Heterodimers and Purification Thereof

Field of the Invention The present invention relates to proteins comprising an Fc fragment monomer with reduced ability to bind to protein G. The invention also relates to proteins comprising heterodimeric Fc fragments and methods for selective purification thereof, based on the modulation of protein A (pA) and protein G (pG) superantigen binding sites found in the Fc domain. Specific modifications are introduced in each Fc region of the heavy chain of the heterodimeric immunoglobulin to eliminate the affinity for pA and / or pG. Substitutions that eliminate the affinity for pA may be present in one heavy chain of the heterodimeric immunoglobulin, and substitutions that eliminate the affinity for pG may be present in the other heavy chain of the heterodimeric immunoglobulin, thereby enabling purification of the heterodimeric immunoglobulin using a combination of pA and pG affinity chromatography.

Substitutions may be introduced to reduce or eliminate from the Fc region the affinity for pA and pG in a monomer or heterodimer, so that purification of antibody like formats may be carried out through the differential binding to other affinity ligands such as protein L (pL), VKappa and/or VLambda light chain affintiy resins.

Background

Protein A (pA) is a 42 kDa surface protein originally found in the cell wall of the bacterium Staphylococcus aureus. It is useful in purification methods because it can bind immunoglobulins. It is composed of five homologous immunoglobulin (Ig)-binding domains that fold into a three -helix bundle. Each domain is able to bind proteins from many mammalian species, most notably Immunoglobulin G (IgG). Protein A binds the heavy chain within the Fc region of most immunoglobulins. It also binds within the Fab region in the case of the human VH3 family.

Protein G (pG) is an immunoglobulin-binding protein expressed in group C and G Streptococcal bacteria, much like pA, but with differing binding specificities. It is a 65 -kDa (G148 protein G) and a 58 kDa (C40 protein G) cell surface protein that has found application in purifying antibodies through its binding to the Fab and Fc region.

Protein L (pL) is a bacterial protein originally derived from Peptostreptococcus magnus and binds to the variable domain of kappa I, III and IV subclasses (Bjorck L (1988) J. Immunol. 140, 1194-97; Nilson BH et al. (1992) J. Biol. Chem. 267, 2234-2239). Immobilised pL and its recombinant derivatives have been used for many years in the purification of immunoglobulins and fragments thereof containing a Vkappa I, III or IV subclass variable domain. Alternatives to these resins have been developed, such as 'KappaSelect' and 'LambdaFabSelect' (GE Healthcare) which bind to the constant domains of the VKappa light chain or Vlambda light chain, respectively.

Immobilized pA from Staphylococcus aureus and pG from G group Streptococci have been used for many years to purify antibodies from a variety of species. The high selectivity and stability of pA and pG have made them a popular choice for the purification of antibodies from a wide range of sample sources, including serum, ascitic fluid, and hybridoma cell culture supernatants. Mammalian antibodies are categorized into five major classes, i.e. IgA, IgD, IgE, IgG, and IgM. IgG is the predominant class of antibody in serum and is generated in large amounts during the secondary immune response. The IgG class of antibody is further divided into subclasses that vary depending upon the species and the properties of the heavy chain component. There are four subclasses of IgG in humans (IgGl, IgG2, IgG3, IgG4) and in mice (IgGl, IgG2a, IgG2b, IgG3). The affinity of pA for IgG varies considerably between species and IgG subtypes and has been extensively characterized (Duhamel RC et al. 1979, J Immunol Methods 31, 211-217; Hober S et al., (2007) J. Chromatog. B Analyt. Technol. Biomed. Life Sci., 848, 40-47). In humans, pA binds with high affinity to IgGl, IgG2, and IgG4, but does not bind the IgG3 Fc fragment. Among the four IgG subtypes in mice, pA has the weakest affinity for IgGl while pG has affinity for all four IgG subclasses.

Van Loghem E et al. (1982) "Staphylococcal protein A and human IgG subclasses and allotypes" Scand J Immunol 15, 275-285) describe that Staphylococcal pA binds molecules belonging to the IgGl, IgG2, and IgG4 subclasses. IgG3 proteins generally do not bind to protein A, except for those coded by the two gamma 3 alleles, which are G3m(u-): G3m(b0,b3,b5,s,v). G3m(u) is located in the CH2 domain. The difference between G3m(u-) and G3m(u+) IgG3 proteins correlates with the sequence at position 339 in the CH2 domain: Ala and Thr respectively. There is another structural difference in the CH3 domain which correlates with pA binding and non-binding: all IgG proteins that bind pA have Histidine (His, H) at position 435, whereas those that do not bind pA, have Arginine (Arg, R) at that position.

WO2011/122011 (Chugai) describes constant domain modifications in terms of altered binding to FcRn, which affect antigen uptake and PK modulation. WO2014/049003 (Glenmark) describes purification of heterodimers and lists contact residues for pG. WO2014/049003 focusses on the 'BEAT' platform (a Fab-Fc/scFv-Fc construct) and methods for purifying this format as a heterodimer.

Methods to produce heterodimeric immunoglobulins are known in the art.

In W095/33844, Lindhofer & Thierfelder describe a simple method in which two distinct immunoglobulin chains were expressed in a single cell. However, this method was limited by the formation of homo-dimeric species over the heterodimer of interest (Kufer P et al.,

(2004) Trends Biotechnol, 22(5): 238-244). Although the immunoglobulins can be engineered to enhance heavy chain heterodimerization (Merchant AM et al., (1998) Nat.

Biotechnol, 16(7): 677-681), and thus favour heterodimer production, these methods still resulted in the production of significant amounts of the unwanted homodimeric immunoglobulins (Jackman J et al., (2010) J Biol Chem., 285(27):20850-9, Klein C et al,

(2012) MAbs, 4(6):653-63).

According to another approach described in WO96/27011, a pair of antibody molecules can be engineered to maximize the percentage of heterodimers that are recovered from recombinant cell culture. In this method, one or more small amino acid side chains from the CH3 interface of the first antibody molecule are replaced with larger side chains (e.g. tyrosine or tryptophan). Compensatory "cavities" of identical or similar size to the large side chain(s) are created on the interface of the second antibody molecule by replacing large amino acid side chains with smaller ones (e.g. alanine or threonine). This provides a mechanism for increasing the yield of the heterodimer over other unwanted end-products such as homodimers[US00573 168A, US007 83076B2, Ridgway JB, Presta LG, Carter P. Protein Eng 1996 Jul; 9(7): 617-21 ; Atwell S, Ridgway JB, Wells JA, Carter P. J Mol Biol 1997 Jul 4; 270(1 ): 26-35.] Techniques for recovery of heterodimers from homodimers based on a differential affinity of the heterodimers for an affinity reagent have been described. Lindhofer H et al., (1995) J Immunol, 155(1): 219-225 described a differential affinity technique that used two different heavy chains from two different animal species, one of which did not bind the affinity reagent pA. Lindhofer et al. (ibid, and US 6, 551, 592) also described the use of two different heavy chains that originated from two different human immunoglobulin isotypes (IgHGl and IgHG3), one of which (IgHG3) does not bind the affinity reagent pA. WO10/151792 (Davis S et al.) described a modified version of this technique which involved the use of the two amino acid substitutions H435R/Y436F described by Jendeberg et al. (Jendeberg et al., (1997) J. Immunol. Methods, 201(1): 25-34) to abrogate the affinity for the reagent pA in one of the heterodimer heavy chains.

A drawback of differential purification techniques based on pA is that some VH3 domains present in the heavy chains may possess pA binding sites that will interfere with the purification methods.

A combination of differential purification techniques has been proposed that is based on a modification of one CHI domain of a heterodimeric antibody for reduced binding to the CaptureSelect® IgG-CHl affinity reagent (PCT Publication No: W013/136186; Fischer N et al.). However, a drawback of this technique is that at least one heavy chain needs to encompass a CHI region to remove both homo-dimers, thereby limiting the scope of this technology. Hence there is need for a technique complementary to the differential pA purification technique, to create a difference in binding to a second affinity reagent, that would ideally bind a region confined to the Fc region of immunoglobulins thereby avoiding the modification of antigen-binding sites, and which is amendable to any antigen binding Fc scaffold format.

The amino acid residues which are involved in pA or pG binding can be deduced from the experimentally-solved crystal structures of immunoglobulins in complex with the bacterial surface proteins (Protein Data Bank (PDB) database; www.pdb.org); however, since the binding sites for pA, pG and FcRn receptor overlap at the same CH2-CH3 domain interface, it is impossible to predict the outcome of any substitution in terms of its effect towards the affinity for either pA or pG, its impact on FcRn affinity or its impact on properties such as stability and melting temperature. Summary of Invention

According to a first aspect, there is provided a protein comprising a human IgG Fc variant fragment monomer capable of binding superantigen protein A but not superantigen protein G [pA+/pG-] comprising serine at position 428, serine at position 434 and optionally histidine at position 436, with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, such substitutions in the IgG Fc variant fragment monomer at positions 428, 434 and optionally 436 reduces or prevents the binding of protein G, enhancing the purification of the protein.

In the context of the present invention, an Fc variant fragment capable of binding protein A but not protein G is represented as [pA+/pG-], an Fc variant fragment capable of binding protein G but not protein A is represented as [pA-/pG+] and an Fc variant fragment capable of binding neither protein A nor protein G is represented as [pA-/pG-] .

In one embodiment, the protein may be a monomeric protein.

According to a second aspect, there is provided a heterodimeric protein comprising a heterodimeric human IgG Fc variant fragment, wherein the heterodimeric Fc variant fragment comprises: a first Fc variant fragment monomer, capable of binding superantigen protein A but not superantigen protein G [pA+/pG-], comprising serine at position 428, serine at position 434 and optionally histidine at position 436 or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24; and a second, Fc variant fragment monomer capable of binding superantigen protein G but not superantigen protein A [pA-/pG+], comprising arginine at position 435 and optionally phenylalanine at position 436, with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, such substitutions in the IgG Fc variant fragment monomer at positions 428, 434 and optionally 436 reduces or prevents the binding of protein G, enhancing the purification of the heterodimeric protein.

In an embodiment of the invention, the heterodimeric Fc variant fragment has one Fc which binds pA, but has very reduced or no binding to pG, while its other Fc binds pG, but has reduced or no binding to pA. In another approach, each heterodimeric Fc variant fragment has reduced or no binding to pA and pG and purification can be achieved via differential affinity of different proteins fused to each Fc of the heterodimeric proteins. Thus, the protein or heterodimeric protein of the invention has increased purity when compared with the wild type.

Preferably, the protein of the first aspect or the heterodimeric protein of the second aspect comprises a human IgG Fc variant fragment monomer capable of binding superantigen protein A but not superantigen protein G, comprising serine at position 428, serine at position 434 and histidine at position 436. Advantageously, the substitutions at these positions in the IgG Fc variant fragment enhance the purification of the protein or heterodimeric protein.

Previous methods to prepare heterodimeric Fc fragments have involved use of the technique of engineering through the CH3 interface. However, a problem is that homodimer contaminants persist (see Table A in Cabrera et al. WO2012058768A1). Advantageously, using the protein or heterodimeric protein of the present invention, as each arm binds only to either pA or pG the sequential purification ensures a highly homogeneous product. This prevents and/ or avoids the need for additional purification steps and associated loss of yield thus generating a purer product. Homodimeric contaminants are likely to interfere with the therapeutic action of the heterodimeric Fc format. In some cases, receptors are known to be activated through dimerisation (e.g. cMET receptor, TNFR1 receptor) which could inadvertently occur though bivalent interaction of a bivalent homodimeric Fc contaminant. Thus, the protein or heterodimeric protein of the present invention advantageously has an increased purity. Additionally, the protein or heterodimeric protein of this invention has superior stability when compared with the wild type, and to that of other heterodimer formats (see Tables A, 3 and 4 in Cabrera et al. WO2012058768A1).

In one embodiment, biologically-active proteins, such as binding scaffolds and receptor domains may be fused to termini of the heterodimeric variant Fc fragment. The native Fc fragment offers many advantages as a module providing tuneable half-life and effector functions. However being homodimeric, consisting of two identical chains, means there are limitations on how protein fusions can be displayed, which by default are bivalent and monospecific like native IgG immunoglobulins. By contrast a heterodimeric Fc enables mono- and multi-specificity, and mono- and multi-valency. Hence, virtually any protein or peptide molecule can be incorporated in a number of combinations and orientations which opens up new targeting and therapeutic possibilities. Recent work with antigen binding scaffolds has shown the dramatic effect that both relative positioning and bi- and tri- specificity can have (Bracks S et al. (2014) Mol. Cancer Ther. 13, 2030-2039; Jost C et al. (2013) Structure 21, 1979-1991; Spangler JB et al. (2012) J. Mol. Biol. 422, 532-544).

Substitutions that eliminate the affinity for both pA and pG can be introduced into each heavy chain of the heterodimeric immunoglobulin, so that purification can be based on the affinity of proteins fused to each Fc sequence, rather than the binding properties of the Fc in the heteroduplex. In embodiments in which both variant Fc of the heterodimer are modified to reduce binding to both pA and pG, differential purification can be achieved based on the binding properties of the biologically-active fusion proteins at the termini of the variant Fc, e.g., differential affinity for pA, pG, protein L, kappa or lambda binding resins. Typically, the Fc variant fragment monomer may be modulated to increase half-life and enable effector functions through the Fc domain.

