TITGAQTEDEAIYFCALWYSNHWVFGCGTKLTVLGQPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAVTVA WKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWKSHRSYSCQVTHEGSTVEKTVAPTEAS 10 With the sequences above, following bispecific IgE constructs combinations were made: 1. Trastuzumab x MOv18 (i.e. anti-Her2 heavy chain SEQ ID NO:1; anti-Her2 kappa light chain SEQ ID NO:8; anti-FRα heavy chain SEQ ID NO:12; and anti-FRα kappa light chain SEQ ID NO:17); 2. EPS226 (V26) x MOv18 (i.e. anti-Her2 heavy chain SEQ ID NO:9; anti-Her2 lambda 15 light chain SEQ ID NO:11; anti-FRα heavy chain SEQ ID NO:12; and anti-FRα kappa light chain SEQ ID NO:17); 3. Trastuzumab x Blinatumomab (i.e. anti-Her2 heavy chain SEQ ID NO:1; anti-Her2 kappa light chain SEQ ID NO:8; anti-CD3 heavy chain SEQ ID NO:19; and anti-CD3 kappa light chain SEQ ID NO:22); 20 4. EPS226 (V26) x Blinatumomab (i.e. anti-Her2 heavy chain SEQ ID NO:9; anti-Her2 lambda light chain SEQ ID NO:11; anti-CD3 heavy chain SEQ ID NO:19; and anti- CD3 kappa light chain SEQ ID NO:22). 5. Trastuzumab x Runimotamab (i.e. anti-Her2 heavy chain SEQ ID NO:1; anti-Her2 kappa light chain SEQ ID NO:8; anti-CD3 heavy chain SEQ ID NO:24; and anti-CD3 25 kappa light chain SEQ ID NO:27); 6. EPS226 (V26) x Runimotamab (i.e. anti-Her2 heavy chain SEQ ID NO:9; anti-Her2 lambda light chain SEQ ID NO:11; anti-CD3 heavy chain SEQ ID NO:24; and anti- CD3 kappa light chain SEQ ID NO:27); 49

[0050] 7. Trastuzumab x Tarlatamab (i.e. anti-Her2 heavy chain SEQ ID NO:1; anti-Her2 kappa light chain SEQ ID NO:8; anti-CD3 heavy chain SEQ ID NO:29; and anti-CD3 lambda light chain SEQ ID NO:32); 8. EPS226 (V26) x Tarlatamab (i.e. anti-Her2 heavy chain SEQ ID NO:9; anti-Her2 5 lambda light chain SEQ ID NO:11; anti-CD3 heavy chain SEQ ID NO:29; and anti- CD3 lambda light chain SEQ ID NO:32); 9. Trastuzumab x NIP (i.e. anti-Her2 heavy chain SEQ ID NO:1; anti-Her2 kappa light chain SEQ ID NO:8; anti-NIP heavy chain SEQ ID NO:34; anti-NIP lambda light chain SEQ ID NO:39); 10 10. EPS226 (V26) x NIP (i.e. anti-Her2 heavy chain SEQ ID NO:9; anti-Her2 lambda light chain SEQ ID NO:11; anti-NIP heavy chain SEQ ID NO:34; anti-NIP lambda light chain SEQ ID NO:39); 11. NIP x Runimotamab, (i.e. anti-NIP heavy chain SEQ ID NO:36; anti-NIP lambda light chain SEQ ID NO:38; anti-CD3 heavy chain SEQ ID NO:24; and anti-CD3 kappa light 15 chain SEQ ID NO:27); 12. NIP x Tarlatamab, (i.e. anti-NIP heavy chain SEQ ID NO:36; anti-NIP lambda light chain SEQ ID NO:38; anti-CD3 heavy chain SEQ ID NO:29; and anti-CD3 lambda light chain SEQ ID NO:32); Bispecific IgE expression in expi293 cells and purification 20 Expi293F resuspension cells (Gibco) were grown in the expi293 media (Gibco) to a cell density of 3 x 106 / mL before transfection. The transfection reaction was prepared by mixing 1 μg of DNA construct per million of cells, OptiMEM media (Gibco), and expiFectamine transfection kit (Gibco) as recommended by the manufacturer. The transfection mixture was mixed thoroughly by swirling, incubated for 10 min at room temperature and poured directly into the 25 flask.18 hours after the transfection, the expiFectamine enhancer mixtures were added to the culture, as recommended by the manufacturer. The culture was incubated at 125 rpm and 37 °C. The cells were harvested 5 days after transfection by centrifuging at 4,000 g for 20 mins at 4 °C. The pellet was discarded while the supernatant was filtered through 0.22 μm membrane and kept at 4 °C before the purification steps. 50

[0051] The filtered supernatant was applied at contact time of 10 min to a 5 mL pre-packed column containing CaptureSelect IgE resin (ThermoFisher) pre-equilibrated in PBS buffer pH 7.4. The column was washed with 5 column volumes (CVs) of PBS pH 7.4 before eluting the bound bispecific IgE with 5 CVs of 0.1 M sodium citrate pH 3.1. 5 The pooled eluate from CaptureSelect IgE step was concentrated to ~2 mL and applied to a Superdex 20016 / 600 size-exclusion column (Cytiva) pre-equilibrated in PBS pH 7.4. The peak fractions were pooled and kept at 4 °C. Bio-Layer Interferometry (BLI) of bispecific antibodies Bio-Layer Interferometry (BLI) experiments for anti-Her2 x anti-CD3 IgE antibodies were 10 performed in the Octet R8 system (Sartorius). The 96-well plates were shaking at 1000 rpm at 30 °C throughout the experiments. The buffer used for dilution, sensor washing, baseline, association, and dissociation steps was PBS pH 7.4, with 0.05% (v / v) Tween20. Streptavidin SA biosensor (Sartorius) was saturated with 0.1 μg / mL of biotinylated CaptureSelect™ biotin anti-IgE conjugate (ThermoFisher) for 600 s and washed for 120 s in 15 buffer. After 60 s immersion in the wells containing buffer to get the baseline recording, the loaded sensors were immersed for 600 s into the wells containing the purified bispecific IgE antibodies at concentration of 200 nM, followed by dissociation step into the buffer well for 300 s. After 60 s of second baseline recording, the loaded sensors were immersed into the wells for 180 s containing first antigen (Her2 or CD3, from Acro Biosystems) at concentration of 20 200 nM. Subsequently, the second association step into the wells for 180 s containing alternate second antigen (CD3 or Her2, from Acro Biosystems) at concentration of 200 nM. For the final dissociation step, the biosensors were immersed for 30 s into the same wells used for baseline recording. An outline of the experimental set up is shown in Figure 1. Figure 2 illustrates how raw 25 experimental data is normalised to baseline and analysed. Figures 3 and 4 show that the bispecific IgE antibodies trastuzumab x runimotamab and trastuzumab x tarlatamab can bind to both antigens (Her2 and CD3) simultaneously in the OCTET biosensors. The order of antigen additions did not alter the binding profile. 51