Advantageously, the protein or heterodimeric protein of the first or second aspect, or the thirteenth or fourteenth aspect has improved stability, purity and immunogenicity when compared to wild type, with no or minimal loss of half-life when compared to wild type.

Typically, the heterodimeric Fc has a Tm value of between 65.9 °C and 80.8 °C for the CH2 and CH3 domains, respectively. This compares to the values of 71.0 °C and 82.7 °C obtained for WT Fc. Advantageously, the heterodimeric Fc of the present invention has improved stability compared to wild type. Typically, the heterodimeric Fc has a half-life of 99 hours, compared with 95 hours for the wild type. Therefore, advantageously, there is no loss in half life when compared with wild type as a result of the amino acid substitutions. Additionally, minimal weak binders to the DRB 1 allotypes were observed with the heterodimeric protein having the substitutions of the present invention, thus improving stability. In naturally-occurring human immunoglobulins of gamma isotype that are known to bind the bacterial surface pA and pG (IgHGl, IgHG2, and IgHG4; Jendeberg et al., (1997) supra and Nezlin R & Ghetie V, (2004) Advances in Immunology, Academic Press, Vol. 82: 155-215), each heavy chain carries a binding site at the CH2-CH3 domain interface for each of these two bacterial surface proteins. Since the binding sites for pA and pG overlap in heavy chains, specific substitutions that reduce or eliminate pG binding are useful to purify heterodimers of heavy chains in a similar manner to the pA-based methods. Differential affinity methods based on modulation of pG binding offer new strategies for the purification of heterodimeric immunoglobulins. Combining both pA and pG differential affinity methods advantageously enables preparation of heterodimers of heavy chains with a high degree of purity and without the need for gradient elution.

Preferably, the heterodimeric protein further comprises: in the first Fc variant fragment threonine at position 407 and cysteine at position 349 and in the second Fc fragment tyrosine at position 366 and cysteine at position 354; or in the first Fc variant fragment threonine at position 407 and cysteine at position 354 and in the second Fc fragment tyrosine at position 366 and cysteine at position 349, with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, the heterodimeric protein comprising these residues has increased stability when compared with the wild type.

These additional constructs were made to see if the formation of heterodimers in this co- expression system could be biased. The 'Knobs into Hole' (KiH) mutations, which were previously shown to bias the formation heterodimeric Fc by Ridgway et al., (1996) (Protein Engineering, 9: 617-621), were incorporated into the above constructs. A disulphide bridge was also introduced with these KiH mutations to see if this would lead to improved stability. The KiH mutations were previously shown by Ridgway et al. to lower the stability of the Fc heterodimer.

In one embodiment, the heterodimeric protein comprises: (a) in the first Fc variant fragment serine at position 428, serine at position 434, histidine at position 436 and in the second Fc fragment arginine at position 435; (b) in the first Fc variant fragment serine at position 428, serine at position 434, histidine at position 436, threonine at position 407 and cysteine at position 349 and in the second Fc fragment arginine at position 435, tyrosine at position 366 and cysteine at position 354; (c) in the first Fc variant fragment serine at position 428, serine at position 434, histidine at position 436 and threonine at position 407 and in the second Fc fragment arginine at position 435 and tyrosine at position 366; or (d) in the first Fc variant fragment serine at position 428, serine at position 434, histidine at position 436, threonine at position 407 and cysteine at position 354 and in the second Fc fragment arginine at position 435, tyrosine at position 366 and cysteine at position 349; or wherein the heterodimeric protein comprises these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, the heterodimeric protein comprising these residues has increased stability when compared with the wild type.

Preferably, the protein of the first aspect comprises a biologically-active protein connected to one or both termini of the Fc variant fragment. Typically, the protein of the first aspect is a monomeric protein.

Preferably, the heterodimeric protein of the second aspect comprises a biologically-active protein connected to one, two, three or four of the termini of the heterodimeric Fc variant fragment.

Preferably, the biologically-active protein comprises one or more protein selected from the group consisting of an antigen-binding moiety, a receptor or domain of a receptor, a cytokine, a peptide, cyclic peptide or peptide derivative, a protein or peptide that hybridises specifically to a partner protein or peptide to create a non-covalent Fc fusion product, a peptide containing cysteine or lysine for conjugation, a hormone, enzyme, chemokine, ligand, toxin, and/ or a biologically active fragment thereof.

Preferably, the heterodimeric protein comprises a biologically-active protein connected to at least two termini of the heterodimeric Fc variant fragment. It is preferred that the heterodimeric protein comprises a biologically-active protein connected to each of the N- termini of the heterodimeric Fc variant fragment. In one embodiment, the biologically-active protein connected to first Fc variant fragment monomer binds to protein A and not protein G, i.e. [pA+/pG-], and wherein the biologically- active protein connected to the second Fc variant fragment monomer binds to protein G and not protein A, i.e. [pA-/pG+]. Such an arrangement assists in purification of the protein to produce a highly pure protein when compared to wild type.

Typically, each biologically-active protein is an antigen-binding moiety.

In one embodiment, the heterodimeric protein comprises: a Fab connected to each of the N- termini of the heterodimeric Fc variant fragment, a scFv connected to each of the N-termini of the heterodimeric Fc variant fragment, a Fab connected to one N-terminus and a scFv connected to the other N-terminus of the heterodimeric Fc variant fragment, a Fab connected to each of the C-termini of the heterodimeric Fc variant fragment, a scFv connected to each of the C-termini of the heterodimeric Fc variant fragment, and/ or a Fab connected to one N- terminus and a scFv connected to the other C-terminus of the heterodimeric Fc variant fragment.

Typically, the Fc variant fragment of the protein of the first aspect, or the Fc variant fragments of the heterodimeric protein of the second aspect, comprises a modification in one or more Fc non- structural loops that join beta strands, wherein such modification confers antigen binding. Advantageously, such modification assists in antigen binding.

Typically, the Fc variant fragment of the protein of the first aspect, or the Fc variant fragments of the heterodimeric protein of the second aspect comprises at least one modification selected from: a modification to increase or decrease Fcgamma receptor binding and thereby modify effector function, a modification to increase or decrease complement activation, a modification to increase or decrease FcRn binding and hence half-life, and the removal or addition of a glycosylation site. Preferably, the constituent Fc variant monomers are derived from different IgG isotypes.

In one embodiment, the Fc fragment monomer of the protein of the first aspect or the heterodimeric protein of the second aspect capable of binding to protein A but not protein G, i.e. [pA+/pG-], is connected to a VH3 family VH fragment capable of binding to pA. Preferably, the Fc fragment monomer of the heterodimeric protein capable of binding to protein G but not protein A, i.e. [pA-/pG+], is connected to a CHI region of IgG (pG+). In one embodiment, the Fc fragment monomer of the heterodimeric protein capable of binding to protein G but not protein A, i.e. [pA-/pG+], is connected to a Fab arm that comprises a CHI region of IgG, i.e. [pG+].

Preferably, the protein or heterodimeric protein comprises a Vlambda and/or Vkappa VL domain or a fragment thereof.

Typically, the Fc fragment monomer of the heterodimeric protein capable of binding to protein G but not protein A, i.e. [pA-/pG+], is connected to a naturally-occurring or modified VH3 family VH fragment that is not capable of binding to pA, i.e. [pA-].

In one embodiment, the modified VH3 family VH has a modification at one or more positions at the interface of the binding site between pA and VH3.

Preferably, the modified VH3 family VH has a modification at one or more positions selected from: 15, 17, 19, 57, 59, 64, 65, 66, 68, 70, 81, 82a and/ or 82b.

In one embodiment, the Fc fragment monomer of the protein of the first aspect or the heterodimeric protein of the second aspect capable of binding to protein A but not protein G, i.e. [pA+/pG-], is connected to a VH1, 2, 4, 5, 6, or 7, family VH fragment that is not capable of binding to pA.

Preferably, the heterodimeric protein comprises a scFv, Fab, VHH, VH, or dAb comprising a VH or a biologically-active fragment thereof. Preferably, the Fc fragment monomer arm of the heterodimeric protein capable of binding to protein G, i.e. [pG+], is connected to a kappa or lambda light chain.

Typically, the heterodimeric protein comprises a first scFv connected to the first Fc monomer fragment and a second scFv connected to the second Fc monomer fragment. In one embodiment, the first and second scFv comprise different variable heavy domains (VH). In one embodiment, the first and second scFv each comprise the same variable light domain (VL).

Preferably, the first and second scFv each comprise a different variable light domain (VL). Typically, each VL is a kappa VL (VK) or a lambda VL (νλ).

Preferably, the first and second scFv each comprise a different isotype of variable light domain (VL). In one embodiment, the first scFv comprises a VK and the second scFv comprises a νλ.

In one embodiment, the first scFv comprises a νλ and the second scFv comprises a VK.

Preferably, the first Fc fragment monomer [pA+/pG-] is connected to a VH and the second Fc fragment monomer [pA-/pG+] is connected to a νλ or VK.

Preferably, the heterodimeric protein is a half mAb' in which the second Fc fragment monomer [pA-/pG+] is connected to a Fab comprising a hinge region, CHI domain and VH [pA-] associated with a light chain comprising a CL and VL (νλ or VK)

It is preferred that the heterodimeric protein comprises a first Fab connected to an Fc comprising the first Fc monomer fragment [pA+] and a second Fab connected to an Fc comprising the second Fc monomer fragment [pG+] . In one embodiment, the heterodimeric protein comprises: a first Fab connected to an Fc comprising the first Fc monomer fragment which is [pA+], wherein the first Fab is [pA+] or [pA-], and wherein the first Fab comprises a CH I that is [pG-] connected to a VH and an associated light chain that comprises a CL connected to a VL (λ or K); and a second Fab connected to an Fc comprising the second Fc monomer fragment which is [pG+], wherein the second Fab is [pG+] or [pG-] and comprises a CHI that is [pG+] or [pG-] connected to a VH that is [pA-] and an associated light chain that comprises a CL connected to a VL (λ or κ).

In one embodiment, the first and second Fabs are from the same isotype or class.

In another embodiment, the first and second Fabs are from a different isotype or class.

Preferably, the heterodimeric protein comprises a heterodimeric IgG Fc fragment, wherein the heterodimeric IgG Fc fragment comprises first and second Fc fragment monomers, wherein the first and second Fc fragment monomers are incapable of binding superantigen A and superantigen G [pA-/pG-] and comprise serine at position 428, serine at position 434, histidine at position 436 and arginine at position 435; wherein one of the monomers is connected to a [pA+ / pG-] Fab and the other monomer is connected to a [pA-/pG+] Fab; wherein the [pA+ / pG-] Fab comprises a CHI that is [pG-] connected to a VH that is [pA+] and an associated light chain that comprises a CL connected to a VL (λ or K); wherein the [pA- / pG+] Fab comprises a CHI that is [pG+] connected to a VH that is [pA-] and an associated light chain that comprises a CL connected to a VL (λ or κ). Advantageously, the heterodimeric protein has enhanced purity and stability when compared to wild type.

According to a third aspect, there is provided an immunoglobulin (Ig) Fc fragment monomer or heterodimeric protein according to the first or second aspect, wherein each Fc fragment monomer comprises a sequence extending from the start of the first beta strand in CH2 to the final residue of the G strand of CH3.

Preferably, each Fc fragment monomer comprises: a sequence from serine 239 to serine 442 or lysine 447; a sequence from proline 238 to serine 442 or lysine 447; a sequence from glycine 236 to serine 442 or lysine 447; a sequence from alanine 231 to serine 442 or lysine 447; a sequence from cysteine 226 to serine 442 or lysine 447; or a sequence from threonine 225 to serine 442 or lysine 447; with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, the Fc fragment monomer has enhanced purity and stability when compared to wild type. Preferably, the protein of the first aspect or the heterodimeric protein of the second aspect comprises a first Fc (full length) monomer comprising the first Fc fragment monomer and a second Fc (full length) monomer comprising the second Fc fragment monomer. Preferably, the first and/ or second Fc monomers are each selected from an IgG 1, 2, 3, or 4.

Preferably, the first and/ or second Fc monomers are human or variant human sequences.

It is preferred that the heterodimeric protein comprises a first antigen-binding domain and a second-antigen binding domain, wherein the first and/ or second antigen-binding domains bind the same target or bind different targets.

Preferably, the heterodimeric protein comprises a first antigen-binding domain and a second antigen-binding domain, wherein the first and/ or second antigen-binding domains bind different epitopes on the same target.

Typically, the target or targets are selected from: a tumour-associated antigen, bacterial, fungal or viral antigen, cytokine, interleukin, growth factor, immunomodulatory, immune- checkpoint molecule and/ or an antigen enabling the retargeting of effector cells.

According to a fourth aspect, there is provided a vector or set of vectors comprising nucleic acid encoding a protein or a heterodimeric protein according to the first, second or third aspect. According to a fifth aspect, there is provided an in vitro host cell comprising nucleic acid encoding a protein or heterodimeric protein according to the first, second or third aspect.

According to a sixth aspect, there is provided a composition comprising a protein or a heterodimeric protein according to the first, second or third aspect, further comprising an excipient.

According to a seventh aspect, there is provided a heterodimeric protein according to the second or third aspect for use as a medicament. According to an eighth aspect, there is provided the use of a heterodimeric protein according to the second or third aspect in the manufacture of a medicament for the treatment of a disease or condition. According to a ninth aspect, there is provided a method of treatment of the human or animal body comprising administration of a heterodimeric protein according to the second or third aspect or a composition according to the sixth aspect. According to a further aspect, there is provided use of a heterodimeric protein according to the second or third aspect or a composition according to the sixth aspect for treatment of a disease or condition.

According to a tenth aspect, there is provided a method for purification of a heterodimeric protein according to the second aspect, the method comprising the steps of: (a) providing a mixture comprising the first and second monomers and/ or heterodimers thereof; (b) chromatographic separation using protein A; and (c) chromatographic separation using protein G; thereby isolating heterodimeric protein comprising the heterodimeric fragment.

Preferably, step (b) precedes step (c). In another embodiment, step (c) precedes step (b).

Preferably, the method comprises purification of a heterodimeric IgG Fc variant fragment, wherein the heterodimeric IgG Fc variant fragment comprises a first [pA+/pG-] monomer and second [pA-/pG+] monomer, said first [pA+/pG-] monomer comprising a first IgG Fc variant fragment monomer, capable of binding superantigen A, but not superantigen G [pA+/pG-], comprising serine at position 428, serine at position 434 and optionally histidine at position 436 and said second [pA-/pG+] monomer comprising a second, IgG Fc variant fragment monomer, capable of binding superantigen G [pG+], comprising arginine at position 435, and optionally phenylalanine at position 436, with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat, said purification methods comprising:

- providing a mixture of comprising the first and second monomers and heterodimers thereof,

- chromatographic separation using protein A.

- chromatographic separation using protein G,

- thereby isolating heterodimeric protein comprising the heterodimeric fragment. Preferably, the step of chromatographic separation using protein A precedes chromatographic separation using protein G, or the step of chromatographic separation using protein G precedes chromatographic separation using protein A.

Preferably, the pA and / or pG chromatographic separation resins are specific to the Fc region. Examples of such resins are: a pA based resin called MabSelect Sure™ (GE Healthcare) which is reportedly Fc specific (GE Healthcare Application Note 29-0515-20 AA); and a protein G that has been modified to bind Fc only (Muir N, (2009) PhD Thesis: Studies on the interaction of a single domain of protein G from Streptococcus with human Fc and Fab, University of Southampton, School of Biological Sciences).

Preferably, the heterodimeric protein comprises a monomer capable of binding protein L and a monomer that is not capable of binding protein L, and / or the heterodimeric protein comprises a monomer capable of binding a kappa light chain resin and a monomer that is not capable of binding a kappa light chain resin, and / or the heterodimeric protein comprises a monomer capable of binding a lambda light chain purification resin and a monomer that is not capable of binding a lambda light chain purification resin. Preferably, the heterodimeric protein comprises a monomer capable of binding protein L and a monomer that is not capable of binding protein L, wherein chromatographic separation using protein L is used in addition to chromatographic separation using protein A/ or and protein G, or as an alternative to a chromatographic separation step using protein A and/ or protein G, thereby isolating heterodimeric protein comprising the heterodimeric fragment.

Preferably, the heterodimeric protein comprises a monomer capable of binding a kappa light chain affinity purification resin and a monomer that is not capable of binding a kappa light chain affinity purification resin, wherein chromatographic separation using a kappa light chain affinity purification resin is used in addition to chromatographic separation using protein A and/ or protein G or as an alternative to a chromatographic separation step using protein A and/ or protein G, thereby isolating heterodimeric protein comprising the heterodimeric fragment. Preferably, the heterodimeric protein comprises a monomer capable of binding a lambda light chain affinity purification resin and a monomer that is not capable of binding a lambda light chain affinity purification resin, and chromatographic separation using a lambda light chain affinity purification resin is used in addition to chromatographic separation using protein A and/ or protein G or as an alternative to a chromatographic separation step using protein A and/ or protein G, thereby isolating heterodimeric protein comprising the heterodimeric fragment.

According to an eleventh aspect, there is provided a method for purification of a heterodimeric protein according to the second aspect of the invention, wherein the first and/ or second monomers are linked to an scFv; wherein, one of the scFv comprises a kappa light chain and the other scFv comprises a lambda light chain, wherein the method comprises the steps of: (a) providing a mixture of comprising the first and second monomers and heterodimers thereof; (b) chromatographic separation using protein G; (c) chromatographic separation using protein L, a kappa light chain purification resin or a lambda light chain purification resin, dependent on the light chain associated with the Fc having the [pG-/pA+] phenotype, thereby isolating heterodimeric protein comprising the heterodimeric fragment.

According to a twelfth aspect, there is provided a method for purification of a heterodimeric protein wherein the first monomer comprises a scFv comprising a kappa light chain VK capable of binding pL and the second monomer comprises a scFv comprising a kappa light chain VK incapable of binding pL such as VKII class, natural VK variants or those modified so as not bind pL (e.g. VKI class carrying an S 12P mutation), wherein positions are defined with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat, wherein the method comprises the steps of: (a) providing a mixture of comprising the first and second monomers and heterodimers thereof; (b) chromatographic separation using protein G; (c) chromatographic separation using protein L and / or a kappa light chain purification resin.

Preferably, the step of chromatographic separation using protein G precedes the step of chromatographic separation using protein L and/or kappa light chain purification resin or wherein the step of chromatographic separation using protein L, and/or kappa light chain purification resin precedes the step of chromatographic separation using protein G. Advantageously, the method of the eleventh or twelfth aspect provides for the enhanced purification of the heterodimeric protein, producing a protein that has increased purity and stability when compared to wild type.

Previous methods to prepare heterodimeric Fc fragments uses the technique of engineering through the CH3 interface. However, a problem is that homodimer contaminants persist (see Table A in Cabrera et al. WO2012058768A1). Advantageously, using the protein of heterodimeric protein of the present invention, as each arm binds only to either pA or pG the sequential purification ensures a highly homogeneous product. This prevents the need for additional purification steps and associated loss of yield to generate a purer product. Homodimeric contaminants are likely to interfere with the therapeutic action of the heterodimeric Fc format. In some cases receptors are known to be activated through dimerisation (e.g. cMET receptor, TNFR1 receptor) which could inadvertently occur though bivalent interaction of a bivalent homodimeric Fc contaminant. Thus, the protein or heterodimeric protein of the present invention has an increased purity. Additionally, the protein or heterodimeric protein has superior stability when compared with the wild type.

According to a thirteenth aspect, there is provided a protein comprising an IgG Fc fragment monomer that is not capable of binding superantigen protein A or protein G [pA-/pG-] comprising serine at position 428, serine at position 434, arginine at position 435 and optionally histidine at position 436, wherein positions are defined with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, the protein comprising these mutations has increased purity and stability when compared with wild type.

According to a fourteenth aspect, there is provided a heterodimeric protein comprising a heterodimeric IgG Fc fragment, wherein the heterodimeric Fc fragment comprises: first and second Fc fragment monomers that are not capable of binding superantigen protein A or superantigen protein G [pA-/pG-] comprising serine at position 428, serine at position 434, arginine at position 435 and optionally histidine at position 436, wherein positions are defined with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, the protein comprising these mutations has increased purity and stability when compared with wild type.

Advantageously, the protein or heterodimeric protein of the first, second, thirteenth or fourteenth aspect allows for generation of an immunoglobulin which is suitable for the target of pathogens that normally evade the immune system through the expression of superantigens. The protein or heterodimeric protein may advantageously be purified through the use of affinity resins which bind the Fab arms and those that are not derived from superantigens. Advantageously, reducing the number of pA and pG sites will reduce the avidity effect that is observed in immunoglobulins that have binding sites in both the Fc and Fab arms, and thus allow for milder and less extreme pH elution conditions so as to improve the likelihood of generating a more homogeneous, monomeric product.

Preferably, the first Fc fragment monomer is connected to a biologically-active protein that is capable of binding superantigen protein A, but not superantigen protein G [pA+/pG-] and the second Fc fragment monomer is connected to a biologically-active protein that is capable of binding superantigen protein G, but not superantigen protein A [pA-/pG+] .

Preferably, the first Fc fragment monomer is connected to a biologically-active protein that is capable of binding protein L and the second Fc fragment monomer is connected to a biologically-active protein that is not capable of binding protein L, but optionally is capable of binding a protein A, G or Vlambda affinity resin.

Preferably, the first Fc fragment monomer is connected to a biologically-active protein comprising a Vkappa that is capable of binding a kappa light chain purification resin and the second Fc fragment monomer is connected to a biologically-active protein that is not capable of binding a kappa purification resin, but optionally is capable of binding a protein A, G or Vlambda affinity resin.

Typically, one Fc fragment monomer is connected to a biologically-active protein comprising Vlambda that is capable of binding a lambda light chain purification resin and the other Fc fragment monomer is connected to a biologically-active protein that is not capable of binding a lambda purification resin, but optionally is capable of binding a protein A, G or L affinity resin or VKappa light chain purification resin. According to a fifteenth aspect, there is provided a method for purification of a heterodimeric protein according to the fourteenth aspect, wherein the first Fc fragment monomer is connected to a biologically-active protein that is capable of binding superantigen protein A, but not superantigen protein G [pA+/pG-] and the second Fc fragment monomer is connected to a biologically-active protein that is capable of binding superantigen protein G, but not superantigen protein A [pA-/pG+], the method comprising the steps of: (a) providing a mixture of comprising the first and second monomers and heterodimers thereof; (b) chromatographic separation using protein A and/ or (c) chromatographic separation using protein G, thereby isolating heterodimeric protein comprising the heterodimeric fragment.

According to a sixteenth aspect, there is provided a method for purification of a heterodimeric protein according to the fourteenth aspect, wherein the first Fc fragment monomer is connected to a biologically-active protein that is capable of binding protein L and the second Fc fragment monomer is connected to a biologically-active protein that is not capable of binding protein L, but optionally is capable of binding a protein A, G or Vlambda affinity resin, the method comprising the steps of: (a) providing a mixture of comprising the first and second monomers and heterodimers thereof; (b) chromatographic separation using protein L; and/ or (c) chromatographic separation using protein A, G or Vlambda affinity resin or gradient elution, thereby isolating heterodimeric protein comprising the heterodimeric fragment.

According to a seventeenth aspect, there is provided a method for purification of a heterodimeric protein according to the fourteenth aspect, wherein the first Fc fragment monomer is connected to a biologically-active protein comprising Vkappa that is capable of binding a kappa light chain affinity purification resin and the second Fc fragment monomer is connected to a biologically-active protein that is not capable of binding a kappa light chain affinity purification resin, but optionally is capable of binding a protein A, G or Vlambda affinity resin, the method comprising the steps of: (a) providing a mixture of comprising the first and second monomers and heterodimers thereof; (b) chromatographic separation using a kappa light chain affinity purification resin; and (c) optionally a chromatographic separation using protein A, G or Vlambda affinity resin or gradient elution, thereby isolating heterodimeric protein comprising the heterodimeric fragment. According to an eighteenth aspect, there is provided a method for purification of a heterodimeric protein according to the fourteenth aspect, wherein one Fc fragment monomer is connected to a biologically-active protein comprising Vlambda that is capable of binding a lambda light chain purification resin and the other Fc fragment monomer is connected to a biologically-active protein that is not capable of binding a lambda purification resin, but optionally is capable of binding a protein A, G or L affinity resin or VKappa light chain purification resin, the method comprising the steps of: (a) providing a mixture of comprising the first and second monomers and heterodimers thereof; (b) chromatographic separation using a lambda light chain affinity purification resin; and/ or (c) chromatographic separation using protein A, G or L affinity resin or VKappa light chain purification resin or gradient elution, thereby isolating heterodimeric protein comprising the heterodimeric fragment.

Advantageously, the method of the fifteenth to eighteenth aspects provides for the enhanced purification of the heterodimeric protein, producing a protein that has increased purity and stability when compared to wild type.

In contrast to naturally-occurring homodimeric immunoglobulins, the heterodimeric immunoglobulins of the present invention may comprise Fc fragments composed of Fc monomers that differ in their affinity to pA and pG (e.g., Fc heterodimers of pA-/pG+ and pA+/pG- variants described herein) or may comprise Fc fragments (pA-/pG-) with differential affinity for pA, pG, pL, lambda or kappa light chain conferred by protein fused to the pA-/pG- Fc dimer or heterodimer. Fusions may include, but are not limited to, full length bispecific antibodies, monovalent Fab-Fc fusions and bispecific scFv-Fc, scFab-Fc and Fab- Fc fusions.