[0052] Figure 5 shows that the IgE bispecific antibodies V26 x Runimotamab and V26 x Tarlatamab can bind to both antigens (Her2 and CD3) simultaneously in the OCTET biosensors. In Fig.5, CD3 is added before Her2. The results demonstrate that bispecific antibodies as described herein are capable of 5 simultaneous binding to two different antigens, e.g. (i) Her2 and FRα or (ii) Her2 and CD3. In addition, it was demonstrated that the bispecific IgE antibodies trastuzumab x runimotamab and trastuzumab x tarlatamab are capable of binding to FcεR1 (see Figure 6). Thus the knob- into-hole mutations (i.e. A26W and F69G in Cε4) present in these antibodies do not impair interaction with FcεR1. 10 Sandwich ELISA for bispecific IgE An enzyme-linked immunosorbent assay (ELISA) for bispecific IgEs was developed. The assay was tested with anti-Her2 x anti-FRα (MOv18) and anti-Her2 x anti-CD3 (runimotamab) bispecific IgEs. The method is described in Figures 7 and 8. 100 μL sample of His-tagged CD3 or His-tagged 15 FRα (Acro Biosystems) at 5 μg / mL was coated for each well for 18 h. In the following day, each well was blocked with 150 μL of SuperBlock buffer (ThermoScientific) for 1 h. After 3x wash with 1x PBS, a 100 μL sample of purified bispecific antibodies (and monospecific controls) was added to the well and left incubating for 1 h. Subsequently, the well was washed thrice with 1x PBS + 0.05% (v / v) Tween20, before adding 100 μL of biotinylated Her2 (Acro 20 Biosystems) at 5 μg / mL in each well. After another 3x wash with 1x PBS + 0.05% (v / v) Tween20, 100 μL of HRP-conjugated streptavidin at 0.05 μg / mL was added to each well for 1 h. To develop the ELISA, each well was stringently washed with 3x wash with 1x PBS + 0.05% (v / v) Tween 20, followed by 3x wash with 1x PBS, before adding 100 μL to each well a 3, 3’, 5, 5’-tetramethylbenzidine solution for 5 mins before stopping the reaction with ELISA 25 StopSolution (Invitrogen). The plate was read in 450 nm wavelength using VarioSkan plate reader (ThermoScientific). All steps were done at room temperature. The results are shown in Figures 9 and 10. Mean is shown of two technical replicates, error bars are SD. Trastuzumab x MOv18 IgE, V26 x MOv18 IgE (see Fig.9) and runimotamab (anti-Her2 x anti-CD3, see Fig.10) IgE bispecific antibodies were tested. The results show 30 that bispecific IgE antibodies bind to both targets (FRɑ and Her2, or Her2 and CD3) 52

[0053] simultaneously by ELISA. In contrast, monospecific IgE antibodies containing knobs-into- holes (KiH) mutations (i.e. Trastuzumab KiH mono IgE, Mov18 KiH mono IgE, V26 KiH mono IgE, anti-CD3 KiH mono IgE or anti-Her2 KiH mono IgE) did not bind to both antigens in this ELISA. 5 Cell-based assays Flow sandwich cell binding assay CD3+ expressing Jurkat cells were incubated with the Her2+ expressing SKBR3 cell line at an effector to target (E:T) ratio of 1:1 for 1 hour at 37°C, alone with the monospecific control and bispecific antibodies. Monospecific Trastuzumab KiH IgE (Tras KiH mono), monospecific 10 blinatumomab KiH IgE (Blina KiH mono), and Runimotumab (anti-CD3 and anti-Her2 bispecific IgG antibody) were added to cells at a final concentration of 100 nM. For the Tras IgE + Blina IgE condition (Blina + Tras mono), 100 nM of each antibody was added to cells. Bispecific Blina x Tras IgE (Blina x Tras) were added to cells at a final concentration of 200 nM. Followed the 1-hour incubation, cells were washed twice with PBS and harvested by 15 trypsin. Each sample was stained with live / dead dye and CD45, a common marker expressing on Jurkat cells. Cells were then fixed with 2% paraformaldehyde and analysed by flow cytometry (BD FACSymphony A1). As shown in Figure 11, staining indicates binding of Jurkat cells to SKBR3 cells, e.g. via crosslinking of CD3 / HER2 by bispecific antibody. The percentage of CD45+ Jurkat cells were 20 plotted to compare the potential binding effect of the antibodies. The results are shown in Figure 12. Mean is shown of three technical replicates, error bars are SEM. Figure 12 indicates that Blina x Tras (i.e. a bispecific IgE antibody that binds Her2 and CD3) induces high levels of crosslinking of Jurkat and SKBR3 cells (compared to monospecific antibodies alone or in combination and runimotamab IgG). 25 Activation of T-cells and cytotoxicity induced by bispecific antibodies T cell activation and target cell cytotoxicity were then measured in a similar assay. Her2+ SKBR3 cells were labelled with CFSE (1 µM) and cultured with CD8+ T cells isolated from healthy human blood at an effector to target (E:T) ratio of 2:1 for 48 hours in the presence of antibodies stated. Monospecific Trastuzumab KiH IgE (Tras KiH mono) and monospecific 53

[0054] blinatumomab KiH IgE (Blina KiH mono) were added to cells at a final concentration of 100 nM. Bispecific Blina x Tras IgE (Blina x Tras) was added to cells at a final concentration of 200 nM. After 48 hours incubation, SKBR3 and CD8+ T cells were harvested, stained for CD45 and a live / dead dye, fixed with 2% paraformaldehyde and analysed by flow cytometry 5 (BD FACSymphony A1) with CountBright absolute counting beads. SKBR3 target cells (CFSE+ CD45-) were normalised to absolute counting beads and then made relative to the no antibody condition to calculate relative SKBR3 survival. The results are shown in Figures 13 and 14. Mean is shown of two technical replicates, error bars are SEM. 10 In a further assay, Her2+ SKBR3 cells were labelled with CFSE (1 µM) and cultured with CD8+ T cells isolated from healthy human blood at an E:T ratio of 2:1 for 48 hours in the presence of antibodies stated. Monospecific Trastuzumab KiH IgE (Tras IgE) and monospecific blinatumomab KiH IgE (Blina IgE) were added to cells at a final concentration of 100 nM. For the Tras IgE + Blina IgE condition (i.e. two monospecific antibodies), 100 nM 15 of each antibody was added to cells. Bispecific Blina x Tras IgE was added to cells at a final concentration of 200 nM. After 48 hours incubation, SKBR3 and CD8+ T cells were harvested, stained for CD45, CD69, and a live / dead dye, fixed with 2% paraformaldehyde and analysed by flow cytometry (BD FACSymphony A1). The percentage of CD8+ T cells (CD45+ CFSE- ) expressing CD69 was calculated. 20 The results are shown in Figure 15. Mean is shown of three technical replicates, error bars are SEM. The results in Figures 13 to 15 show that Blina x Tras (i.e. a bispecific IgE antibody that binds Her2 and CD3) reduces survival of SKBR3 cells and increases T cell activation, compared to monospecific antibodies alone or in combination. 25 In a further study, T cell activation in the presence and absence of target (SKBR3) cells is measured. Jurkat T cells were cultured in the presence or absence of Her2+ SKBR3 cells at an E:T of 1:1 for 48 hours in the presence of antibodies stated. Monospecific Trastuzumab KiH IgE (Tras IgE) and monospecific blinatumomab KiH IgE (Blina IgE) were added to cells at a final concentration of 100 nM. For the Tras IgE + Blina IgE condition, 100 nM of each 30 monospecific antibody was added to cells. Bispecific Blina x Tras IgE was added to cells at a final concentration of 200 nM. After 48 hours incubation, SKBR3 and Jurkat T cells were 54