The invention utilises the residue changes found in human IgG3 and feline IgGl Fc, which respectively weaken or abrogate binding to pA and pG, by substituting them into human IgG, in particular human IgGl Fc. Specific substitutions that eliminate the affinity for pA or pG can be introduced in one heavy chain of the heterodimeric immunoglobulin. Substitutions that eliminate the affinity for pA can be introduced in one heavy chain of the heterodimeric immunoglobulin and substitutions that eliminate the affinity for pG can be introduced in the other heavy chain of the heterodimeric immunoglobulin, thereby providing heterodimeric immunoglobulin and methods to readily purify the heterodimeric immunoglobulin using a combination of pA and pG affinity chromatography. A purification scheme in accordance with the present invention enables purification of Fc heterodimer very cleanly by the sequential application to pA and pG resins (verified by mass spectrometer data), the properties of the purified heterodimeric Fc confer a slightly better half-life (when compared to wild type, WT) in the mice PK studies.

It is preferred that substitutions that eliminate the affinity for both pA and pG may be introduced into each heavy chain of the heterodimeric immunoglobulin, so that purification can be based on the affinity of proteins fused to each Fc sequence, rather than the binding properties of the Fc in the heteroduplex. In embodiments in which both variant Fc of the heterodimer are modified to reduce binding to both pA and pG, differential purification can be achieved based on the binding properties of the biologically-active fusion proteins at the termini of the variant Fc, e.g., differential affinity for pA, pG, protein L, kappa or lambda binding resins.

In one embodiment, the variant Fc of the protein or heteroprotein of the first or second aspect may be modulated to increase half-life and enable effector functions through the Fc domain.

Unless otherwise indicated, positions are numbered according to EU numbering system of Edelman GM et al., (1969) Proc Natl Acad Sci USA, 63: 78-85. Numbering of amino acid residues in the VH3 region is according to Kabat numbering (Kabat, EA et al. (1991) Sequences of proteins of immunological interest. 5 ^th Edition- US Department of Health and Human Services, NIH Publication No. 91, 3242). Detailed Description of the Invention

The invention will be further described by way of example and with reference to the following figures, wherein: Figure 1 is a schematic representation of a two-step affinity purification strategy to generate heterodimer Fc and derivatives thereof based on modulation of the superantigen binding sites on co-expressed Fc chains. Figure 2 shows an alignment of the human IgGl Fc region to feline IgG Fc (top) and human IgG3 (bottom). The residues that interact with pA and/or pG are shown in Table 4. SEQ ID NO 22 shows the human IgGl sequence and SEQ ID NO 23 shows the feline IgGl sequence. SEQ ID NO 24 shows the human IgG3 Fc sequence.

Figure 3 shows the results of the biolayer interferometry (Octet) binding kinetic studies of IgGl-Fc homodimer mutants to pA and pG produced from Pichia pastoris. Binding curves of the mutants and WT IgGl-Fc to pA (LHS column) and pG (RHS column) are shown. Figure 4 shows the DSC profiles of the IgGl-Fc homodimer mutants and wild type.

Figure 5 shows the CD spectra of the IgGl-Fc homodimer mutants and wild type. Figure 5A illustrates spectra of the pA-/pG+ mutants compared to wild type, and Figure 5B shows spectra of the pA+/pG- mutants compared to wild type.

Figure 6 shows the SEC profiles IgGl-Fc homodimer mutants and wild type. The bottom profile is that of the molecular weight standards.

Figure 7 shows SDS-PAGE analysis of the steps during the pG and pA purification of the heterodimeric IgGl Fc fragment (Lanes 2-6) and protein G purification of WT IgGl-Fc (Lanes 7-9) from transiently transfected HEK293-E cells supernatant. Lanes 1 and 10: Molecular weight markers. Lane 2: Cell supernatant of co-expressed H435R and M428S/N434S/Y436H IgGl Fc fragments. Lane 3: Protein G flow through. Lane 4: Protein G elution at pH3. Lane 5: Protein A flow through. Lane 6: Protein A elution at pH3.5. Lane 7: WT IgGl-Fc Cell supernatant. Lane 8: Protein G flow through. Lane 9: Protein G elution at pH3.

Figure 8 (A to E) shows the DSC profiles of the IgGl-Fc heterodimer variants and wild type. The table shows the Tm values for wild type and heterodimer IgGl-Fc and those derived from the fitted curves for the KiH mutants.

Figure 9 shows the results of the biolayer interferometry (Octet) binding kinetic studies of the purified IgGl-Fc heterodimer and WT IgGl-Fc to pA and pG. Figure 10 shows biolayer interferometry (Octet) data for the binding of the IgGl-Fc proteins at 1000 nM to anti-Fc coated tips: (A) PBS buffer only, (B) IgGl-Fc WT, (C) H435R IgGl- Fc homodimer, (D) M428S/N434S/Y436H IgGl-Fc homodimer, and (E) IgGl-Fc heterodimer.

Figure 11 shows the surface plasmon resonance (Biacore) results of the binding of purified IgGl-Fc heterodimer to human FcRn. Figure 11 A shows the Biacore binding curves for a 1 :1 dilution series of the heterodimer IgGl-Fc starting at 500nM. Figure 1 IB provides an overlay of the binding curves to FcRn at an Fc fragment concentration of 250nM.

Figure 12 shows the SEC profiles of (A) the affinity purified IgGl-Fc heterodimer and wild type Fc; and (B) affinity purified IgGl-Fc heterodimer and versions carrying the KiH mutations. Figure 13 illustrates the pharmacokinetics of the IgGl-Fc heterodimer, clearance curves of the IgGl-Fc heterodimer and wild type IgGl-Fc in mouse.

Figure 14. SEC of WT and heterodimeric IgGl Fc (lmg/ml) after 1 day at: (A) -80°C, (B) 4°C and (C) 25°C; and after 7 days at: (D) -80°C, (E) 4°C and (F) 25°C. The traces are almost superimposable except for a small peak evident for the heterodimer that appears at 17.5min.

Figure 15A shows the SEC profiles of the affinity purified heterodimers and the parental H10-03-6 Fcab. Figure 15B shows the Biacore binding curves to HER2 of a 1:1 dilution series of the WT/H10-03-6 heterodimer Fcab starting at 125nM.

Figure 15C shows the binding curves of the heterodimeric Fcab constructs in comparison to the parental Fcab HI 0-03-6 and Herceptin. The Fcabs were at 125nM and Herceptin at 4nM.

Figure 16 is a schematic representation of the a protein or heterodimeric protein according to the invention, showing X, Y, A, and B which correspond to a binding protein (e.g., VHH, Fn domain, domain antibody (dAb), Darpin abdurin, adhiron); soluble receptor domain; soluble ligand; peptide; a protein or peptide that hybridises specifically to a partner (protein or peptide) to create a non-covalent Fc fusion product (e.g. leucine zipper); a peptide containing reactive side chains such as cysteine for subsequent conjugation of molecules to; and combinations of the above;. Figure 17 is a schematic representation of the generation of a bispecific scFv heterodimer Fc by sequential pA and pG purification which incorporates any VH or VL class. Alternatively, this format can be generated using pA and pG resins that are reportedly specific to the Fc site only. In this scenario a pA positive VH3 domain can be used on both or either chains. Figure 18 is a schematic representation of the generation of a bispecific scFv heterodimer Fc incorporating a Protein L light chain affinity resin purification step.

Figure 19 is a schematic representation of the VH and VL formats in accordance with the invention. An Fv reformatted as a monovalent binder using the heterodimer Fc platform. As 1: 1 binding is essential for some antigen targets, a sequential pA and pG purification can be used to ensure heterodimerisation which incorporates any VH or VL class.

Figure 20 is a schematic representation of the generation of a 'half mAb' (mAb 1/2 ) format using the platform of the invention. As noted previously considerations of Fab arm binding properties can be ignored if pA and pG based resins are available which are specific to the Fc site only.

Figure 21 is a schematic representation of a VH and VL format example based on the platform showing common light chain bispecific mAb - Fabs from the same isotype or class. As noted considerations of Fab arm binding properties can be ignored if pA and pG based resins are available which are specific to the Fc site only.

Figure 22 is a schematic representation of the purification of a scFvFab- Fc format based on scFv incorporating a VH3 which is pA+ and opposing full heavy chain possessing a CHI which is pG+.

Figure 23 is a schematic representation of the purification of a pA-/pG- IgG-Fc fusion (H435R, M428S, N434S, Y436H) is dependent on the properties of the molecules fused to the N- and/or C-termini. Figure 24 shows a human IgGl to 4 subclass amino acid sequence alignment, residues 428, 434, 435 and 436 modified according to the invention are indicated in bold. These sequences are shown by SEQ ID NOs 25 to 28, respectively.

Figure 25 shows the sequences of the WT human IgGl-Fc in pPICZaA. (A) Amino acid sequence of the IgGl-Fc (bold type face). (B) SEQ ID NO 29 shows the amino acid sequence and SEQ ID NO 30 shows the corresponding nucleotide sequence. (C) Alignment of the amino acid and nucleotide sequences.

Figure 26 shows the IgGl-Fc amino acid and DNA coding sequences used in the pTT5 vector. (A) Amino acid sequence of the IgGl-Fc (bold type face), as shown in SEQ ID NO 31. The leader sequence is shown in italics, this is cleaved from the final product. (B) Nucleotide sequence of the IgGl-Fc shown in A, as shown in SEQ ID NO 32. The underlined sequence is Sapl restrictions site. (C) Nucleotide sequence of the IgGl-Fc showing the introduced EcoNl restriction site, as shown in SEQ ID NO 33. (D) An alignment of the amino acid and DNA coding sequence.

Figure 27 shows the amino acid sequences of human IgGl constant region (Uniprot P01857) incorporating CH2 and CH3 domains (as defined in Uniprot) of the Fc fragment, showing:

(a) full Uniprot sequence of the constant region of the wild-type human IgGl, as shown in SEQ ID NO 34;

(b) wild-type human IgGl Fc fragment incorporating CH2 and CH3 domains (as defined in Uniprot), as shown in SEQ ID NO 35;

(c) human IgGl Fc fragment with M428S/N434S modification, as shown in SEQ ID NO 36;

(d) human IgGl Fc fragment with M428S/N434S/Y436H, as shown in SEQ ID NO 37; and

(e) human IgGl Fc fragment with M428S/N434S/H435R/Y436H modification as shown in SEQ ID NO 38. Figure 28 shows the amino acid sequences of human IgG2 constant region (Uniprot P01859) incorporating CH2 and CH3 domains (as defined in Uniprot) of the Fc fragment, showing: (a) full Uniprot sequence of the constant region of the wild-type human IgG2, as shown in SEQ ID NO 39; (b) wild-type human IgG2 Fc fragment incorporating CH2 and CH3 domains (as defined in Uniprot), as shown in SEQ ID NO 40;

(c) human IgG2 Fc fragment with M428S/N434S modification, as shown in SEQ ID NO 41;

(d) human IgG2 Fc fragment with M428S/N434S/Y436H, as shown in SEQ ID NO 42; and (e) human IgG2 Fc fragment with M428S/N434S/H435R/Y436H modification, as shown in SEQ ID NO 43.

Figure 29 shows the amino acid sequences of human IgG3 constant region (Uniprot P01860) incorporating CH2 and CH3 domains (as defined in Uniprot) of the Fc fragment, showing: (a) full Uniprot sequence of the constant region of the wild-type human IgG3, as shown in SEQ ID NO 44;

(b) wild-type human IgG3 Fc fragment incorporating CH2 and CH3 domains (as defined in Uniprot), as shown in SEQ ID NO 45;

(c) human IgG3 Fc fragment with M428S/N434S modification, as shown in SEQ ID NO 46; and

(d) human IgG3 Fc fragment with M428S/N434S/F436H, as shown in SEQ ID NO 47.

Figure 30 shows the amino acid sequences of human IgG4 constant region (Uniprot P01861) incorporating CH2 and CH3 domains (as defined in Uniprot) of the Fc fragment, showing: (a) full Uniprot sequence of the constant region of the wild-type human IgG4, as shown in SEQ ID NO 48;

(b) wild-type human IgG4Fc fragment incorporating CH2 and CH3 domains (as defined in Uniprot), as shown in SEQ ID NO 49;

(c) human IgG4 Fc fragment with M428S/N434S modification, as shown in SEQ ID NO 50; (d) human IgG4 Fc fragment with M428S/N434S/Y436H, as shown in SEQ ID NO 51 ; and

(e) human IgG4 Fc fragment with M428S/N434S/H435R/Y436H modification, as shown in SEQ ID NO 52.

Figure 31 shows the biolayer interferometry (Octet) data detects no interaction of PBS buffer only control to protein A coated tips (A) and protein G coated tips (B).

Figure 32 shows biolayer interferometry (Octet) data for the binding of the IgGl-Fc proteins at 1000 nM to protein A coated tips: (A) WT, (B) M428S/N434S/Y436H, and (C) H435R. Figure 33 shows biolayer interferometry (Octet) data for the binding of the IgGl-Fc proteins at xxx nM to protein G coated tips: (A) WT, (B) M428S/N434S/Y436H, and (C) H435R. Figure 34 shows biolayer interferometry (Octet) data for the binding of the IgG2-Fc proteins at 1000 nM to protein A coated tips: (A) WT, (B) M428S/N434S/Y436H, and (C) H435R.