[0055] harvested, stained for CD45, CD69, and a live / dead dye, fixed with 2% paraformaldehyde and analysed by flow cytometry (BD FACSymphony A1). The percentage of Jurkat T cells (CD45+) expressing CD69 was calculated. The results are shown in Figure 8. Mean is shown of three technical replicates, error bars are 5 SEM. The results show that Blina x Tras (i.e. a bispecific IgE antibody that binds Her2 and CD3) requires target cells for T cell activation, in contrast to monospecific antibodies alone or in combination. Identification of IgE-specific sites to generate IgE heavy chain heterodimers A set of protein structures containing IgE domains were selected (n = 15): 1) IgE in complex 10 with FcεRI (PDB IDs: 8C1C, 2Y7Q, 1F6A), 2) IgE in complex with FcεRII (PDB IDs: 5LGK, 4EZM, 4GKO, 4KI1), and 3) IgE alone (PDB IDs: 1O0V, 2WQR, 5MOL, 3H9Z, 3HA0, 5MOI, 5MOJ, 5MOK). The interacting residues between the two heavy chains of IgE within each structure were determined using PyMOL. Within each group (1-3), the interacting residues common to all structures were derived. As structures within each group contained 15 different numbers of IgE domains (i.e. Cε1-Cε2-Cε3-Cε4, Cε2-Cε3-Cε4, or Cε3-Cε4), structures contributed unequally at some sites. Interacting residues common across all groups were then determined. PyMOL (Schrödinger) was used to analyse properties of each amino acid identified and pairs or groups that would be amenable to mutation to produce KiH-based IgE bispecifics. Models of KiH mutations and image analyses were generated using PyMOL 20 and scripts run on PyMOL. Using this method, the pairs of mutations shown in Table 4 were identified as suitable for promoting heterodimerization of two IgE heavy chains: 55

[0056] Table 4 56

[0057] These positions are shown in the Cε2, Cε3 and Cε4 domain sequences of IgE below in bold and underlining: Cε2 (1-107) (SEQ ID NO:4): 5 VCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLS TASTTQEGELASTQSELTLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCA Cε3 (1-107) (SEQ ID NO:5): DSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWSRASGKPVNHSTR KEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTS 10 Cε4 (1-110) (SEQ ID NO:7): GPRAAPEVYAFATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTT QPRKTKGSGFFVFSRLEVTRAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK It is noted that in the mutation pairs shown in Table 4 above, at least one mutation in each pair is in Cε3 or Cε4. Preferably both mutations are in Cε4. 15 Although Table 4 shows preferred specific mutations, alternative residues may be introduced at the identified positions. For example, where phenylalanine, tryptophan, isoleucine, leucine or valine is indicated (e.g. as a “knob” mutation), any of that group of larger amino acid residues may alternatively be used at that position. Similarly where glycine, serine or threonine is indicated as a “hole”, any of that group of smaller amino acid residues may be used at that 20 position. Expression of asymmetric IgE to assess IgE heavy chain heterodimer formation The IgE-specific mutations in Table 4 were tested experimentally as listed, but also in further combinations as shown in Table 5. To assess heterodimer formation, asymmetric IgE were produced by expression of three 25 different chains: 1) the heavy chain of trastuzumab IgE (SEQ ID NO:178), 2) the light chain 57

[0058] of trastuzumab IgE (SEQ ID NO:179), and 3) a truncated heavy chain of IgE covering the Fc region (Cε2-Cε3-Cε4, SEQ ID NO:180). The extent of heterodimer formation (i.e. all three chains) was evaluated by the overall expression level of the heterodimer and by the relative expression level of the heterodimer compared to trastuzumab IgE homodimer (i.e. the heavy 5 chain and light chain of trastuzumab IgE) and Fc homodimer (i.e. the heavy chain of IgE Fc). From Tables 4 and 5, ‘Heavy chain 1’ mutations were incorporated into the heavy chain sequence for trastuzumab IgE and ‘Heavy chain 2’ mutations into the heavy chain sequence of IgE Fc. No mutations were made to the light chain of trastuzumab IgE. Expression of heterodimeric IgE was performed as described for bispecific IgE in Expi293F cells. Proteins 10 were expressed in 2.5 mL cultures in 24-well 10 mL plates and incubated at 250 rpm. The DNA plasmids separately encoding the three chains were transfected into cells in a 1:1:1 ratio (n = 2). Supernatants were harvested and heterodimer formation was visualized by SDS-PAGE (Novex 4-12 % Tris-Glycine, 1.0 mm, 15 WedgeWell) using InstantBlue Coomassie protein stain. Each mutational set favoured different levels of heterodimer formation, as shown in 15 Figures 17-20 and 25. Optimization and identification of further IgE-specific sites to generate IgE heavy chain heterodimers The sets of IgE-specific mutations that produced the highest levels and purity of IgE heavy chain heterodimers were (Heavy chain 1 / Heavy chain 2): 1) A10V, A26L, S54L, T56W / 20 A10V, F11Y, F67A, F69V; 2) T13W / Y9T; 3) A26Y / F69G, 4) R71Y / F67T; and 5) R71F / F67T. The pairs of KiH mutations in 2) to 5) were selected for further optimization and for identification of additional mutations that may improve heterodimerization. DDGun3D (Montanucci L, et al. (2019) BMC Bioinformatics.20 (Suppl 14): 335) was used to screen in silico the effect of mutations on the stability of IgE (PDB ID: 5MOL). First, every 25 combination of amino acids for each pair was screened (n = 400 per pair). Some combinations were prioritized for experimental testing. Selected combinations had a predicted increase in stability (T_DDG[3D], Stability[3D] cutoff ≥ 0.3) and combined a ‘larger’ and a ‘smaller’ amino acid. Second, groups of three amino acids were screened whereby: 1) ‘Heavy chain 1’ mutation (or ‘knob’) was at position T13, A26, or R71; 2) ‘Heavy chain 2’ mutation (or ‘hole’) 30 was, respectively, at position Y9, F69, or F67; and 3) another ‘Heavy chain 2’ mutation was added that was defined as an additional amino acid interacting with the position defined in 58