Figure 35 shows biolayer interferometry (Octet) data for the binding of the IgG2-Fc proteins at 1000 nM to protein G coated tips: (A) WT, (B) M428S/N434S/Y436H, and (C) H435R.

Figure 36 shows biolayer interferometry (Octet) data for the binding of the IgG4-Fc proteins at 1000 nM to protein A coated tips: (A) WT, (B) M428S/N434S/Y436H, and (C) H435R.

Figure 37 shows biolayer interferometry (Octet) data for the binding of the IgG4-Fc proteins at 1000 nM to protein G coated tips: (A) WT, (B) M428S/N434S/Y436H, and (C) H435R. Figure 38 shows the analyses of the IgGl pA-/pG- (M428S/N434S/H435R/Y436H) homodimer. Figures (A) and (B) show SDS-PAGE analysis. (A) Supernatant of HEK cells transfected with the following IgGl-Fc variants. Lane 1 and 7: Molecular weight markers. Lane 2: M428S/N434S/H435R/Y436H homodimer. Lane 3: H435R homodimer. Lane 4: wild-type. Lane 5: M428S/N434S/Y436H homodimer. Lane 6: Control HEK supernatant without transfection. (B) Application of the M428S/N434S/H435R/Y436H homodimer HEK supernatant to protein A and protein G columns. Lanes 1, 5 and 10: Molecular weight markers. Lane 2: Cell supernatant of the M428S/N434S/H435R/Y436H IgGl Fc fragment. Lane 3: Protein A elution at 3.5. Lane 4: Protein A elution at pH2.5. Lane 5: Protein A flow through. Lane 7: Protein G elution at pH3.5. Lane 8: Protein G elution at pH2.5. Lane 9: Protein G flow through. (C) Biolayer interferometry analysis showing binding of wild-type and M428S/N434S/H435R/Y436H homodimer HEK supernatant to anti-Fc coated tips. lOx PBS was added to supernatant to a final IxPBS prior to loading. The control is supernatant from HEK cells only, without a transfection.

With reference to the drawings, there is provided a protein comprising a human IgG Fc variant fragment monomer capable of binding superantigen protein A but not superantigen protein G [pA+/pG-], comprising serine at position 428, serine at position 434 and optionally histidine at position 436, with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, such substitutions in the IgG Fc variant fragment monomer at positions 428, 434 and optionally 436 reduces or prevents the binding of protein G, enhancing the purification of immunoglobulins.

In the context of the present invention, an Fc variant fragment capable of binding protein A but not protein G is represented as [pA+/pG-], an Fc variant fragment capable of binding protein G but not protein A is represented as [pA-/pG+] and an Fc variant fragment capable of binding neither protein A nor protein G is represented as [pA-/pG-] .

With reference to the drawings, there is provided a heterodimeric protein comprising a heterodimeric human IgG Fc variant fragment, wherein the heterodimeric Fc variant fragment comprises: a first Fc variant fragment monomer, capable of binding superantigen protein A but not superantigen protein G [pA+/pG-], comprising serine at position 428, serine at position 434 and optionally histidine at position 436 or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24; and a second, Fc variant fragment monomer, capable of binding superantigen protein G but not superantigen protein A [pA-/pG+], comprising arginine at position 435 and optionally phenylalanine at position 436, with reference to human IgGl or comprising these residues at the corresponding positions in human IgG 2, 3, or 4 as shown in the alignment in Figure 24, wherein positions are defined according to the EU index of Kabat. Advantageously, such substitutions in the IgG Fc variant fragment monomer at positions 428, 434 and optionally 436 reduces or prevents the binding of protein G, enhancing the purification of heterodimeric immunoglobulins.

In an embodiment of the invention, the heterodimeric Fc has one Fc which binds pA, but has very reduced or no binding to pG [pA+/pG-], while its other Fc binds pG, but has reduced or no binding to pA [pA-/pG+]. In another approach, each heterodimer Fc has reduced or no binding to pA and pG [pA-/pG-] and purification can be achieved via differential affinity of different proteins fused to each Fc of the heterodimeric protein.

Preferably, the protein or the heterodimeric protein comprises a human IgG Fc variant fragment monomer capable of binding superantigen protein A but not superantigen protein G, comprising serine at position 428, serine at position 434 and histidine at position 436. Advantageously, such substitutions in the IgG Fc variant fragment enhance the purification of protein or heterodimeric protein.

In one example, the Fc arms of an IgGl were modified by the mutations shown in Figure 1 below to negate or weaken the binding to either pA or pG. These mutations allow a simple purification scheme to generate pure heterodimeric Fc, as shown in Figure 1.

Table 1

Table 1 illustrates mutations that alter the pA and pG binding properties of the IgG Fc fragment. M428S/N434S/Y436H (pG-) does not bind detectably to pG; M428S/N434S (pG-*) shows much reduced binding to pG, relative to wild type.

A purification scheme in accordance with the present invention enables purification of Fc heterodimer very cleanly by the sequential application to pA and pG resins (verified by mass spectrometer data), the properties of the purified heterodimeric Fc confer a slightly better half-life (when compared to wild type, i.e. WT) in the mice PK studies.

Biologically-active proteins, such as binding scaffolds and receptor domains can be fused to termini of the heterodimeric variant Fc of the invention.

Substitutions that eliminate the affinity for both pA and pG can be introduced into each heavy chain of the heterodimeric immunoglobulin, so that purification can be based on the affinity of proteins fused to each Fc sequence, rather than the binding properties of the Fc in the heteroduplex. In embodiments in which both variant Fc of the heterodimer are modified to reduce binding to both pA and pG, differential purification can be achieved based on the binding properties of the biologically-active fusion proteins at the termini of the variant Fc, e.g., differential affinity for pA, pG, protein L, kappa or lambda binding resins.

Variant Fc of the invention can be modulated to increase half-life and enable effector functions through the Fc domain.

Examples

The production and characterisation of heterodimeric IgGl-Fc variant proteins is described. The strategy for generating the heterodimeric Fc fragments is shown in Figure 1. Rational design and sequence alignment studies were used to identify sites that would lead to a weakening or abrogation of binding to protein G (pG) whilst having a minimal impact on the interaction with protein A (pA). Known substitutions that produce the reverse scenario, i.e. leading to a weakening or abrogation of binding to pA whilst having a minimal impact on the interaction with pG, were integral to the method and were also generated. The properties of these respective substitutions were first assessed in the Fc fragment, which is a homodimer of identical chains consisting of a hinge region joined to CH2 and CH3 domains of the IgGl heavy chain. Initial studies to characterise the rationally-designed mutants were carried out in Pichia pastoris. Those substitutions with the desired properties were used for production of the heterodimeric Fc fragments, made up of two chains that differ in their ability to bind pA and pG. To generate heterodimeric IgGl-Fc, two vectors each containing an Fc construct with either an altered pA or pG binding site were co-transfected and expressed in HEK293 mammalian cells. Although two vectors were used to express the two Fc chains, the heterodimeric Fc could be incorporated into the same vector in a bicistronic arrangement under the same or different promoters using standard molecular biology techniques. The IgGl-Fc heterodimer was purified from the supernatant by stepwise pA and pG affinity chromatography. The heterodimer was characterised both biophysically and pharmacokinetically. Strategies were also investigated to bias formation of the heterodimeric species. The properties of the Fc heterodimer were found to be similar to that of wild type Fc and summarised in Table 2.

* % yield in comparison to WT Fc following expression and purification

Table 2. A comparison of the properties of the heterodimer Fc format with WT Fc

The heterodimer (M428S/N434S/Y436H Fc paired with H435R Fc) had a melting temperature Tm (°C) Tml/Tm2 of 66 /81. The thermostability of Fc heterodimere of the present invention is affected to a lesser degree than many of the other technologies.

Figure 21 includes the use of Fabs arms with no binding to either pA or pG, or both, such as those from other isotypes (IgA, IgD, IgE) and other species, and lack of pG binding in many IgG2 (Perosa et al. (1997) Clin. Exp. Immunol. 109(2), 272-8). The figure could be made of two different IgA Fab arms on the heterodimeric IgGl-Fc. The Fabs can carry different VH that bind different epitopes on the same or different antigen. Note that if pA and pG based resins are used which are specific to the Fc site only then any isotype Fab arms (including IgG), or combinations of, can be used irrespective of their ability to bind pA or pG.

The scFv of Figure 22 could be oriented as a VL-VH fusion (as depicted) or VH-VL. For purification of such formats it is possible to substitute the use of a light chain affinity resin or Fab-specific resin for one or both of the pA or pG steps if particular light chain classes are present only on one of the Fc arms; a VKappal and a VLambda associated with each Fc arm would enable a stepwise purification using pL and VLambda affinity resins.

With reference to Figure 23, a common light chain bispecific mAb is shown where the distinct binding properties of the heavy chain (VHCH1) of the Fab arms enables a two-step purification. One VHCH1 of the heavy chain is pA+/pG- whereas the other has the opposite phenotype, pA-/pG+, e.g., a VH3 subclass with an IgA CHI generates a VHCH1 which is pA+/pG-, and a VH that is not of the VH3 subclass with an IgGl CHI forms a VHCH1 which is pA-/pG+. If scFv were substituted for the Fab arms to create a scFv-Fc fusion then a combination of pA and a light chain affinity resin could be employed, whereby a VH3 is used only on one chain to make it pA+ and different light chains, that differ in their ability to bind to a light chain affinity resin, are used on each chain, such as a VLambda and a VKappal.

Figures 27 - 30 show the amino acid sequences of IgGl (Figure 27), IgG2 (Figure 28), IgG3 (Figure 29) and IgG4 (Figure 30) incorporating CH2 and CH3 domains (as defined in Uniprot) of the Fc fragment, showing (a) full uniprot sequence of the constant region of the wild-type human IgG (b) Fc fragment with M428S/N434S modification, (c) Fc fragment with M428S/N434S/Y436H (d) Fc fragment with M428S/N434S/H435R/Y436H modification. Example 1: Substitution of residues to abrogate binding to protein G (pG) whilst retaining binding to protein A (pA) in an IgG-Fc fragment (pA+/pG-) and to abrogate binding to protein A (pA) whilst retaining binding to protein G (pG) in an IgG-Fc fragment (pA-/pG+). Substitution of residues to abrogate pA binding and pG binding were introduced into the wild type (WT) human IgGl-Fc sequence (Figure 25). A reference amino acid sequence for the Fc region used herein can be found in the Uniprot database with identifier P01857 and the nucleotide sequence in Genbank accession J00228 for the gene IGHGl. Standard molecular biology procedures were used for cloning. Mutagenesis was performed using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies); the primers are listed in Table 3.

Name

Primer Sequence 5' -> 3'

H435R-fw (SEQ ID NO: 1) GAGGCTCTGCACAACCGCTACACACAGAAGAGC

H435R-rev (SEQ I D NO: 2) G CTCTTCTGTGTGTAG CG GTTGTG CAG AG CCTC

H435R-Y436F-fw (SEQ I D NO: 3) GAGGCTCTGCACAACCGCTTCACACAGAAGAGCCTC

H435R-Y436F-rev (SEQ ID NO:4 ) GAGGCTCTTCTGTGTGAAGCGGTTGTGCAGAGCCTC

M428S-N434S-fw (SEQ I D NO: 5) CTTCTCATGCTCCGTGAGCCATGAGGCTCTGCACAGCCACTACACACAG

M428S-N434S-rev (SEQ ID NO:6 CTGTGTGTAGTGGCTGTGCAGAGCCTCATGGCTCACGGAGCATGAGAAG )

M428S-N434S- Y436H-fw CTTCTCATGCTCCGTGAGCCATGAGGCTCTGCACAGCCACCACACACAGAAGAG (SEQ ID NO: 7)

M428S-N434S- Y436H -rev CTCTTCTGTGTGTGGTGGCTGTGCAGAGCCTCATGGCTCACGGAGCATGAGAAG (SEQ ID NO: 8)

Table 3. Primers for introduction of substitutions into a wild type (WT) human IgGl-Fc sequence to abrogate pA binding or pG binding. The codon substitutions are underlined.

All primers used were synthesised by Sigma Aldrich. IgGl-Fc wild-type sequence, cloned into the EcoRl and Notl restriction sites in the expression vector pPICZaA (Invitrogen), was used as a template. A rational design approach was used to create an IgGl-Fc that shows weak or no binding to pG. Table 4 below (Sauer-Eriksson et al. (1995) (Structure 3: 265- 278)) lists the amino acids involved in the binding of human IgGl-Fc to pG and pA.

Fc Protein G (pG) Protein A (pA)

Residue Interaction Residue Interaction Residue Interaction

L251 MC K31 SC Q129 SC

M252 H K28 H F124 H

1253 MC E27 SC Q129 SC

1253 H W43 H F132 H

1253 H T44 H L153 H

S254 MC E27 SC

S254 SC E27 SC Q128 SC

Q311 SC E42 SC N137 SC

L314 H L136 H

E380 SC K28 SC

E382 SC K28 SC

M428 H K28 H

L432 MC Y133 SC

H433 SC N35 SC

N434 SC N35 SC N130 SC

N434 SC V39 MC

N434 SC W43 SC

H435 H Y133 H

H435 H L136 H Y436 MC N35 SC

Y436 H K28 H

Q.438 SC Q32 SC

Table 4. A modified version of Table 2 described by Sauer-Errikson et al. (1995). The residues of the Fc domain that interact with either pA or pG are listed. Main chain polar interactions (MC); other polar and charged interactions (SC); and hydrophobic interactions (H).