[0059] ‘Heavy chain 1’ (cutoff ≤ 5 Å). Every combination of amino acids for all three positions of a given set were screened simultaneously (n = 8000 per group). Some combinations with a predicted increase in stability as above were prioritized for testing. Finally, groups of four amino acids were screened that again included the four sets of prioritized KiH pairs, but also 5 had two mutations on ‘Heavy chain 1’ and two mutations on ‘Heavy chain 2’ that interact in a small group (n = 160000 per group). Table 6 shows additional mutations that were predicted to increase the stability of IgE from this analysis and that were selected for expression in Table 7. As before, further combinations were also tested in Table 7. The further sets of mutations were expressed and assessed as described for the first round as 10 shown in Figures 21 to 30. A subset was also re-expressed in 25 mL cell cultures as shown in Figures 31 to 32. Supernatants were harvested and heterodimer formation was visualized by SDS-PAGE (Novex 4-20 % Tris-Glycine for Figures 21-30 and Novex 4-12 % in Figures 31 to 33). Several groups of mutations favoured heterodimerization of IgE heavy chains. Examples of mutations that showed improved IgE heterodimer formation after optimization 15 are shown in Figures 31 to 32. From Figures 31 and 32, the best performing groups of mutations after optimisation based on heterodimer formation and expression were: 1. Position T13 (Y or W) in combination with Y9F, L28V 2. Position A26 (Y and F) in combination with L28V, F69I 20 3. Position R71 (Y and F) in combination with L28V, F67I 4. Position F11W and T13W in combination with positions Y9F and F11I. SEQ ID NO: 210 - sequence of expressed trastuzumab IgE HC (wild type) EVQLVESGGGLVQPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNG 25 YTRYADSVKGRFTISADTSKNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDYW GQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLGCLATGYFPEPVMVTWDTGSL NGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSV CSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLST 59

[0060] ASTTQEGELASTQSELTLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVS AYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWSRASGKPVNHSTRKEEKQRNG TLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFATPE WPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRL 5 EVTRAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK SEQ ID NO: 211 - sequence of expressed trauzumab kappa LC (wild type) DIQMTQSPSSLSASVGDRVTITCRASQDVNTAVAWYQQKPGKAPKLLIYSASFLYSG VPSRFSGSRSGTDFTLTISSLQPEDFATYYCQQHYTTPPTFGQGTKVEIKRTVAAPSVF IFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVDNALQSGNSQESVTEQDSKDSTY 10 SLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC SEQ ID NO: 212 - sequence of expressed IgE Fc HC (wild type) AARDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLST ASTTQEGELASTQSELTLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVS AYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWSRASGKPVNHSTRKEEKQRNG 15 TLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFATPE WPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRL EVTRAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK 60

[0061] .sreimdoretehinahcyvaehEg I 4feonio 4eC,44 444444 44tnaiamromoCeCeCeCeCeCeCee eeC22,233, ,4, , , ,C,C,C, ,4fDeCeCeCeCeC3eC4eC4eCeC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eCeCrofnde iotstaWe ttu65yM4T lltanne ioitmaretuL4pxM35Sesnn iioo ttaatWtu1ui6L6K6F6F6nM2 2A2A2A2A2AM.ah5lCne ybviotaD DW2IIV WWYYYYFaTaetHuI71M11S21G2G01305R1S010T- 1K1Y A1F31W T31Y T31W T31Y T31W T316T26A266666C A2A2A2A2A2A31C T31T

[0062] G9G9LLDFLL6 69691665 99F F F6FT T6F6F26K1K1T7AATDTTAA676767603T97676767D DF F F FQ YF F F6F4e4eCe,C4C4e ,C4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC K37A7E6FY62W6Y A2A17F R17Y R17F R17K R42H T42Y4Y T5S65Y4Y T5S65T

[0063] esaerc r rn io,o,o t W de3t1W3cidT1,T,er 2Y31Y3pnT1sn iot,T,ioatYYYY tatu31uM T31T31T31T mfospuorFFFFFFFFrgor r r r,Io,Io,Io,Io,I I I I I I roroe31l T31313131313131 Ip rTTTTTTT626262626217R17R17R17R17R17, ,RW W max1iorn,or r r r r r,o,o,o,o,o,o,A,A,A,A,A, ror r r r r,o,o,o,o,o,111F1,FE.ah1W31W31W3W3W3W3W3W3W62W62W62W62W62F17F17F17F17F17F1,7Y Y 6lCneyiot T,T1,T1,T1,T1,T1 1,T,T,A,bv FA,A,A,A,R,R,R,R,R,R,111F1,F,aaatuY3Y3Y3Y3Y3Y3Y3Y3Y66Y6F6Y6F6Y6F6Y6F6Y1Y1Y1Y1Y1Y1 I1W1I1W1TeH M1T1T1T1T1T1T1T1T2A2A2A2A2A2A2A2A2A2A7R7R7R7R7R7R1F1F1F1F

[0064] 4e4Ce,C4e ,C4eC V1I1F11F L9IY9Y464e4Ce,eC4,C4eC r or,o,W31W3T1,TY,31Y3T1,TY,31Y T31Y T31Y T31T ro, ro,W11W1F1,FY,11Y1F1,IF1,1W1I1W1F1F1F1F

[0065] 4eC 56 nio ahtnaCiomyrvtaoafetuTLTI FI I FI FFI FI I FI FFI FI I FST roH M1 9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9IY9FY9FY96Y2A82L fdetset ylltane niaimremoD4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4e4e4e4e4e4e4e4e4e4e4epxC C C C C C C C C C C esnn iioo ttaattuu1inM2M.ah7lCneybviota FFFFFFFFFFFaaYYYYYYYYWWWWWFFTetHuM131T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31T31W T31W T31W T316T26A2A

[0066] 4eC 66 L9V9I9I1VG9STL9V9I IVTLVI IVLVI IVTAAYI WVAI Y6661 82662826696118282969696182999182777777820377F F F FLFA LF F F FL LF F F1FL6F6F6F1FL6F6F6F6F6F6FL Q6F6F4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC F 6F26F26F26F26I26I26I I I I I I2626262626262W62W62W62W62W62W62Y62Y6Y6Y6Y6I1I1I1I1I1I1I1I1I1 Y1AA A A A A A A A A A A A A A A A A A A2A2A2A2A7R7R7R7R7R7R7R7R7R7R

[0067] 76I7I7I9I7I7I9VI WI VI WI VI WI VI WI666 6661111111111111111111 1 1 1 1 1 1F F F F F F F F F F F F F F F1F1F1F1F1F1F1FI7W67VA F6F820I7Y7I7W7VA0I7LI I FLI I FLI I FLIL3Q6F6F6F6F82L3Q6F9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9Y9IY9FY9Y 4e4e4e4444444444444eC C Ceeeeeeeeee eee, , ,C C C C C C C C C C C C C4C4eC4e , , , , , , , , , , , , ,C4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC4eC Y31Y T31Y T31Y T31Y T31Y T31Y T31Y T31W T31W T31W T31W T31Y T31Y T31Y T31Y T31T Y17Y1YYYFFFFFFI1I1I1I1Y1Y1Y1Y1W1W1W1W1W1W1W1W1R7R17R17R17R17R17R17R17R17R17R1F1F1F1F1F1F1F1F1F1F1F1F1F1F1F1F

[0068] Example 2 - Design and cloning of bispecific IgEG constructs In this example, human IgEG bispecific antibodies were constructed. Each DNA encoding the antibody chain was synthesised and subsequently cloned into an expression vector pTWIST CMV Beta Globin (Twist Inc.) together with secretion signal peptide. 5 For the heavy chain sequence, the secretion signal is as follows: MEWSWVFLFFLSVTTGVHS (SEQ ID NO:178) For the light chain sequence, the secretion signal is as follows: MSVPTQVLGLLLLWLTDARC (SEQ ID NO:179) IgEG heavy chain Anti-Her2 arm (Trastuzumab) with “knob” mutation in IgE Fc portion 10 and “knob” mutation in IgG Fc portion [IgEG cis] The following amino acid sequence (SEQ ID NO:180) describes an IgEG heavy chain with anti-Her2 (Trastuzumab) variable domain, “knob” mutation in the Cε4 domain, and “knob” mutation in the CH3 domain. The knob mutations are A26W (numbering from the first residue of Cε4) and T366Y (numbering based on Ridgway et al, 1996) and are highlighted in bold and 15 underlining. SEQ ID NO:180: EVQLVESGGGLVQPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNGYTRYADSVKGRFTISADTS KNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDYWGQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLGCL ATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVC 20 SRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTL SQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTW SRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFA TPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVTRAEWEQKDEFIC RAVHEAASPSQTVQRAVSVNPGKRSEPKSSDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVVD 25 VSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKAK GQPREPQVYTLPPSREEMTKNQVSLYCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLYSKLTVDKS RWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK SEQ ID NO:180 comprises the following domains: 30 68

[0069] Trastuzumab heavy chain variable region (SEQ ID NO:181): EVQLVESGGGLVQPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNGYTRYADSVKGRFTISADTS KNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDYWGQGTLVTVSS 5 Cε1 (SEQ ID NO:182): ASTQSPSVFPLTRCCKNIPSNATSVTLGCLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVS GAWAKQMFTCRVAHTPSSTDWVDNKTFS Cε2 (SEQ ID NO:183): VCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSEL 10 TLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCA Cε3 (SEQ ID NO:184): DSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWSRASGKPVNHSTRKEEKQRNGTLTVTSTLP VGTRDWIEGETYQCRVTHPHLPRALMRSTTKTS Cε4 comprising mutation A26W (SEQ ID NO:185): 15 GPRAAPEVYAFATPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVT RAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK A linker (SEQ ID NO:186): RSEPKSS 20 CH2 (SEQ ID NO:187): DKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVVDVSHEDPEVKFNWYVDGVEVHNAKTKPREEQ YNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKAK and CH3 comprising mutation T366Y (SEQ ID NO:188): 25 GQPREPQVYTLPPSREEMTKNQVSLYCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLYSKLTVDKS RWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK The original (wild type) Cε4 sequence is shown in SEQ ID NO:189: GPRAAPEVYAFATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVT 30 RAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK 69

[0070] The original (wild type) CH3 sequence is shown in SEQ ID NO:190: GQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLYSKLTVDKS RWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK 5 Kappa light chain of anti-Her2 arm (Trastuzumab) The following amino acid sequence (SEQ ID NO:191) describes the kappa light chain with anti-Her2 (Trastuzumab) variable domain. SEQ ID NO:191: DIQMTQSPSSLSASVGDRVTITCRASQDVNTAVAWYQQKPGKAPKLLIYSASFLYSGVPSRFSGSRSGTDFTLTI 10 SSLQPEDFATYYCQQHYTTPPTFGQGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKV DNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC IgEG heavy chain anti-FRα arm (MOv18) with “hole” mutation in IgE Fc portion and “hole” mutation in IgG Fc portion [IgEG cis] 15 The following amino acid sequence (SEQ ID NO:192) describes an IgEG heavy chain with anti-FRα (MOv18) variable domain, “hole” mutation in the Cε4 domain, and “hole” mutation in the CH3 domain. The hole mutations are F69G (numbering from the first residue of Cε4) and Y407T (numbering based on Ridgway et al, 1996) and are highlighted in bold and underlining. 20 This sequence further comprises mutations at positions 44 and 135, which aid correct pairing of heavy chain SEQ ID NO:192 with light chain SEQ ID NO:197. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted in bold 25 and underlining). SEQ ID NO:192: QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQGLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLGC LATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSV 30 CSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELT LSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLT 70

[0071] WSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAF ATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFI CRAVHEAASPSQTVQRAVSVNPGKRSEPKSSDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVV DVSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKA 5 KGQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLTSKLTVDK SRWQQGNVFSCSVMHEALHNHYTQKSLSLSPG SEQ ID NO:192 comprises the following domains: MOv18 heavy chain variable region, comprising mutation G44C (SEQ ID NO:193): QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQCLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS 10 SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSS Cε1 comprising mutation C14G (SEQ ID NO:194): ASTQSPSVFPLTRGCKNIPSNATSVTLGCLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVS GAWAKQMFTCRVAHTPSSTDWVDNKTFS Cε2 (SEQ ID NO:183) 15 Cε3 (SEQ ID NO:184) Cε4 comprising mutation F69G (SEQ ID NO:195). GPRAAPEVYAFATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVT RAEWEQKDEFICRAVHEAASPSQTVQRAVSVNPGK Linker (SEQ ID NO:186) 20 CH2 (SEQ ID NO:187) and CH3 comprising mutation Y407T (SEQ ID NO:196) GQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLTSKLTVDKS RWQQGNVFSCSVMHEALHNHYTQKSLSLSPG The original MOv18 heavy chain variable domain sequence is shown in SEQ ID NO:197:25 QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQGLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSS Kappa light of chain anti-FRα arm (MOv18) 71

[0072] The following amino acid sequence (SEQ ID NO:198) describes a kappa light chain with anti- FRα (MOv18) variable domain. It further comprises mutations at positions 100 and 214, which aid correct pairing of light chain SEQ ID NO:198 with heavy chain SEQ ID NO:192. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond 5 from the constant domain to the variable region (G100C and C214del, numbering from first residue of variable domain; mutated residues highlighted in bold and underlining). SEQ ID NO:198: DIQMTQTTSSLSASLGDRVTISCRASQDINNFLNWYQQKPDGTVKLLIYYTSRLHSGVPSRFSGSGSGTDYSLTI INLEQEDIAIYFCQQSSTIPRTFGCGTKLEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKV 10 DNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGE* (* theCat the C-terminus from the original sequence is deleted) The original MOv18 kappa light chain sequence is shown in SEQ ID NO:199: DIQMTQTTSSLSASLGDRVTISCRASQDINNFLNWYQQKPDGTVKLLIYYTSRLHSGVPSRFSGSGSGTDYSLTI INLEQEDIAIYFCQQSSTIPRTFGGGTKLEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKV 15 DNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC IgEG heavy chain anti-CD3 arm (Runimotamab) with “hole” mutation in IgE Fc portion and “hole” mutation in IgG Fc portion [IgEG cis] The following amino acid sequence (SEQ ID NO:200) describes an IgEG heavy chain with 20 anti-CD3 (Runimotamab) variable domain, “hole” mutation in the Cε4 domain, and “hole” mutation in the CH3 domain. The hole mutations are F69G (numbering from the first residue of Cε4) and Y407T (numbering based on Ridgway et al, 1996) and are highlighted in bold and underlining. This sequence further comprises mutations at positions 44 and 135, which aid correct pairing 25 of heavy chain SEQ ID NO:200 with light chain SEQ ID NO:203. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted in bold and underlining). 30 SEQ ID NO:200: 72