To aid design, naturally-occurring IgGs with the desired property of binding pA, but little or no binding to pG, were investigated; cat (feline) IgG was found to have these characteristics. An alignment of human IgGl-Fc with feline IgG-Fc is shown in Figure 2. With reference to the alignment of the human IgGl Fc region to feline IgG Fc (top) and human IgG3 (bottom) as shown in Figure 2, feline IgG binds weakly to pG and three positions (underlined) were chosen to be mutated in the human IgGl-Fc, M428S-N434S and M428S-N434S-Y436H ('cat triplet'). Human IgG3 is known for its weak pA binding properties. H435 was selected and histidine was exchanged to arginine, H435R (underlined).

From an analysis of the known structure of pG and the Fc fragment (Sauer-Eriksson et al. (1995); PDB entry 1FCC), the amino acids involved in pG binding were compared with sequence differences between human IgGl and feline IgG. Two combinations of mutations were chosen for further assessment: M428S-N434S-Y436H (referred to as 'cat-triplet') and M428S-N434S. Amino acid substitutions and vector construction were performed by standard molecular biology procedures. Each combination was introduced into an IgGl-Fc chain so as to generate an Fc fragment that binds to pA, but not to pG (pA+/pG-). Human IgG3 Fc does not bind to pA, but retains binding to pG (Reis et al. (1984) J. Immunol 132: 3098-3102); we denote this as pA-/pG+. The alignment of the Fc region of human IgG3 with IgGl is shown in Figure 2. From analysis of the structure of pA with IgG-Fc it was previously suggested that the H435R substitution in IgG3 is responsible for preventing pA binding (Deisenhofer (1981) Biochemistry 20, 2361-2370; PDB entry 1FC2), which was further corroborated by the findings from the structure of pG/IgG-Fc (Sauer-Eriksson, 1995) and by identification of positions which had previously been used to abolish the binding of IgGl-Fc to pA (Jendeberg et al., (1997) J Immunol Methods, 201: 25-34). The substitutions H435R and H435R/Y436F were introduced into the Fc region in this study to generate pA- /pG+ mutants.

The properties of the substitutions described above were each first assessed in the Fc fragment, which is a homodimer of identical chains consisting of a hinge region joined to CH2 and CH3 domains of the IgGl heavy chain. Example 2: Expression and purification of pA+/pG- and pA-/pG+ IgGl Fc Fragments pPICZaA is an expression vector used for expressing and secreting recombinant proteins from Pichia pastoris. The AOX1 promoter can lead to a high level of methanol-induced expression of the gene of interest. Expression was carried out in the Pichia pastoris strain X33 according to the manufacturer's instructions (Invitrogen). In brief, a 7 days' expression was conducted. lOmL pre-cultures of selected production clones were incubated for 24 hours at 28 °C and 180rpm. After 24h the pre-culture were harvested, transferred into lL-baffled- shaking flasks, mixed with YPG medium to 500ml and incubated for a further 24 hours at 28 °C and 180rpm. After 24h the cultures were centrifuged (3000g, lOmin) and YPG was discarded. The cell pellets were resuspended to 500mL with YPM medium. The cultures were incubated at 25 °C with shaking at 180rpm for 4 days. 5mL methanol was added each day. The final culture was centrifuged at 5000g for lOmin. 1M NaP04-buffer (pH7) was then added in a 1: 10 v/v ratio and the solution centrifuged again at a higher speed of 8000g for 45min. Finally the supernatant was filtered (pore size 0.45μιη).

The IgGl-Fc proteins were purified from the supernatant using either a pA or a pG affinity column, depending on their mutations. A HiTrap Protein A HP 5mL column and HiTrap Protein G HP lmL column was used on an AKTA purifier (GE Healthcare). The filtered supernatant was applied with a flow rate of 4mL/min on the 5mL column and with a flow rate of lmL/min on the lmL column. To ensure that all of the supernatant had passed through the column, the system was washed with lx PBS again until UV signal maintained a constant level. Bound protein was eluted with glycine buffer (0.1M, pH3.5) and neutralised to pH7.4 with Tris-buffer (pH12). The protein sample was dialysed against lxPBS. A large-scale expression and purification was carried out using the four constructs as described. The M428S-N434S-Y436H ('cat-triplet') and M428S-N434S IgG Fc constructs were purified using pA affinity resin. The 'cat-triplet' was named after its feline source. The H435R and H435R/Y436F IgG Fc constructs were purified using pG affinity resin. Table 5 gives a representative example of the amounts of Fc fragment that were acid eluted (glycine buffer, 0.1M, pH3.5) from the different affinity resins. The yields are similar to those seen for wild-type (WT) IgGl-Fc fragment at approximately 20mg/ 0.5L culture using Pichia pastoris.

Table 5. A representative example of the amounts of Fc fragment that were acid eluted from the pA and pG affinity resins.

Example 3: Characterization of pA+/pG- and pA-/pG+ IgGl Fc Fragments

Testing the binding and non-binding of each Fc construct to pG and to pA was carried out by Bio-layer Interferometry. A protein stability test was performed with differential scanning calorimetry and the effect of the mutations on the secondary structure was tested using circular dichroism. Size-exclusion chromatography was also carried out. 3.1 Bio-layer Interferometry

The binding kinetics were investigated to determine the extent to which the mutations were effective in binding to pG, but not to pA, and vice versa. Testing the binding and non- binding of each Fc construct to pG and to pA was carried out by Bio-layer Interferometry with the forteBIO Octet system and capture tips. For analysing the binding kinetics of the Fc mutants, a 1:2 dilution series with lxPBS was made starting with a concentration of ΙΟΟΟηΜ Fc protein. Before and after every sample dilution measurement, a run with PBS was conducted providing a base line before measuring the IgGl-Fc sample and achieving a dissociation of the sample afterwards. Measurements comprised an association of the protein (600sec, different IgGl-Fc sample dilutions) and the dissociation (600sec, lx PBS).

Each profile shows a dilution series of 1:2 with running buffer (Figure 3). All mutants were analysed for binding to pA and pG. The binding kinetics of the Fc fragment with M428S- N434S-Y436H and M428S-N434S mutations are displayed at the bottom of Figure 3. A qualitative estimation showed good binding properties to pA, whereas there was no binding of the M428S-N434S-Y436H mutant to pG. The M428S-N434S mutant showed weak binding characteristics to pG and a very fast off-rate. The mutation combination cat-triplet (M428S-N434S-Y436H) showed the desired property of binding to pA and no detectable binding to pG (pA+/pG-).

A qualitative estimation showed as expected that the mutants H435R and H435R-Y436F did not bind to pA, whereas binding to pG was only slightly reduced compared to human WT IgGl-Fc. The signal intensity decreased with decreasing protein concentration for the mutants (as for the wild-type) for binding to pG. The specific knockout of the pA binding site was successful for both mutants.

In summary, both approaches led to the generation of homodimeric IgGl-Fc variants M428S- N434S and M428S-N434S-Y436H, with weak and no detectable binding to pG, respectively, that bound to pA (pA+/pG-) and to generation of IgGl-Fc variants H435R and H435R- Y436F with no detectable binding to pA, but that bound to pG (pA-/pG+).

3.2 Differential Scanning Calorimetry Differential scanning calorimetry (DSC) was used to investigate the stability and unfolding process of the constructs. Several hydrogen bonds, disulphide bridges and hydrophobic interactions maintain native protein structure and stability. Slowly increasing the temperature led first to partial denaturation of domains until eventually the whole protein was denatured. This process resulted in a thermogram displaying a heat capacity profile of the protein and its domains.

For running DSC, protein solutions were diluted to an OD280 of 0.35, which corresponded to a concentration of 5μΜ. Dilutions were made with 1 x PBS. Differential scanning calorimetry was carried out using a MieroCal VP-DSC differential scanning calorimeter (GE Healthcare). The heating-rate was 60 °C/h with a temperature scale ranging from 20 °C to 110 °C. After the first run a rescan was conducted. The data obtained was interpreted using the Software OriginLab and a non-2-state fit applied.

As shown in Figure 4, the T _m values of all mutants varied and differed from the wild type. Three peaks were observable which were typical for a human IgGl-Fc fragment expressed in Pichia pastoris. The IgGl-Fc weak pG binder M428S-N434S and non-pG binder M428S- N434S-Y436H showed decreased T _m values. Therefore the denaturation process started approximately 7 °C earlier than that of the WT. T _m 2 and T _m 3 values were closer to the WT values, which indicated less impact of the mutations on the CH3 domain of the protein. T _m values of the IgGl-Fc weak pA binder H435R and H435R-Y436F were more similar to the T _m values of the WT Fc. The protein denaturation process started only approximately 4.5 °C earlier than the WT. T _m 3 even showed a slight increase. Traxlmayr et al. (2012) (Biochim. Biophys. Acta 1824: 542-549) studied various scaffold- stabilizing mutations in the Fc and the DSC values were comparable with the values of the homodimeric variant constructs described herein. Changes in temperature stability were expected due to the mutations introduced into the Fc fragment. The largest shift was in T _m l, which represented the CH2 domain of the Fc fragment, although all mutations are located in the CH3 domain. The two domains are in contact with each other and this result highlighted structural "cross talk" (i.e- interaction) between the two domains.

3.3 Circular Dichromism Circular Dichromism (CD) was carried out to analyse the secondary structure of the constructs in comparison to WT IgG Fc. Circular Dichromism (CD) was carried out on a Chirascan (Applied Photophysics). Protein samples were diluted with 1 x PBS to an OD ₂ 8o of 0.8. CD spectra (205-280nm) were recorded using a quartz cuvette with a path length of 1mm. Each sample was scanned 3 times. The spectral bandwidth was 0.5 nm, the step size was 1 nm with a time per point of 3 sec.

In Figure 5A, CD spectra of H435R and H435R-Y435F variants are shown. In Figure 5B, CD spectra of M428S-N434S-Y436H and M428S-N434S variants were compared to wild type IgGl Fc. The CD spectrum of wild type IgGl-Fc showed two minima at 218nm and 229nm, which are characteristic for the beta-barrel structure of the immunoglobulin fold. The IgGl- Fc weak pA binders, H435R and H435RY436F, showed a shift in the minima at 218nm and 229nm which indicated small differences in concentration but the overall spectra were comparable to the shape of the Fc wild-type protein. With the IgGl-Fc weak or non-pG binding constructs the non-pG binding M428S-N434S-Y436H ('cat-triplet') showed a small shift in the ellipticity at the minima 218nm, whereas at 229nm both variants showed this small deviation. In summary, the CD spectra of the variant constructs are comparable with that of the wild-type protein. 3.4 Size-exclusion Chromatography

For comparing the size and shape of the IgGl-Fc homodimer mutants to WT IgGl-Fc, size- exclusion chromatography was conducted. Size exclusion chromatography (SEC) was performed on a Shimadzu LC20 HPLC system. Protein of wild type and variants was applied to a Superdex 200 lOx 300mm SEC Column (GE Healthcare) at a flow rate of 0.75mL/min in lxPBS supplemented with 200mM NaCl as running buffer. Elution was monitored using a UV/VIS photodiode array detector and detecting the absorption at 280nm.

As shown in Figure 6, the Fc constructs showed almost identical retention times to those of the WT IgGl-Fc. The wild-type Fc elutes as a single peak at a retention time of approximately 20min. The small peak seen for the M428S-N434S-Y436H and H435R- Y436F located close to the main peak implied that there was a low percentage of multimeric species. Reapplication to SEC of the fractions from the main peak resulted in a single peak showing that these multimeric species could be removed. SEC, CD and DSC protein characterization indicated only small differences in the properties of each construct compared to WT IgGl-Fc. The decision to select the H435R and M428S- N434S-Y436H constructs for validating the heterodimerisation approach was based on the results of Bio-layer Interferometry. The Fc with H435R mutation was used for abrogating binding to pA. As M428S-N434S showed a slight interaction with pG, the M428S-N434S- Y436H mutant was selected for heterodimerisation studies.

Example 4: Transient expression in HEK293-E cells, two-step purification and characterisation of the heterodimeric IgGl-Fc (M428S-N434S-Y436H/H435R).