[0073] EVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYIHWVRQAPGQCLEWIGWIYPGDGNTKYNEKFKGRATLTADTS TSTAYLELSSLRSEDTAVYYCARDSYSNYYFDYWGQGTLVTVSSASTQSPSVFPLTRGCKNIPSNATSVTLGCLA TGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVCS RDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTLS 5 QKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTWS RASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFAT PEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFICR AVHEAASPSQTVQRAVSVNPGKRSEPKSSDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVVDV SHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKAKG 10 QPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLTSKLTVDKSR WQQGNVFSCSVMHEALHNHYTQKSLSLSPGK SEQ ID NO:200 comprises the following domains: Runimotamab anti-CD3 heavy chain variable region, comprising mutation G44C (SEQ ID NO:201): 15 EVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYIHWVRQAPGQCLEWIGWIYPGDGNTKYNEKFKGRATLTADTS TSTAYLELSSLRSEDTAVYYCARDSYSNYYFDYWGQGTLVTVSS Cε1 comprising mutation C14G (SEQ ID NO:194): Cε2 (SEQ ID NO:183) Cε3 (SEQ ID NO:184) 20 Cε4 comprising mutation F69G (SEQ ID NO:195). Linker (SEQ ID NO:186) CH2 (SEQ ID NO:187) and CH3 comprising mutation Y407T (SEQ ID NO:196) The original Runimotamab anti-CD3 heavy chain variable region is shown in SEQ ID NO:202:25 EVQLVQSGAEVKKPGASVKVSCKASGYTFTNYYIHWVRQAPGQGLEWIGWIYPGDGNTKYNEKFKGRATLTADTS TSTAYLELSSLRSEDTAVYYCARDSYSNYYFDYWGQGTLVTVSS Kappa light chain of anti-CD3 arm (Runimotamab) 73

[0074] The following amino acid sequence (SEQ ID NO:203) describes a kappa light chain with anti- CD3 (Runimotamab) variable domain. It further comprises mutations at positions 105 and 219, which aid correct pairing of light chain SEQ ID NO:203 with heavy chain SEQ ID NO:200. This is achieved by moving a cysteine residue involved in formation of an inter-chain 5 disulphide bond from the constant domain of to the variable region (Q105C and C219del, numbering from first residue of variable domain; mutated residues highlighted in bold and underlining). SEQ ID NO:203: DIVMTQSPDSLAVSLGERATINCKSSQSLLNSRTRKNYLAWYQQKPGQPPKLLIYWASTRESGVPDRFSGSGSGT 10 DFTLTISSLQAEDVAVYYCTQSFILRTFGCGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAK VQWKVDNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGE* (* the C at the C-terminus from the original sequence is deleted) The original runimotamab anti-CD3 kappa light chain sequence is shown in SEQ ID NO:204: DIVMTQSPDSLAVSLGERATINCKSSQSLLNSRTRKNYLAWYQQKPGQPPKLLIYWASTRESGVPDRFSGSGSGT 15 DFTLTISSLQAEDVAVYYCTQSFILRTFGQGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAK VQWKVDNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC IgEG heavy chain Anti-Her2 arm (Trastuzumab) with “knob” mutation in IgE Fc portion and “hole” mutation in IgG Fc portion [IgEG trans] The following amino acid sequence (SEQ ID NO:205) describes an IgEG heavy chain with 20 anti-Her2 (Trastuzumab) variable domain, “knob” mutation in the Cε4 domain, and “hole” mutation in the CH3 domain. The knob mutation is A26W (numbering from the first residue of Cε4) and hole mutation is Y407T (numbering based on Ridgway et al, 1996) and are highlighted in bold and underlining. SEQ ID NO:205: 25 EVQLVESGGGLVQPGGSLRLSCAASGFNIKDTYIHWVRQAPGKGLEWVARIYPTNGYTRYADSVKGRFTISADTS KNTAYLQMNSLRAEDTAVYYCSRWGGDGFYAMDYWGQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLGCL ATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSVC SRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELTL SQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLTW 30 SRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAFA TPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVTRAEWEQKDEFIC 74

[0075] RAVHEAASPSQTVQRAVSVNPGKRSEPKSSDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVVD VSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKAK GQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLTSKLTVDKS RWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK 5 SEQ ID NO:205 comprises the following domains: Trastuzumab heavy chain variable region (SEQ ID NO:181) Cε1 (SEQ ID NO:182) Cε2 (SEQ ID NO:183) Cε3 (SEQ ID NO:184) 10 Cε4 comprising mutation A26W (SEQ ID NO:185) A linker (SEQ ID NO:186) CH2 (SEQ ID NO:187) and CH3 comprising mutation Y407T (SEQ ID NO:196) 15 IgEG heavy chain Anti-Her2 arm (EPS226) with “knob” mutation in IgE Fc portion and “hole” mutation in IgG Fc portion [IgEG trans] The following amino acid sequence (SEQ ID NO:206) describes an IgEG heavy chain with anti-Her2 (EPS226, also referred to herein as V26) variable domain, “knob” mutation in the Cε4 domain, and “hole” mutation in the CH3 domain. The knob mutation is A26W (numbering 20 from the first residue of Cε4) and hole mutation is Y407T (numbering based on Ridgway et al, 1996) and are highlighted in bold and underlining. SEQ ID NO:206: QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYAMHWVRQAPGQRLEWIGWINAGNGNTKYSQKFQGRVTITRDTS ASTAYMELSSLRSEDTAVYYCARDFSSQVATAAVDYWGQGTLVTVSSASTQSPSVFPLTRCCKNIPSNATSVTLG 25 CLATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFS VCSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSEL TLSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNL TWSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYA FATPEWPGSRDKRTLWCLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVFSRLEVTRAEWEQKDEF 75

[0076] ICRAVHEAASPSQTVQRAVSVNPGKRSEPKSSDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVV VDVSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISK AKGQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLTSKLTVD KSRWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK 5 SEQ ID NO:206 comprises the following domains: EPS226 heavy chain variable region (SEQ ID NO:207) QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYAMHWVRQAPGQRLEWIGWINAGNGNTKYSQKFQGRVTITRDTS ASTAYMELSSLRSEDTAVYYCARDFSSQVATAAVDYWGQGTLVTVSS Cε1 (SEQ ID NO:184) 10 Cε2 (SEQ ID NO:183) Cε3 (SEQ ID NO:184) Cε4 comprising mutation A26W (SEQ ID NO:185) A linker (SEQ ID NO:186) 15 CH2 (SEQ ID NO:187) and CH3 comprising mutation Y407T (SEQ ID NO:196) Lambda light chain of anti-Her2 arm (EPS226) The following amino acid sequence (SEQ ID NO:208) describes the lambda light chain with anti-Her2 (EPS226) variable domain. 20 SEQ ID NO:208: QSVLTQPASVSGSPGQSITISCTGTSSDVGSYNLVSWYQQHPGKAPKLMIYEVSNRPSGVSNRFSGSKSGNTASL TISGLQAEDEADYYCSSYTSSSTLVFGGGTKLTVLGQPKAAPSVTLFPPSSEELQANKATLVCLISDFYPGAVTV AWKADSSPVKAGVETTTPSKQSNNKYAASSYLSLTPEQWKSHRSYSCQVTHEGSTVEKTVAPTECS 25 IgEG heavy chain anti-FRα arm (MOv18) with “hole” mutation in IgE Fc portion and “knob” mutation in IgG Fc portion [IgEG trans] The following amino acid sequence (SEQ ID NO:209) describes an IgEG heavy chain with anti-FRα (MOv18) variable domain, “hole” mutation in the Cε4 domain, and “knob” mutation 76