4.1 Transient Expression

Transient expression was carried out using the mammalian expression vector pTT5, which contains a CMV promoter (NRC, Canada). A pTT5 vector containing the WT IgG-Fc sequence with an EcoNl restriction site in the C-terminal region of the CH2 domain was used for sub-cloning (Figure 26C and D). The EcoNl site was introduced using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies) with the following primers: pTT5 EcoN lfw (SEQ I D NO: 9)

5'-GGTGAGCGTGCTGACCGTCCTGCACCAGGACTGGCTGAATGGCAAGGAGTACAAGTG C-3' pTT5 EcoN lrev (SEQ I D NO: 10)

5'-GCACTTGTACTCCTTGCCATTCAGCCAGTCCTGGTGCAGGACGGTCAGCACGCTCAC C-3'

The pPICZaA constructs detailed in Example 1, which contain an EcoNl site, were used as a DNA template to release the EcoNl/Sapl insert with the desired mutations. In this way the M428S-N434S-Y436H variant IgFc sequence was sub-cloned into EcoNl and Sapl of the pTT5 WT IgG-Fc vector. The insert containing the H435R mutation was also separately cloned into the pTT5 WT IgG-Fc vector following this method. The two pTT5 vector constructs were then co-transfected using PEI (polyethylenimine) into HEK2936E cells (NRC, Canada) and the cells were cultured following standard procedures. The supernatant was harvested after 120h and filtered (pore size 0.45μιη). Although two vectors were used to express the two Fc chains which make up the heterodimeric Fc, using standard molecular biology techniques both Fc chains can be incorporated into the same vector in a bicistronic arrangement under the same or different promoters.

Additional constructs were also made to see if the formation of heterodimers in this co- expression system could be biased. The 'Knobs into Hole' (KiH) mutations, which were previously shown to generate heterodimeric Fc fragments by Ridgway et al., (1996) (Protein Engineering, 9: 617-621), were incorporated into the above constructs. A disulphide bridge was also introduced into these KiH mutations to see if this would lead to improved stability. The primers used for site-directed mutagenesis, using the QuikChange Kit and standard molecular biology techniques, can be found in Table 6. The T366Y and Y407T substitutions on separate Fc chains create a knob-into-hole format. Y349C and S354C were changed on separate Fc chains to enable creation of an inter-chain disulphide bridge.

Primer Name Primer Sequence 5'→ 3

T366Y (SEQ ID CAAGAACCAGGTCAGCCTGTATTGCCTGGTCAAAGGCTTCT NO: 11 )

T366Y-rev (SEQ AGAAGCCTTTGACCAGGCAATACAGGCTGACCTGGTTCTTG ID NO: 12)

Y407T (SEQ ID CGGCTCCTTCTTCCTCACCAGCAAGCTCACCGTG

NO: 13)

Y407T-rev (SEQ CACGGTGAGCTTGCTGGTGAGGAAGAAGGAGCCG

ID NO: 14)

Y349C (SEQ ID CGAGAACCACAGGTGTGCACCCTGCCCCCATCCCGG

NO: 15)

Y349C-rev (SEQ CCGGGATGGGGGCAGGGTGCACACCTGTGGTTCTCG

ID NO: 16)

S354C (SEQ ID ACCCTGCCCCCATGCCGGGATGAGCTGACCAAGAAC

NO: 17)

S354C-rev (SEQ GTTCTTGGTCAGCTCATCCCGGCATGGGGGCAGGGT

ID NO: 18) Table 6. Primers used to introduce KiH mutations and interdisulphide bonds into IgG-Fc. The codon substitutions are underlined. The IgGl-Fc construct containing the M428S-N434S-Y436H mutations was cloned into the mammalian expression vector pTT5. The IgGl-Fc construct H435R was cloned into a separate pTT5 vector. The two pTT5 vector constructs were then co-transfected into HEK293-E6 cells, as described above. Additional constructs were also made to see if the formation of heterodimers in this co- expression system could be biased. The theoretical maximum amount is 50% heterodimer and 25% for each of the homodimers for this purification based strategy. The 'Knobs into Hole' (KiH) mutations, which were previously shown to generate heterodimeric Fc fragments (Ridgway et al. 1996), were incorporated into the construct. In another version, a cysteine was also included into these KiH mutations to enable formation of a disulphide bridge and to assess if this would lead to improved stability.

Table 7. Heterodimeric constructs with KiH mutations

The construction of these clones is described above. The HEK293E-expressed and purified IgGl-Fc heterodimer constructs were characterised by: DSC, binding to pA, pG and FcRn, SEC and half-life studies in mouse. 4.2 Purification

Purification was carried out essentially as described previously (Example 2), but in this case involved two steps; the supernatant was first applied to the pA column (HiTrap Protein A HP 5mL column) and the bound protein subsequently eluted and applied to the pG column (HiTrap Protein G HP lmL column), both using an AKTA purifier (GE Healthcare). The filtered supernatant was applied with a flow rate of 4mL/min on the 5mL column and with a flow rate of lmL/min on the lmL column. After the first purification of the HEK293-E6 supernatant with a pA column, the protein was eluted in a single step with ΙΟΟηΜ glycine buffer pH3.5, as a single symmetrical peak. After neutralisation and dialysis against lxPBS, this sample was then applied to the pG column and thus the second purification step was performed using the pG column. The conditions for elution were identical, bound sample was eluted with ΙΟΟηΜ glycine buffer pH3.5. Fraction concentrations were measured and those with a protein concentration over 0.2mg/mL were pooled and dialyzed against lxPBS. After dialysis, heterodimers were characterized. The pure IgGl-Fc heterodimer, having one Fc chain carrying the M428S-N434S-Y436H mutations (pA+/pG-) and the other the H435R mutation (pA-/pG+), was used for characterisation. In a separate second purification run the sequence of the columns was reversed, the supernatant was first applied to the pG column (HiTrap Protein G HP 5mL column) and the bound protein subsequently eluted and applied to the pA column (HiTrap Protein A HP lmL column). The procedure was identical to that described above except for the change in the sequence of columns used. An SDS PAGE analysis of the steps during this procedure is shown in Figure 7. The fraction that first bound to the pG column and then to the pA column, as identified in Lane 6, is the pure IgGl-Fc heterodimer.

4.3 Characterisation by DSC, Bio-layer Interferometry, SPR and SEC

DSC, Bio-layer Interferometry and SEC were carried out as described in the examples above. The DSC profile of the purified IgGl-Fc heterodimer gave a result close to wild type Fc (see Figure 8 and Table 8). Two sharp peaks were observable, these were also seen in the in the wild-type spectrum (A). T _m l of the IgGl-Fc heterodimer was 65.9 °C, about 5 °C below the Tm of the wild-type protein. T _m 2 was lower by only 2 °C compared to the wild-type protein, which indicated that the amino acid substitutions had only a small effect on the stability of the CH3 domain. This showed that the thermal stability of the IgGl-Fc heterodimer decreased only slightly as a result of the mutation that abrogated pA binding on one chain and mutations that abrogated pG binding on the other chain.

Table 8. T _m l °C and T _m 2 °C peaks identified by DSC

As shown in Figure 8C-E, heterodimers with additional 'Knob-into-Hole' (KiH) mutations demonstrated a marked decrease in the T _m (Fig 8C; Heterol-KiH). The addition of the disulphide bridge across the interface of the two CH3 domains in this construct led to a modest improvement (Figure 8D and E; Hetero2-KiH and Hetero3-KiH, respectively). The 'Hetero2-KiH' Fc heterodimer ('cat-triplet' M428S-N434S-Y436H, Y407T, Y349C on one Fc arm and H435R, T366Y, S354C on the opposing Fc arm) gave the best Tm values for the constructs containing the KiH mutations.

The binding affinity of the purified IgGl-Fc heterodimer to pA and pG was analysed by Bio- layer Interferometery, as previously described in Section 3.1, (Figure 9). The IgGl-Fc heterodimer bound to pA and pG with KD values greater than those for WT IgGl-Fc. This was due to the two protein A/G binding sites found in the Fc homodimer being reduced to only one pA and one pG binding site in the Fc heterodimer. The IgGl-Fc heterodimer, WT and homodimer versions showed binding to the anti-Fc positive control (Figure 10).

The human FcRn binding properties of the heterodimers were tested by Surface Plasmon Resonance using a Biacore 3000 (GE Healthcare). Samples were prepared and run in HBS-P buffer. Human FcRn (R&D System) was immobilised on a CM5 chip at a coating density of 200 RU. Varying concentrations of the Fcab were injected onto the chip to determine the affinity, ranging from 4000nM to InM. The binding interaction was measured at pH6.0 at a flow rate of 20μ1/ηιίη and checked for dissociation at pH7.4. The KD values were calculated with a 1: 1 fit or steady-state affinity, where applicable, using BIAevaluation 3.2 software. The human FcRn binding properties of the heterodimer tested with Surface Plasmon Resonance are shown in Figure 11A. The binding curve of the IgGl-Fc heterodimer differed only slightly from the wild-type protein (Figure 1 IB).

Table 9. Estimated KD and KA values of the respective Fc fragments.

The K _D value for the IgGl-Fc heterodimer (142nM) was slightly higher than that of the wild- type (55nM), showing that there is some impact on the FcRn binding which overlaps the pA and pG binding sites. The Heterol-KiH Fc fragment (IgGl-Fc heterodimer with Knob-into- Hole mutations) was also tested. All of the Fc fragments showed an equivalent pH dependency of association and dissociation to FcRn. All Fc fragments dissociated rapidly at pH7.4.

Size-exclusion chromatography was carried out as previously described to determine the size and solution properties of the IgGl-Fc heterodimer. As shown in Figure 12A, the heterodimer had a major single peak running at a retention time equivalent to wild type of approximately 20min. The KiH variants of the heterodimer were also analysed, of these Hetero2-KiH gave a favourable profile (Figure 12B). Hetero2-KiH also had an improved DSC profile over the others (Figure 8D) suggesting that it should be used preferentially if a KiH mutation is introduced to bias heterodimer formation.

Pharmacokinetic studies Pharmacokinetic studies were carried out on the IgGl-Fc heterodimer to determine the serum half-life in mouse. During the study clinical signs were monitored. Animals were inspected for any evidence of reaction to the treatment or ill health. No deviations from normal behaviour were recorded, the mice were healthy during the entire study. The serum proteins obtained were then tested in an ELISA. An enzyme-linked immunosorbent assay (ELISA) was conducted with the protein harvested from the murine mouse model. ELISA raw data was analysed using PKSolver (Zhang et al. (2010) Computer Methods and Programs in Biomedicine 99, 306-314).

The pharmacokinetics studies were carried out by the Department of Toxicology, Slovak Medical University study No. 20140505. Male mice were used, body weight 20.1 - 23.3g at the start of the treatment. 12 mice were used per test protein. Test proteins were applied at a concentration of lOmg/kg body weight (BW). The tested substance was administered intravenously into the tail vein. The volume was adjusted individually and ranged from 210- 233μί of the appropriate stock solution (lmg/mL in PBS) per mouse based on mouse body weight. Blood collection time points post-administration were 15, 30 minutes; 1, 8, 24 hours; 3, 7, 10, 14, 18, 21 and 28 days, resulting in 12 time points. Three mice (No. 1-3) out of 12 in each group were bled from sinus orbitalis at the first time point (1. TP = 15min), a second group of three mice (No. 4-6) at the second TP (30min) and so on. Mice bled for the 1TP were bled again at 5TP (= 24h) and so on. Each subgroup of mice (3/12) underwent three blood collections in total. Blood amounts collected ranged from 80-140μΕ. Serum was prepared after storage of blood samples for lh at room temperature and subsequent centrifugation of the samples in a bench centrifuge (4 °C 500g, lOmin) twice. Serum samples were stored at -18 °C before further procedures. Samples were analysed by enzyme-linked immunosorbent assay (ELISA). The ELISA used a mouse monoclonal antibody specific to the CH2 domain of human IgGl-Fc (Clone 8A4, AbD Serotec) to capture the human IgGl-Fc constructs. 96-well plates were coated with mouse IgG 8A4, lOOng/well in 50μί and incubated at 4 °C overnight. Washing of plates 4-times with 1 x PBS, subsequent 250μί blocking buffer (3% non-fat dry milk in PBS) was added and the plates were incubated at 37 °C for lh. Purified IgGl-Fc Heterodimer and IgGl-Fc wild type were diluted with 1 x PBS, supplemented with mouse serum (Sigma Aldrich). A standard curve was obtained (PBS + 10% mouse serum). The well plates were then washed four times with PBST and ΙΟΟμί secondary antibody, goat anti-human IgG (Fc specific) peroxidase (Sigma Aldrich SAB3701270) used at 1: 1.000, which was added to blocking buffer. Plates were incubated at RT °C for 2h and subsequently washed four times with PBS. ΙΟΟμί TMB substrate (Sigma Aldrich) was added and incubated for lOmin before stopping the reaction with 50μί 30% H2S04. The plates were read in a 96-well-plate reader at a wavelength of 405nm. ELISA raw data was analysed using PKsolver and Sigmaplot.

The clearance curves of the IgGl-Fc heterodimer and the IgGl-Fc wild-type are shown in Figure 13A. Half-life values for the a-phase and β-phase are displayed in Table 10.

Table 10. Half-life values for the a-phase and β-phase

Results were comparable to the values for the half -life of the WT IgG-Fc fragment and Fc fragments with mutations at H435 in IgGl (Kim et al. (1999) J. Immunol. 29: 2819-2825; Wozniak-Knopp et al. (2010) Protein Eng. Des. Sel. 23: 289-297). The results indicated that the binding to FcRn is preserved. Compared to the wild type, only minimal differences were observed. Stability Studies

The heterodimer and wild type were purified by size exclusion chromatography, as described previously. The Fc fragments were then incubated at different temperatures for 1 and 7 days at a concentration of lmg/ml in PBS before being reanalysed by size exclusion chromatography (Figure 14). The results showed the heterodimer to behave similarly to the wild type protein and that prolonged incubation had a minimal effect.