[0077] in the CH3 domain. The hole mutation is F69G (numbering from the first residue of Cε4) and knob mutation is T366Y (numbering based on Ridgway et al, 1996) and are highlighted in bold and underlining. This sequence further comprises mutations at positions 44 and 135, which aid correct pairing 5 of heavy chain SEQ ID NO:209 with light chain SEQ ID NO:198. This is achieved by moving a cysteine residue involved in formation of an inter-chain disulphide bond from the Cε1 constant domain of to the variable region (G44C, numbering from first residue of variable domain; C14G, numbering from the first residue of Cε1; mutated residues highlighted in bold and underlining). 10 SEQ ID NO:209: QVQLQQSGAELARPGASVKLSCKASDYIFTNYDITWVKQRPGQCLEWIGEIDPRSGKSYYNEKFKGKSTLTADKS SSTAYMELRSLTSEDSAVYFCATMYYYGSSPPMDYWGQGTSVTVSSASTQSPSVFPLTRGCKNIPSNATSVTLGC LATGYFPEPVMVTWDTGSLNGTTMTLPATTLTLSGHYATISLLTVSGAWAKQMFTCRVAHTPSSTDWVDNKTFSV CSRDFTPPTVKILQSSCDGGGHFPPTIQLLCLVSGYTPGTINITWLEDGQVMDVDLSTASTTQEGELASTQSELT 15 LSQKHWLSDRTYTCQVTYQGHTFEDSTKKCADSNPRGVSAYLSRPSPFDLFIRKSPTITCLVVDLAPSKGTVNLT WSRASGKPVNHSTRKEEKQRNGTLTVTSTLPVGTRDWIEGETYQCRVTHPHLPRALMRSTTKTSGPRAAPEVYAF ATPEWPGSRDKRTLACLIQNFMPEDISVQWLHNEVQLPDARHSTTQPRKTKGSGFFVGSRLEVTRAEWEQKDEFI CRAVHEAASPSQTVQRAVSVNPGKRSEPKSSDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVV DVSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKA 20 KGQPREPQVYTLPPSREEMTKNQVSLYCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLYSKLTVDK SRWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK SEQ ID NO:209 comprises the following domains: MOv18 heavy chain variable region, comprising mutation G44C (SEQ ID NO:193): Cε1 comprising mutation C14G (SEQ ID NO:194): 25 Cε2 (SEQ ID NO:183) Cε3 (SEQ ID NO:184) Cε4 comprising mutation F69G (SEQ ID NO:195). Linker (SEQ ID NO:186) CH2 (SEQ ID NO:187) 77

[0078] and CH3 comprising mutation T366Y (SEQ ID NO:188) With the sequences above, following bispecific IgEG constructs combinations were made: 1. Trastuzumab x MOv18 IgEG cis [i.e. anti-Her2 heavy chain SEQ ID NO:180; anti- 5 Her2 kappa light chain SEQ ID NO:191; anti-FRα heavy chain SEQ ID NO:192; anti- FRα kappa light chain SEQ ID NO:198] 2. Trastuzumab x Runimotamab IgEG cis [i.e anti-Her2 heavy chain SEQ ID NO:180; anti-Her2 kappa light chain SEQ ID NO:191; anti-CD3 heavy chain SEQ ID NO:200; anti-CD3 kappa light chain SEQ ID NO:203] 10 3. Trastuzumab x MOv18 IgEG trans [i.e. anti-Her2 heavy chain SEQ ID NO:205; anti- Her2 kappa light chain SEQ ID NO:191; anti-FRα heavy chain SEQ ID NO:209; anti- FRα kappa light chain SEQ ID NO:198] 4. EPS226 x MOv18 IgEG trans [i.e. anti-Her2 heavy chain SEQ ID NO:206; anti-Her2 kappa light chain SEQ ID NO:208; anti-FRα heavy chain SEQ ID NO:209; anti-FRα 15 kappa light chain SEQ ID NO:198] Bio-Layer Interferometry (BLI) of bispecific antibodies Bispecific IGEGs were expressed as described above for bispecific IgE antibodies. Bio-Layer Interferometry (BLI) experiments for anti-Her2 x anti-FRα IGEGs were performed in the Octet R8 system (Sartorius), using similar methods to those described above for bispecific IgE 20 antibodies (see e.g. Figs.1 and 2). Figures 34 and 35 show the results of BLI experiments for bispecific IGEGs binding to Fc receptors. The results indicate that the bispecific IGEG trastuzumab x MOv18 “cis” (i.e. construct 1 above, anti-Her2 x anti-FRα bispecific IGEG with knobs-into-holes mutations in cis form) is capable of binding to human FcεR1, FcγR1, and FcRn receptors. Thus the knob- 25 into-hole mutations present in IGEGs do not impair interaction with FcεR1, FcγR1, and FcRn. Sandwich ELISA for bispecific IgE An enzyme-linked immunosorbent assay (ELISA) for bispecific IGEGs was used to test for simultaneous binding of (a) anti-Her2 x anti-FRα (cis or trans) and (b) anti-Her2 x anti-CD3 78

[0079] IGEGs to Her2 and FRα or CD3, using similar methods to those described above for bispecific IgE antibodies (see e.g. Figure 7). The following IGEGs were tested: (i) trastuzumab x MOv18 IGEG trans (i.e. construct 3 above; anti-Her2 x anti-FRα bispecific 5 IGEG with knobs-into-holes mutations in trans form); (ii) V26 x MOv18 IGEG trans (i.e. construct 4 above; anti-Her2 x anti-FRα bispecific IGEG with knobs-into-holes mutations in trans form; “V26” is an alternative nomenclature for “EPS226”); (iii) trastuzumab x MOv18 IGEG cis (i.e. construct 1 above; anti-Her2 x anti-FRα bispecific 10 IGEG with knobs-into-holes mutations in cis form); and (iv) trastuzumab x runimotamab IGEG cis (i.e. construct 2 above; anti-Her2 x anti-CD3 bispecific IGEG with knobs-into-holes mutations in cis form); The results are shown in Figures 36 to 38. Mean is shown of two technical replicates, error bars are SD. The results show that all four bispecific IGEGs bind to both targets (FRɑ and 15 Her2, or Her2 and CD3) simultaneously by ELISA. In contrast, monospecific IGEGs containing knobs-into-holes (KiH) mutations (i.e. Trastuzumab (KiH) mono IGEG, MOv18 (KiH) mono IGEG, V26 (KiH) mono IGEG or runimotamab (KiH) mono IGEG) did not bind to both antigens in this ELISA. 79

Claims

CLAIMS 1. A heteromultimer comprising at least a first polypeptide and a second polypeptide, wherein: a) the first polypeptide and the second polypeptide each comprise (i) a Cε3 domain 5 and / or a Cε4 domain and (ii) a Cγ2 and / or a Cγ3 domain; b) the Cε3 and / or Cε4 domains of the first and second polypeptides differ in amino acid sequence; optionally wherein the Cγ2 and / or a Cγ3 domains of the first and second polypeptides differ in amino acid sequence; c) the Cε3 and / or Cε4 domains of the first and second polypeptides preferentially 10 heterodimerize, optionally wherein the Cγ2 and / or a Cγ3 domains of the first and second polypeptides preferentially heterodimerize; such that the first and second polypeptides form the heteromultimer.