Example 5: Generation of a heterodimeric HER2-binding Fcab

The substitutions enabling purification of the heterodimeric Fc fragment were tested in the previously reported Fcab format, an Fc fragment with a binding site engineered into framework loops of the CH3 domain (Wozniak-Knopp et al. (2010) Protein Eng. Des. Sel. 23: 289-297). The Fcab H10-03-6, which binds HER2/neu with a high affinity, was used in the studies described herein. Fcab H10-03-6 was produced by randomization of amino acids in the AB and EF loops of the CH3 domain, which conferred binding with high affinity to HER2/neu. The mutations for establishing a heterodimer, the pA-/pG+ (H435R) or pA+/pG- (M428S, N434S and Y436H; 'cat triplet') variants, were inserted into Fcab H10-03-6 by site-directed mutagenesis using PCR. The EcoNl primer was used in combination with either the H435R primer or 'cat triplet' primer for PCR using the HI 0-03-6 DNA sequence as template.

The primers are shown in Table 11.

Table 11. Primers used to abrogate pA binding and pG binding, and to introduce an EcoNl site into the Fcab H10-03-6 sequence. The codon substitutions are underlined for H435R and "cat- triplet" primers. The EcoNl site is underlined in the EcoNl primer.

The PCR product was digested with EcoNl and Sapl and inserted into pTT5 carrying the WT IgG Fc vector, as described above. Each Fc chain was inserted individually into a pTT5 expression vector to create H435R or cat triplet Fcab H10-03-6 pTT5 constructs. Transfections into HEK2936E were carried out with different combinations of these and WT Fc chains (Table 12).

Table 12. Substitutions made in Fcab H10-03-6 to establish pA-/pG+ and pA+/pG- heterodimers.

One heterodimer was expressed from pTT5 vectors that contained the HI 0-03 -6 antigen binding sites on each Fc template; whilst two heterodimer constructs contained the H10-03-6 antigen binding site on one chain and the wild type sequence on the other.

Protein expression and purification of heterodimers was carried out as described in the examples above. Methods as described in the examples above were carried out to characterise the proteins with the exception of binding studies to HER2 by surface plasmon resonance using a Biacore 3000.

Size-exclusion chromatography was carried to comparing the size and purity of the IgGl-Fc mutants to IgGl-Fc wild-type heterodimer and the parental H10-03-6 Fcab. 50μg of wild- type and mutants were applied onto the Superdex 200 lOx 300mm SEC Column (GE Healthcare). As shown in Figure 15A, the heterodimeric constructs that have one H10-03-6 Fc binding arm and one wild type Fc arm have a single symmetrical peak with a slightly increased retention time in comparison to the wild type Fc heterodimer. This profile is a significant improvement to those constructs that contain the H10-03-6 sequence in both Fc arms, namely the parental Fcab 10-03-6 and the heterodimeric version, which elute much later as a broader peak. Studies have previously been carried out on Fcab H10-03-6 to improve its properties and SEC profile (Traxlmayr et al., (2012) Protein Eng. Des. Sel. 26: 255-265). Here, pairing up of one Fcab binding arm with a WT Fc arm was shown to improve the SEC profile. The pairing of a single Fcab binding Fc arm with a WT Fc arm also created one binding site per Fcab molecule, which may be beneficial in targeting some antigens. Likewise two Fc chains from two different Fcab binders can be paired by this method to create a bispecific heterodimeric Fcab.

The purified Fcab H10-03-6 heterodimers were characterised by binding studies to HER2 via surface plasmon resonance. Samples were prepared and run in HBS-P buffer. Soluble monomeric human HER2 (Bender Medsystems, Austria) was immobilised on a CM5 chip at a coating density of 1000 RU. Varying concentrations of the Fcab were injected onto the chip to determine the affinity, ranging from 2000nM to InM. A flow rate of 50μ1/ηιίη was used for binding to HER2. The KD values were calculated with a 1: 1 fit or steady state affinity, were applicable, using BIAevaluation 3.2 software (BIAcore).

The K _D values are given in Table 13. Figure 15B shows the binding curve profiles from a dilution series for the WT/H10-03-6 heterodimer. The heterodimer fragments made up of a WT and HI 0-03 -6 arm show only a small reduction in affinity to HER2, with a slightly faster off-rate than the constructs made up of two H10-03-6 arms with two possible antigen binding sites. In Figure 15C the binding curves of the heterodimeric constructs are shown in comparison to the parental Fcab H10-03-6 and Herceptin. The K _D values are consistent with those for the parental H10-03-6 Fcab described by Wozniak-Knopp et al., (2010).

Table 13. K _D values for the purified Fcab H10-03-6 and Fcab H10-03-6 heterodimers Example 6 Immunogenicity prediction of pA+/pG- IgGl Fc variants

Prediction of immunogenicity was performed using the publicly available NetMHCIIpan-3.1 server provided by the Center for Biological Sequence (CBS) Analysis from the Technical University of Denmark (Andreatta, M et al. (2015) Immunogenetics 67, 641-650). The following subset of fairly common and distinct DRB 1 alleles, which are present in over 90% of the global population, were used for the analysis: DRB 1_ 0101, DRB 1_0301, DRB 1_ 0401, DRB 1_ 0701, DRB 1_ 0801, DRB 1_ 1101, DRB 1_ 1301, DRB 1_ 1501. As mutations were incorporated at positions 428, 434, 435 and 436 (underlined below) analysis was carried out on the stretch of sequence from position 419 - 445:

419 QGNVFSCSVMHEALHNHYTQKSLSLSP 445 The above sequence is shown in SEQ ID NO 53.

In addition, as a control the middle nine residue sequence from hemagglutinun (HA) peptide of the influenza A virus H3N2 (underlined) was substituted for residues 428-436: 419 QGN VFSC S V Y VKQNTLKLTQ KS LS LS P 445

The above sequence is shown in SEQ ID NO 54.

This peptide is known to bind to many DR alleles (Hennecke J et al. (2000) EMBO J 19, 5611-5624).

The peptide length was set at 10 residues for the analysis. For each Fc variant sequential 10 residue overlapping peptides are generated and their predicted binding affinity calculated (Table 14A). Binding level is given as strong binding (SB) if the % rank or predicted affinity is below 0.5% and 50nM, respectively. The peptide is a weak binder (WB) if the % rank is above 0.5% and below or equal to 2% or the predicted affinity is above 50nM and below ΙΟΟΟηΜ. The % Rank is the predicted affinity ranked as a percentage from a set of 200,000 random natural peptides. Self-peptides corresponding to human germline amino sequences are omitted from the results with the exception of one peptide shown for the WT sequence in Table 14A.

The data showed there to be minimal immunogenic potential for the mutations incorporated in the Fc variants. The one peptide region shown to have a predicted weak binding is also identified in the WT sequence which is included in Table 14A. The pA+/pG- mutations M428S/N434S/Y436H bound to only one DRB l allele and gave the weakest predicted binding compared to the M428S/N434S variant or other mutations such as M428G/N434A (e.g. Glenmark, WO2014/049003). No binders were detected, above the set threshold, for the pA-/pG- mutations M428S/N434S/H435R/Y436H. In contrast strong and weak binders were predicted for the control WT/HA variant. Only the first three results generated for WT/HA are shown in Table 14A. In total the WT/HA variant had 3 strong binders and 29 weak binders present throughout all the DRB l alleles tested. A summary of the number of predicted binders for the variants is shown in Table 14B.

Table 14A. Predicted immunogenicity for a subset of common DRBl alleles as determined using the NetMHCIIpan 3.1 server for binding of peptides to human MHC class II alleles

Fc Variant Peptide Allele Predicted % Binding

(DRB1J Affinity Rank Level

(nM)

M428S/N434S SHYTQKSLSL 0101 235 3 WB

(SEQ ID NO 55) 0701 657 3 WB

M428S/N434S/Y436H HHTQKSLSLS(SEQ 0101 901 12 WB

ID NO 56)

M428S/N434S/H435R/Y436H - - - - -

M428G/N434A AHYTQKSLSL 0101 231 2.5 WB

(SEQ ID NO 57) 0701 639 2.5 WB

0801 998 12 WB

WT NHYTQKSLSL 0101 247 3 WB

(SEQ ID NO 58) 0701 669 3 WB

WT/HA VYVKRNTLKL 0101 74 0.5 SB

(SEQ ID NO 59)

YVKRNTLKLT 0101 148 1.4 WB

(SEQ ID NO 60)

LKLTQKSLSL 0101 277 3.5 WB

(SEQ ID NO 61) Table 14B. Comparison of the number of predicted MHCII binders to the Fc variants

Example 7 Analysis of the pA-/pG+ mutation (H435R) and the pA+/pG- mutations (M428S/N434S/Y436H) in different IgG Subclasses

Substitution of residues to abrogate pA binding and pG binding were introduced into the wild type (WT) IgG2-Fc and IgG4-Fc fragments (Figure 24). The IgGl-Fc changes were constructed as described earlier in Example 4. Wild type IgG3-Fc does not bind pA due to the presence of the R435; addition of the pG- mutations to an Fc carrying this change is detailed in Example 8. The wild type sequence of the IgG2-Fc and IgG4-Fc was cloned in the pTT5 vector system (SEQ ID NO 62 and 63, respectively). The Fc DNA sequences were synthesised with a leader sequence and EcoRl and BamHl sites to allow cloning into the pTT5 vector. Standard molecular biology procedures were used for cloning. Mutagenesis was performed using the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies) using primers carrying the H435R and M428S/N434S/Y436H substitutions as follows: H435R Forward (SEQ ID NO 64):

GAGGCTCTGCACAACCGCTACACACAGAAGAGCCTCTCCCTGTCTC

H435R Reverse (SEQ ID NO 65):

GAGACAGGGAGAGGCTCTTCTGTGTGTAGCGGTTGTGCAGAGCCTC M428S/N434S/Y436H Forward (SEQ ID NO 66):

TTCTCATGCTCCGTGAGCCATGAGGCTCTGCACAGCCACCACACACAGAAGAGC CTCTCCCTGTCTC M428S/N434S/Y436H Reverse (SEQ ID NO 67):

GAGACAGGGAGAGGCTCTTCTGTGTGTGGTGGCTGTGCAGAGCCTCATGGCTCA CGGAGCATGAGAA

Transient transfection, purification and analysis were carried out according to the methodology described in earlier sections for IgGl-Fc. H435R IgG-Fc (pA-) fragments were purified using a pG affinity column, whereas the M428S/N434S/Y436H IgG-Fc (pG-) fragments were purified using a pA affinity column. The purified IgG-Fc variants were analysed by biolayer interferometry (Octet) for the binding to protein A and protein G coated tips (Figures 31-37) and are summarised in Table 15.

* WT lgG3 has Arginine (R) at position 435 and has the pA-/pG+ phenotype.

# These \gG3 equivalent changes (M428S/N434S/H435R/Y436H) are pA-/pG- in the IgGl setting (see Example 8).

Table 15. Summary of the protein A (pA) and protein G (pG) phenotypes of the various IgG residue changes as determined by bio-layer interferometry.

DSC was carried out as described previously and the data is summarised in Table 16.

* Tm values are in °C.

# The first unfolding event identified two Tm values. Table 16. Summary of the DSC data of the IgG-Fc subclasses carrying either the H435R or M428S/N434S/Y436H mutations.

Example 8 Analysis of the combined pA- mutation and the pG- mutations (M428S/N434S/H435R/Y436H)

Substitution of residues to abrogate both pA binding and pG binding were introduced into the wild type (WT) human IgGl-Fc sequence. This combination is equivalent to introducing the pG- changes M428S/N434/Y436H in an IgG3-Fc, as this has arginine (R) at position 435. The pTT5 vector containing the WT IgG-Fc sequence, as described previously, was used as a template to introduce the substitutions with the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies). The following primers, with codon changes in bold type, were used: Forward (SEQ ID NO: 68)

5 ' -TGCTCCGTGAGCCATGAGGCTCTGCACAGCCGCCAC AC ACAGAAGAGCCTCTCCC-3 '

Reverse (SEQ ID NO: 69)

5 ' -GGGAGAGGCTCTTCTGTGTGTGGCGGCTGTGCAGAGCCTC ATGGCTCACGGAGC A-3 '

Transfection into HEK2936E was carried out and the expressed protein characterised for binding to protein A and G (Figure 38). The M428S/N434S/H435R/Y436H IgGl-Fc expressed at a similar level to WT and the pA- and pG- proteins as judged on SDS-PAGE analysis of the supernatant (Fig. 38A). The M428S/N434S/H435R/Y436H IgGl-Fc did not bind to pA nor pG affinity matrix as shown in Figure 38B. Biolayer interferometry analysis showed that binding to anti-Fc coated tips was equivalent for wild-type and M428S/N434S/H435R/Y436H IgGl-Fc HEK supernatants (Fig. 38C). The data show that the combination of mutations result in a pA-/pG- IgG-Fc phenotype.