2. A heteromultimer according to claim 1, wherein the first polypeptide binds specifically to a first target molecule or epitope, and the second polypeptide binds specifically to a 15 second target molecule or epitope different from the first target molecule or epitope.

3. A heteromultimer according to claim 1 or claim 2, wherein the first and second polypeptides each comprise: (i) a modified Cε3 and / or Cε4 domain, wherein the modified Cε3 and / or Cε4 domain of the first polypeptide preferentially associates with the modified Cε3 and / or Cε4 domain 20 of the second polypeptide via a modified interface to form the heteromultimer; and / or (ii) a modified Cγ2 and / or a Cγ3 domain, wherein modified Cγ2 and / or a Cγ3 domain of the first polypeptide preferentially associates with the modified Cγ2 and / or a Cγ3 domain of the second polypeptide via a modified interface to form the heteromultimer.

4. A heteromultimer according to any preceding claim, wherein: 25 (a) (i) the first polypeptide comprises at least one engineered protuberance in the Cε3 and / or the Cε4 domain, (ii) the second polypeptide comprises at least one engineered cavity in the Cε3 and / or the Cε4 domain; and / or (iii) the engineered protuberance and cavity preferentially associate at an engineered interface to promote heterodimerization of the Cε3 and / or the Cε4 domains; and / or 30 (b) (i) the first polypeptide comprises at least one engineered protuberance in the Cγ2 and / or a Cγ3 domain, (ii) the second polypeptide comprises at least one engineered cavity in the Cγ2 and / or a Cγ3 domain; and / or (iii) the engineered protuberance and cavity 80preferentially associate at an engineered interface to promote heterodimerization of the Cγ2 and / or a Cγ3 domains.

5. A heteromultimer according to any preceding claim, wherein: (a) (i) the first polypeptide comprises at least one larger amino acid residue substituted 5 in the Cε3 and / or the Cε4 domain, (ii) the second polypeptide comprises at least one smaller amino acid residue substituted in the Cε3 and / or Cε4 domain; and / or (iii) the larger and smaller amino acid residues preferentially associate to promote heterodimerization of the Cε3 and / or Cε4 domains; and / or (b) (i) the first polypeptide comprises at least one larger amino acid residue substituted 10 in the Cγ2 and / or a Cγ3 domain, (ii) the second polypeptide comprises at least one smaller amino acid residue substituted in the Cγ2 and / or a Cγ3 domain; and / or (iii) the larger and smaller amino acid residues preferentially associate to promote heterodimerization of the Cγ2 and / or a Cγ3 domains.

6. A heteromultimer according to any preceding claim, wherein the first and second 15 polypeptides comprise (i) at least one pair of knobs-into-holes mutations in the Cε3 and / or Cε4 domains; (ii) at least one pair of knobs-into-holes mutations in the Cγ2 and / or a Cγ3 domains; and / or (iii) at least one pair of knobs-into-holes mutations in the Cε3 and / or Cε4 domains and at least one pair of knobs-into-holes mutations in the Cγ2 and / or a Cγ3 domains. 20 7. A heteromultimer according to any preceding claim, wherein: (a) (i) the first polypeptide and the second polypeptide each comprise a Cε4 domain; (ii) the Cε4 domains of the first and second polypeptides differ in amino acid sequence; and / or (iii) the Cε4 domains of the first and second polypeptides preferentially heterodimerize; and / or 25 (b) (i) the first polypeptide and the second polypeptide each comprise a Cγ3 domain; (ii) the Cγ3 domains of the first and second polypeptides differ in amino acid sequence; and / or (iii) the Cγ3 domains of the first and second polypeptides preferentially heterodimerize; such that the first and second polypeptides form the heteromultimer. 30 8. A heteromultimer according to any preceding claim, wherein the first and / or second polypeptides further comprise a Cε1 and / or Cε2 domain.

9. A heteromultimer according to any preceding claim, wherein the first and / or second polypeptides further comprise an immunoglobulin variable domain. 8110. A heteromultimer according to any preceding claim, wherein the first and / or second polypeptides comprise an immunoglobulin E heavy chain fused to a Cγ2 and / or a Cγ3 domain.

11. A heteromultimer according to any preceding claim, further comprising one or more 5 immunoglobulin light chains.

12. A heteromultimer according to any preceding claim, which is a bispecific antibody.

13. A heteromultimer according to any preceding claim, wherein the Cε4 domain of the first or second polypeptide comprises an amino acid sequence as defined in SEQ ID NO:185 or SEQ ID NO:195, or a variant thereof having at least 85%, at least 90%, at least 95% 10 or at least 99% sequence identity to SEQ ID NO:6 or SEQ ID NO:

15.

14. A heteromultimer according to any preceding claim, wherein the first or second polypeptide comprises an amino acid substitution at position 26 and / or 69 of the Cε4 domain as defined in SEQ ID NO:

189.

15. A heteromultimer according to any preceding claim, wherein the first polypeptide 15 comprises a larger amino acid residue at position 26 of the Cε4 domain, and the second polypeptide comprises a smaller amino acid residue at position 69 of the Cε4 domain.

16. A heteromultimer according to any preceding claim, wherein the first or second polypeptide comprises (i) a tryptophan or phenylalanine residue at position 26 of the Cε4 domain and / or (ii) a glycine residue at position 69 of the Cε4 domain. 20 17. A heteromultimer according to any preceding claim, wherein the first polypeptide comprises an amino acid substitution A26W or A26F in the Cε4 domain, and / or the second polypeptide comprises an amino acid substitution F69G in the Cε4 domain.

18. A heteromultimer according to any preceding claim, wherein the first or second polypeptide comprises an amino acid sequence as defined in any of SEQ ID NO:s 180, 25 185, 188, 192, 193, 194, 195, 196, 200, 201, 205, 206, or 209, or a variant thereof having at least 85%, at least 90%, at least 95% or at least 99% sequence identity thereto.

19. A heteromultimer according to any preceding claim, wherein the first or second polypeptide binds specifically to a cancer antigen, preferably HER2 or folate receptor alpha. 30 20. A heteromultimer according to any preceding claim, wherein the first or second polypeptide binds specifically to a T cell antigen, preferably CD3.

21. A pharmaceutical composition comprising a heteromultimer as defined in any preceding claim and a pharmaceutically acceptable excipient, diluent or carrier. 8222. A heteromultimer or pharmaceutical composition as defined in any preceding claim, for use in preventing or treating cancer.

23. A nucleic acid sequence that encodes an amino acid sequence as defined in any of SEQ ID NOs.180, 185, 188, 192, 193, 194, 195, 196, 198, 200, 201, 203, 205, 206, or 209. 5 24. An expression vector comprising one or more nucleic acid sequences as defined in claim 23 operably linked to a promoter suitable for expression in eukaryotic cells.

25. A host cell comprising one or more nucleic acid sequences as defined in claim 23 or one or more expression vectors as defined in claim 24.

26. A method of producing a heteromultimer as defined in any of claims 1 to 20, comprising 10 culturing a host cell as defined in claim 25 under conditions for expression of the first and second polypeptide sequences, and recovering the heteromultimer from the host cell culture. 83

Images